tierlist.tech
中EN

09.24 Today's Picks

405 new projects found today, 254 rated.

Created 2026-09-22 – 2026-09-24Fetched 600Filtered 50Rated 41

A · Worth watching

A

cleanupper

SewCabinSpoutJavaScript

Solves the daily pain of macOS developers' disks being eaten by Xcode DerivedData, node_modules and Homebrew/pip caches, terminal-native, trash-first and zero telemetry, more trustworthy than CleanMyMac, and naturally distributable via Homebrew and social word of mouth. Commercialization could be a Pro rule library or team device cleanup, with the risk that Apple's built-in storage management and open-source peers like Mole already hold the space, and the anomalous 700+ stars in three days needs watching.

🧹 Free, open-source macOS disk cleanup CLI. Scan & safely remove caches, logs, Xcode DerivedData, npm/Homebrew/pip junk and stale node_modules from your terminal. Trash-first, zero telemetry the termi…

#macOS#CLI#Productivity

★723

A

vellum.nvim

blackhat-7Lua

Renders a GitHub-style Markdown preview beside Neovim buffers and draws mermaid diagrams as real images in the terminal. nvim users writing docs are missing exactly this, and it adds mermaid over markdown-preview.nvim; stability remains to be verified.

Live, beautiful GitHub-markdown preview beside your buffer in Neovim — mermaid diagrams as real images, in the terminal.

#Editors#Plugins#Documents#Terminal

★37

A

magpie

yetoneGo

This menu-bar tool unifies model backends across agents (Codex on DeepSeek, Claude Code on Kimi), solving multi-subscription, multi-key switching and vendor lock-in pain. Author yetone has developer influence and it started fast; it could add team config hosting and gateway value, but faces established routing layers like LiteLLM and claude-code-router and must differentiate on experience.

Every agent's model. One place. Codex on DeepSeek, Claude Code on Kimi, from the menu bar.

#AI#LLMs#CLI#AI Agents

★55

A

cc-usage

haishishushuRust

It aggregates quota, tokens and spending from coding assistants like Claude and Codex into a cross-platform desktop app, an immediately useful need for heavy users. The scope is clear; the hard part is keeping up with each platform's API, and several similar CLI tools already exist, so it's a race on maintenance speed.

A cross-platform desktop app for monitoring AI coding assistant usage, giving a unified view of quotas, tokens, costs, and request logs across Claude, Codex, and other platforms.

#LLMs#Monitoring#macOS

★28

A

grit

DasDarkiC++

A Godot GDScript compiler written in C++, aimed squarely at the performance limits of interpreted GDScript; the direction is rare and the target concrete. It is still an early implementation, so do not expect to drop it into production.

GDScript Compiler for Godot.

#Game Dev#Build#Performance#godot

★30

A

shapeshift

anishfnTypeScript

A single input box that morphs by semantics into the right control—date picker, slider, selector—an experiment in form and natural-language interaction paradigms that, if made solid, could become a general input component for agent UIs and be licensed as a design system or SDK. Risks: heavy dependence on the undisclosed Jev model, offline promises and recognition accuracy, plus a strong gimmick factor, and mainstream design systems may simply build it in.

An input that becomes what you mean: one text box that morphs into the right UI as you type. Powered by TypeSafe Jev, works offline.

#AI#UI Design#Web Frontend#Prompting

★404

A

oink

fabriziosestitoRust

A Rust text-adventure (gamebook) engine for e-ink devices, small in scope and clearly defined, with rendering and page turns evidently designed around e-ink traits. It suits developers wanting to write interactive fiction on Kindle-like devices; for now it lacks sample books and docs, so watch whether it can run a full work.

A librogame (gamebook) engine for e-ink devices.

#Games#Game Dev#Linux

★11

A

PianoStudio

drJhonatan00TypeScript

A browser piano workbench that turns sheet-music images into playable sequences, with real samples and an optional OMR pipeline. The barrier to trying it is low; OMR accuracy and sheet-music copyright are its ceiling.

Piano Studio is a responsive browser-based piano workspace that turns a sheet-music image into a playable sequence. It combines a polished virtual piano interface, real acoustic piano samples, local f…

#Audio#Images#Web Frontend#Learning

★37

A

ha-layers

lukab-devPython

Adds a priority layer to Home Assistant lighting, solving the old problem of scenes and manual adjustments overriding each other, with simple installation and clear scope. Immediately useful for heavy HA users, though it's just scheduling logic and differs little from existing blueprint approaches.

Priority layers for Home Assistant lights

#IoT#Self-Hosted#home-assistant

★23

A

sentinel

guiIhermevieiraTypeScript

It turns anti-money-laundering transaction monitoring into a NestJS service plus a separately installable rules engine, with idempotent ingestion, case management and audit logs. Open-source compliance tools are rare and useful to fintech backends; but whether the rules match real regulatory practice can't be proven by the README, so check the tests first.

Open-source AML transaction monitoring: a NestJS service and a framework-agnostic rules engine on npm (@sentinel-aml). Idempotent ingestion, case management, audit log, runtime rule tuning.

#Fintech#APIs#SDKs#Self-Hosted

★11

A

myDATA

palavogikos-collabPython

A Claude skill that issues Greek AADE e-invoices with MARK, saves a JSON per invoice, and syncs to ERP — an extremely narrow niche but a real compliance workflow. Teams doing business in Greece and connecting to SAP B1 can use it directly; the limitation is ERP support for only one vendor, with the rest left to the community.

Issue Greek invoices via AADE timologio, get the MARK, keep one JSON per invoice, sync to your ERP (SAP B1 today; SoftOne, Pylon, Megasoft wanted). Claude skill, not a ΥΠΑΗΕΣ.

#Fintech#E-commerce#AI Agents#Agent Skills

★27

A

rooms

saragordicSwift

On macOS, saves each project's window layout as a "room" and switches the whole set with one shortcut, so you no longer shuffle windows when changing projects. The description is restrained and the scope clear, and 23 stars within hours of creation shows the pain is real; the limitation is macOS-only and a single feature.

Switch between projects on your Mac. Every project is a room: its windows, laid out, one shortcut away.

#macOS#Productivity#UI Design#Automation

★23

A

liteagents

BerriAIPython

A provider-agnostic agent SDK from BerriAI that follows the Claude Agent SDK's query() interface and can switch models per turn; it's a low-cost abstraction for teams already on LiteLLM, with the limitation that the interface tracks upstream and the ecosystem is still shallow.

A provider-independent agent SDK with the same query() interface as the Claude Agent SDK, allowing you to use the right model for every turn.

#AI Agents#SDKs#LLMs#APIs

★9

A

agent-message-board

fstandhartingerPython

It lets multiple coding agents share context through a SQLite message board, with humans watching via a read-only web page; the problem is real, the implementation restrained, and how to run it is obvious at a glance. The limit is that it's only a collaboration patch competing with MCP and shared workspaces in an undecided ecosystem.

Shared SQLite message board for AI coding agents, with a read-only human web UI

#AI Agents#Collaboration#Databases#Self-Hosted

★11

A

talking-head-reel

mariagorskikhTypeScript

A Claude Code skill that turns casual talking-head phone videos into vertical clips, including selection, subtitles, and overlays — a concrete niche with real demand. 40 stars in a day shows the pain point exists, but end-to-end editing like this is sensitive to ffmpeg and model calls, and results are unverified.

Claude Code skill: turn a messy phone recording into an edited vertical reel (best takes, captions, animated overlays, memes)

#Agent Skills#Video Editing#LLMs#ffmpeg

★40

A

NiloPathTracing

ColinLeung-NiloCatLanguage unknown

Adds RTX path tracing and DLSS Ray Reconstruction to Unity URP with one-click setup and NiloToon compatibility; Unity creators doing toon rendering benefit directly, with the limitation of being tied to RTX GPUs and the URP pipeline.

RTX Path Tracing + DLSS Ray Reconstruction for Unity URP. One-click setup, native NiloToon support.

#Game Dev#3D#GPU#unity

★9

A

nafath-php

hamada-emam-techPHP

It wraps Saudi Nafath identity verification's web redirect and app push into a PHP library and documents pitfalls the official docs miss, saving teams doing Middle East logins a few days. The scope is narrow and the description restrained; the downside is serving a single flow in one country, with long-term upkeep on the author.

Nafath identity verification for PHP — web redirect and app-push, with the traps the official docs leave out already handled.

#SDKs#APIs#Privacy

★10

A

OpenWire

AdrescorGitiRust

A native Linux soundboard and effects processor on PipeWire, filling exactly the gap left by Windows-only tools like Soundpad; streamers and voice-chat users can install and use it, with the limitation that it only covers PipeWire desktops.

Native Linux soundpad and effects processor for PipeWire.

#Audio#Linux#Streaming

★15

A

expo-motion-blur

rit3zhTypeScript

Native motion blur for React Native views, implemented on both iOS and Android, a gap the RN ecosystem has long had. 18 stars in half a day shows people were waiting; worth checking on a real device, though it does only this one thing and nothing else.

💨 Native motion blur for React Native views on iOS and Android

#iOS#Android#Animation#UI Design

★18

A

SafariKeePassXC

mgriesslehnerSwift

Uses KeePassXC's own browser integration protocol to fill the gap Safari officially lacks, directly benefiting people who stick with Safari but won't give up a self-hosted password vault. Swift-only and just getting started, so you'll have to verify stability yourself, but the direction is right and the scope restrained.

Safari Web Extension that connects to KeePassXC via its browser integration protocol

#Secrets#Browser Extensions#macOS#Privacy#keepassxc

★8

A

ImageForge

LyraVoidTypeScript

It analyzes, patches, repackages and verifies Android images entirely in the browser, with no upload and no flashing, and the boundaries are clearly stated. Those modifying boot images can use it as a lightweight alternative, though it only handles the image files themselves.

Universal Android image patcher that runs entirely in your browser: analyze, patch, repack, verify and export. No server, no upload, and it never flashes a device.

#Android#WebAssembly#Images

★13

That's today. The other tiers and the boards follow.

BHas a spark55
  • B NotProton It brings a Steam Play/Proton-style compatibility layer to macOS so Mac users can run their Windows game libraries, a long-unsolved and highly shareable pain point, starting fast with 52 stars in 0.9 days. It could go subscription or partner with publishers, but the biggest uncertainty is facing Apple GPTK, CrossOver, Whisky and Valve's stance, with high compatibility and legal risk. #Game Porting#macOS#Games

    Steam Play for macOS

    NotProtonNotSwift★52

  • B 3dicon Generates looping 3D icons with real alpha transparency from a single prompt, directly cutting the asset cost of product landing pages and app icons—transparency is exactly where most AI image tools fall short. 134 stars in 0.2 days shows developers readily accept "skill as tool." Icon-pack subscriptions or a design plugin could monetize; the risk is official AI features from Figma, Spline and others swallowing it. #Agent Skills#3D#Animation#AI

    One prompt in, a looping animated 3D icon out — with real transparency. A Claude Code skill.

    samyost1Python★134

  • B cogsend A self-hosted social media scheduling tool that runs on the user's own Cloudflare account, solving expensive subscriptions, account bans and data leakage—attractive to indie developers and agencies, with 112 stars in 1.9 days from self-hosted community spread. A hosted version and team collaboration could be paid; the risk is that open-source peers like Postiz and Mixpost are ahead, and platform API policies are tightening. #Self-Hosted#Social#Automation#cloudflare

    Self-hosted social media scheduler. Write once, publish to any connected platform. Runs on your own Cloudflare account.

    deepaknessTypeScript★112

  • B FoxDevStudio Provides a modern IDE and 64-bit runtime for Visual FoxPro 9, directly solving the problem that many legacy enterprise systems cannot move to 64-bit and lack modern tooling; such users have strong willingness to pay and few alternatives, and 74 stars in 1.2 days shows the old ecosystem still has heat. The risk is the group naturally shrinking, Microsoft's migration narrative, and the huge compatibility workload for language and runtime. #Editors#Windows#SDKs

    A modern IDE and 64-bit runtime for Visual FoxPro 9 applications.

    FoxDevCommunityRust★74

  • B paragravity Provides non-invasive multi-account isolation and sandboxing for Google Antigravity, solving the problem of developers running multiple agents in parallel with state bleeding between them; it rises naturally with Antigravity's popularity and could extend into a general IDE/agent multi-instance manager charged per seat. Risks: tight binding to a single platform, so an official multi-workspace feature would kill it, and multi-account strategies may violate ToS. #AI Agents#Automation#antigravity

    Native, non-invasive parallel multi-account & sandbox manager for Google Antigravity

    edison-landPython★188

  • B ai-engineering-course A free structured curriculum spanning machine learning, Transformers, LLM fine-tuning, RAG, agents, evaluation and safety, filling the gap in systematic AI engineering teaching material; the author has reach in developer communities, and 75 stars in 0.7 days shows steady growth. Monetization comes from course and corporate training funnels; the risk is content decay requiring constant updates, with DeepLearning.AI and others already occupying the space. #Learning#LLMs#RAG#AI Agents

    AI Engineering Course - A free and complete AI Engineering Course to learn AI Engineering step by step - from Machine Learning, Neural Networks, and Transformers to LLMs, Fine-Tuning, RAG, AI Agents, …

    amitshekhariitbhuMarkdown★75

  • B threejs-architecture-effects Packages the know-how for generating interactive self-assembling Three.js buildings into an agent skill, letting coding agents directly produce usable 3D scene code—hitting the gap where agents lack domain skill packs, and 175 stars in 1.9 days shows skill distribution has momentum. Monetization could be a skill marketplace or subscription; the risk is skills being easily copied or absorbed into model vendors' official skill libraries. #Agent Skills#AI Agents#3D#three.js

    An Agent Skill for building interactive, self-assembling 3D architecture with Three.js. 一个用 Three.js 生成可交互、可动态建造 3D 建筑的 Agent Skill。

    lhlGitHubTypeScript★175

  • B dpapi-toolkit Drop in any Windows DPAPI artifact and it identifies the format and points out the required master key; give it the key and it decrypts offline, with a CLI and web interface. It saves work for forensics and incident response; compared with dpapick it is more oriented to single-file investigation, and its coverage still needs testing. #Windows#Secrets#Reverse Engineering#CLI

    Drop any Windows DPAPI artifact and it identifies the format and the exact master key it needs, then decrypts once you supply the key. Offline, CLI + web UI.

    crypt0p3gPython★10

  • B skipper A git MCP service for agents: local repos are read in-process via gitoxide, while remotes cover GitHub, GitLab, Gitea, and Forgejo by reusing each platform's CLI. It saves you from piling up gh calls yourself; the limitation is that it depends on each platform's CLI already being logged in. #MCP#AI Agents#APIs#git

    MCP server for git repositories: local git in-process via gitoxide, plus GitHub, GitLab, Gitea and Forgejo through their CLIs

    seurosRust★13

  • B YuE2-Studio Packages YuE2 as a native Windows app that generates full songs with vocals locally, and can also edit scores and replay precisely. It removes the Python environment hurdle, greatly lowering the barrier for people who want to play with music generation on their own machine; the limitation is GPU demand, and output quality is up to the model. #Audio#AI#Windows

    Local AI song generator with an editable score — YuE2 on your GPU: full songs with vocals, sheet music, covers, exact replay. Native Windows app, no Python, installer with auto-update.

    timoncoolTypeScript★47

  • B gut Adds an LLM judgment branch to ordinary Elixir control flow, treating model output as one condition in an if; the scope is restrained and the claims are not exaggerated. Useful for people writing BEAM services who want to embed a single classification in business code; how retries and cost are handled is not yet explained, and both scale and reliability remain to be verified. #LLMs#SDKs#APIs

    Use LLM judgment in regular Elixir control flow.

    vinibrslElixir★7

  • B AstrLink A local AI gateway that unifies routing across users' subscriptions and APIs with an on-device privacy emphasis, solving agents filling in keys everywhere and picking models per task. The author maintains New API, bringing ready distribution and engineering credibility, and 54 stars in 1.8 days is steady growth. Monetization is enterprise gateway and quota management, with the risk of direct overlap and severe homogenization against LiteLLM and one-api family products. #AI#LLMs#APIs#Self-Hosted#Privacy

    A local AI gateway for AI agents, unifying your subscriptions and API providers with smart routing and on-device privacy protection.

    Calcium-IonGo★54

  • B llm-multi-agent-game-evaluation It tests LLM strategic reasoning, deception and social norms with reproducible game environments such as RPS, Hawk-Dove and Liar's Bar. But such game benchmarks are already common, and no actual results are given. #LLMs#AI Agents#Benchmarks

    A dynamic multi-agent game framework for evaluating LLM strategic reasoning, communication-driven decision-making, deception, and social-norm behavior across RPS, Hawk-Dove, and Liars Bar environments…

    DaVVinCiPython★20

  • B bendc A self-hosting compiler for Bend 2 written in Bend 2, a rare and hardcore goal. It is a useful reference for language-implementation enthusiasts, but neither completeness nor practicality is there yet. #Build#Performance#Learning

    A self-hosting compiler for Bend 2, written in Bend 2.

    LulzxC★20

  • B qu-ai-wei Merges three sets of Chinese writing rules for "removing AI flavor" and resolves seven places where they contradict each other, packaged as a Claude skill. People who need to polish Chinese drafts can use it directly; the limitation is that the rules come from empirical corpora, so results vary by genre. #Prompting#Content Writing#Agent Skills#AI

    A Claude skill for stripping AI tells from Chinese drafts.

    baibanbaoLanguage unknown★33

  • B codex-routing-detector It probes which model actually answers behind a Codex request, to catch silent swaps. That's practical for users who suspect downgrades and hits an industry black box, but the decision basis and false-positive rate still need checking in the code. #LLMs#Debugging#Monitoring

    Find out which model actually answers your Codex requests (detects silent model substitution, e.g. gpt-6-astra served by gpt-5.6-luna)

    darkdarkcocoaPython★10

  • B hetzner-cloud-audit-skills It packages Hetzner Cloud attack paths, config drift and cost auditing into an agent skill, requiring evidence-backed conclusions, with a narrow and clearly defined scope. It suits small teams on Hetzner for periodic checkups; it depends on an agent host and covers a limited set of resource types. #DevOps#App Security#Agent Skills#hetzner

    Evidence-backed attack paths, drift, and cost auditing for Hetzner Cloud.

    jpolecPython★7

  • B windows-disk-cleaner-skill Brings WizTree-like fast NTFS/ReFS scanning to Windows disk cleanup and packages it as an agent-callable skill. People who want to script cleanup on Windows benefit directly, and 22 stars in a day shows demand; the limitation is that it only covers those two filesystems. #Windows#Agent Skills#Productivity#AI Agents

    A Windows disk cleanup Skill based on Wiztree-like fast NTFS/ReFS scanning

    std-microblockRust★22

  • B deepreseach-learing A LangGraph deep-research agent emphasizing resumable interruptions, evidence management, and cited reports, with engineering more complete than a demo; but such projects are everywhere, the author positions it as a learning reference, and the incremental value is limited. #AI Agents#LLMs#Learning

    Recoverable and auditable deep-research agent built with LangGraph. It plans research, searches the web, manages evidence and context, resumes interrupted runs, and generates cited reports—also design…

    1612535983Python★15

  • B tpu-megakernels A collection of TPU megakernels compiled by Inferact, a genuinely cutting-edge direction. But the repository has only one line of explanation, with no kernel list, performance numbers or usage, so outsiders gain nothing. #Performance#ML Research#Hardware

    A collection of megakernels for TPU

    InferactPython★25

  • B ClaudeAnimationBase A starter kit that turns the process of animating hand-drawn Claude cartoons with p5.js and p5.brush into a template, including 31 emotions and a prompt guide. It is closer to a template-plus-prompt collection, and 28 stars in 0.1 days seems fast. #Animation#AI#LLMs#Web Frontend

    A starter kit for animating hand-painted cartoons with Claude: p5.js + p5.brush, the Clawd character, 31 acted emotions and a guide for the model

    JohnHeibelJavaScript★28

  • B jianying-automation Turns AI-generated storyboard assets into directly editable JianYing drafts, hitting a real short-video pain point: easy to generate, hard to get into editing software. The toolchain is concrete and the license is non-commercial; pyJianYingDraft already exists, so its edge depends on how long draft-format compatibility holds. #Video Editing#Video#AI#capcut

    Jianying (CapCut) automated editing pipeline: AI storyboard assets -> editable Jianying drafts. cutcli+capcut-cli, assets keep their original names and full original audio, non-commercial attribution …

    zhihe4869Python★11

  • B qwen-image-2.1-studio A local Qwen-Image studio for Apple Silicon with a CLI and web UI, addressing the lack of CUDA on Macs. But Mac local image generation already has mature options like mflux and Draw Things, and a one-day-old repo hasn't proven it runs reliably. #Image Generation#macOS#CLI

    Local Qwen-Image-2.1 studio for Apple Silicon: text-to-image, multi-image editing and prompt enhancing, with a CLI and a helmstudio web UI.

    janisharPython★11

  • B drizzle A Drizzle integration module from the official NestJS org, giving teams already on Nest a maintained option for replacing TypeORM/Prisma. Little novelty; its value is entirely in the words 'officially maintained,' and the API design is worth a quick look. #Databases#APIs#nestjs

    Drizzle module for Nest framework (node.js) 🗄️

    nestjsTypeScript★16

  • B NinjaMesher A generator that directly writes OpenFOAM polyMesh hexahedral grids via right-angle cuts, replacing the most tedious step in CFD preprocessing. The description is precise and the scope restrained, so OpenFOAM users can judge fit in ten minutes; the limitation is that it only covers this one mesh type. #Science#3D#openfoam

    Hex-dominant Cartesian cut-cell mesher for CFD, writing OpenFOAM polyMesh directly

    nicolasbadanoC++★7

  • B jev-browser Lets Jev decide control actions while Playwright executes, doing indexed browser automation to reduce the instability of agents operating web pages. The direction is right but the space is crowded (Playwright MCP, browser-use, Stagehand), 77 stars in 0.7 days is modest momentum, and it's tightly bound to the single CodexQA/Jev stack. Monetization is limited to enterprise automation services, and Playwright already holds the standard slot. #AI Agents#Automation#Testing

    Jev Browser — indexed browser automation. Jev chooses the control, Playwright acts. A CodexQA skill.

    openqa-cnTypeScript★77

  • B PrettyPitch Estimates F0 directly from singing voice without relying on scores, note onsets, or pitch annotations, a rare angle for people working on singing synthesis, transcription, and MIR research. The repo covers only this one thing and is quite restrained; but accuracy, datasets, and reproducibility are unverified, so for now it is more like a code appendix to a paper. #Speech#Audio#ML Research

    A singing F0 model trained without scores — no note onsets, no note pitches, no reference score

    wavtechyukkyPython★13

  • B termirai A remote management and AI ops platform for multi-device terminals, solving unified management of scattered server sessions and letting agents execute commands, but the track is crowded (Warp, Termius, tmux setups, various SSH MCPs). 52 stars in 0.5 days is middling momentum; commercialization can be team seats, with the risk of heavy security responsibility and high trust barriers, making it hard to become a standard in the gaps between giants. #Terminal#DevOps#AI Agents

    Multi-device terminal remote management and AI operation platform.

    RaysunKRTypeScript★52

  • B quiron A 'de-AI-ify' skill for Claude Code and Cursor that sells itself on quantifying rewrite results against a human baseline, somewhat more honest than slogan-style peers. But such prompt-layer tools depend heavily on context and many similar ones exist, so long-term differentiation is hard. #Agent Skills#Content Writing#LLMs

    AI humanizer skill for Claude Code, Codex and Cursor. Removes AI slop and the signs of AI writing, then measures the rewrite against human baselines.

    ilien-devPython★20

  • B DataFlow-AgentMM A framework from the DataFlow lineage for collecting and evaluating multimodal agent trajectories, a real direction. But the description stops at "framework," without saying how to install it or which visual environments it supports, and agent evaluation frameworks are already numerous, so the difference is unclear. #AI Agents#Multimodal#Benchmarks

    A framework for collecting and evaluating multimodal agent trajectories in visual environments.

    OpenDCAIPython★18

  • B omacontra A five-level arcade boss rush on Linux and Hyprland, with dependencies and gameplay clearly written; it's a small but complete indie game. Whether it's worth it depends only on feel, and being tied to one window manager limits its audience, with many similar free action games around. #Games#Linux#hyprland

    Omacontra: standalone five-stage arcade boss rush for Linux and Hyprland

    gardnmiPython★14

  • B mu A coding agent that uses a small model as a judge to route work and a large model to do it, a sensible approach, and 49 stars in a day and a half shows people care about cost. But it is assembled on pi and AionUi, the coding agent space is already crowded, and there is no real comparison of results yet. #AI Agents#LLMs#CLI

    mu (μ): a coding agent that thinks before it acts. A small, fast judge makes the routine calls, the big model does the work. Built on pi and AionUi.

    qybaiheTypeScript★49

  • B roboto-physical-ai-actions Reviews robot-collected data and converts it into LeRobot v2.1/v3.0 datasets, with merging and version conversion. People building embodied-AI data pipelines can skip writing their own conversion scripts, and the scope is clear; the limitation is that it only covers the LeRobot ecosystem. #Robotics#Data Analysis#ML Research#lerobot

    Roboto actions for physical AI: curate and audit robotics data, convert it to LeRobot (v2.1, v3.0) datasets, merge, up- and down-convert them

    roboto-aiPython★8

  • B llm-typesafe Simon Willison's Jev/TypeSafe integration plugin for the llm command line; the code quality is probably reliable and the style as restrained as ever. But it is only a thin adapter, and whether it is worth a look depends entirely on whether you have decided to use that model. #LLMs#Plugins#CLI

    LLM plugin for accessing Jev and other TypeSafe AI models

    simonwPython★12

  • B typist Converts photos into text art in the browser that can be pasted into Instagram or Telegram, with no install and a clear purpose, a clean little tool. But character-art generators already have many implementations, and it offers no better algorithm or new twist, so it is useful but replaceable. #Images#Web Frontend#Social

    Turn any photo into text art you can paste into Instagram, X and Telegram. Runs in your browser.

    winchxyzJavaScript★13

  • B yanwai An LSPosed module that shows emotion probabilities, subtext and reply suggestions in WeChat chats. The angle is specific and rare, but it relies on root/Xposed and a specific WeChat version, and the privacy cost is significant, so it's only for trying out. #Android#LLMs#Privacy#Social

    A subtext assistant for WeChat chats: an Android LSPosed / Xposed module that shows emotion probabilities, implied meanings and communication suggestions based on context.

    YIRC99Kotlin★11

  • B dfirswarm The goal is multi-agent collaboration for digital forensics and incident response, and recording each step in a hash chain is a good idea. But talk of 'letting agents build their own tools' drowns out deployment and supported data sources, and a one-day-old project can hardly be called auditable. #AI Agents#App Security#Self-Hosted

    A swarm of forensic agents, on your machine, on the record. Open-source agentic digital forensics and incident response under the AGPL: agents divide the case on a shared board, forge their own tools,…

    halilozturkciTypeScript★13

  • B pasame A near-field file transfer tool with no install needed on the receiving end. The scenario is not new, since LocalSend and Snapdrop already cover it, so it needs clearly better discovery or transfer experience to be worth readers' time. #Networking#Downloaders#CLI

    Pasá archivos a cualquier celular o computadora que esté cerca. Sin cable, sin cuentas, sin instalar nada del lado de quien recibe.

    KonixDevGo★9

  • B JevBystander An Android accessibility screen reader that interprets WeChat messages and only pops up three hints: intent, emotion and urgency, without generating or sending anything, showing good restraint. But the scenario is very narrow, the basis and accuracy of the interpretation are unstated, and privacy is sensitive. #Accessibility#Android#LLMs

    Android accessibility-based WeChat interpretation: screen reading only, only pops 3 Toasts (intent / emotion / urgency / suggestion), generates no reply text and sends nothing · zero third-party depen…

    Nisaka520Kotlin★7

  • B obsidian-taskflow It adds an action workbench layer to Obsidian Tasks to 'reduce decision cost', which genuinely appeals to people with piled-up tasks. But the Tasks ecosystem already has many plugins and templates, it looks more like an interaction redesign than a new solution, and a day-old project has no usage feedback yet. #Plugins#Productivity#Knowledge Base

    An action workbench for Obsidian Tasks that reduces decision cost and helps you start faster. 一个为 Obsidian Tasks 打造的行动工作台，减少决策成本，让你更快开始行动。

    ichris007TypeScript★8

  • B droidmem An Android process memory read/write library using a kernel module with a /proc/pid/mem fallback, with a concrete technical path. But such tools are almost only used to cheat in games or scrape data, and the docs don't cover compatible kernel versions or the anti-cheat cost, so think carefully before installing. #Android#Reverse Engineering#Debugging

    Android memory reader/writer — kernel module + /proc/pid/mem fallback

    ScxttVXC++★9

  • B PickleChart An open-source order-flow and liquidity terminal, a rarity since such tools are usually paid software. But the Windows XP skin, 3D trade bubbles and '60FPS' promotion overshadow any explanation of data definitions and latency, making it look more like a demo. #Fintech#Crypto#Web Frontend

    🥒 High-performance, open-source Order Flow & Liquidity Trading Terminal styled in nostalgic Windows XP Luna. Features 60 FPS canvas charts, 3D trade spheres, resting limit walls, drawing tools, and lo…

    ikariszTypeScript★10

  • B yard-sale A Wes Bos demo project: photograph secondhand goods to identify and price them, running on React plus Cloudflare Workers, good for learning how to write full-stack AI apps. But it is a teaching sample, not a product, the pricing itself is crude, and wrapper apps like this are a dime a dozen. #AI#Images#Web Frontend

    Yard Sale Gold — AI-powered yard sale scanner built with React and Cloudflare Workers

    wesbosTypeScript★38

  • B rewindDV-LAB A MiniDV capture and repair tool for Apple Silicon, releasing only alpha binaries and test reports, with no source. Digitizing old DV tapes is indeed a gap, but a closed-source alpha isn't suitable for entrusting precious footage. #Video#Hardware#macOS

    rewindDV LAB — native MiniDV preservation for Apple Silicon. Engineering alpha downloads, documentation and tester reports. No application source.

    bdubs426Language unknown★12

  • B go-jev The author is a trustworthy Go developer, but the content is just an SDK and CLI wrapped around TypeSafe Jev's typed decisions. Its value depends entirely on whether you've already chosen this model; the code itself offers no judgment worth reading, and 11 stars in half a day is just release noise. #SDKs#CLI#LLMs#APIs

    Go SDK and CLI for TypeSafe Jev: typed decisions (yes/no, choice, score) from a model

    mattnGo★11

  • B Kaghez A comic reader for GNOME written in Python, with an honest, understated description. But Komikku and similar readers already cover the need on Linux, so unless you are on GNOME and dislike the existing options, there is no reason to seek it out. #Linux#Images#gnome

    A manga reader for GNOME

    riyaniiiPython★13

  • B gpu-fan-control A script that controls fan speed by GPU temperature for a Corsair 1000D case with dual BOSS Storm3 controllers, written honestly and extremely narrow in scope. Only people with this exact hardware can use it, but the script is short enough to copy and tweak the curve directly. #Hardware#Linux#Monitoring

    GPU temperature-based fan control for Corsair 1000D + BOSS Storm3 (per-GPU curves, dual-controller)

    maoenping9Shell★8

  • B qwen-image-studio A local interface for running Qwen-Image on Apple Silicon, built on mflux plus ComfyUI's GGUF pipeline. It saves Mac users configuration work, but it is essentially a wrapper, and the description is a single Korean sentence. #Image Generation#AI#macOS#comfyui

    An image generation studio for running Qwen-Image-2.1 locally on Apple Silicon (Next.js + shadcn/ui, ComfyUI GGUF, mflux)

    revfactoryTypeScript★19

  • B GitComment It adds a message board and visitor messages to a personal homepage using GitHub Discussions, with nine themes. It is a decorative gimmick that is easy to build but of limited value. #Social#Web Frontend#github

    A comment section and Visitor Messages for your GitHub profile, powered by GitHub Discussions. Nine themes.

    LISK819129TypeScript★8

  • B llama-chat A native Android client for llama.cpp, with local inference that doesn't go through the cloud; the direction is right. But several mature clients already exist, the repo doesn't explain differentiating features like model management or background inference, and the star count doesn't show real usage. #Android#LLMs#Privacy

    Native Android client for llama.cpp

    haydenkzKotlin★7

  • B jev-guard Adds a scoring guardrail layer to Claude Code to block unauthorized changes and unrequested deployments — a direction aimed at a real pain point. But the description piles on terms like 'probabilistic scoring' and 'evidence checking' without saying how to install it or what it blocks, and similar hook approaches are already plentiful. #AI Agents#App Security#Plugins#claude-code

    Probability-scored guardrails for Claude Code: deny rule-breaking edits and unasked-for deploys, route your docs into each prompt, and check the final answer against the turn's own evidence.

    muratcakmakTypeScript★7

  • B ad-radar A browser extension that collapses Xiaohongshu, Weibo, and Zhihu feeds by keyword and blogger, then uses your own model key to judge ads and topics. The idea is practical, but results depend on prompts and your own wallet, and similar blocking extensions are already plentiful. #Browser Extensions#AI#Social#Privacy

    Open-source browser extension: collapse content by keyword and blogger on Xiaohongshu, Weibo, X, and Zhihu; use your own Jev API key to identify topics such as ads, AI, military, and politics.

    pengchujinJavaScript★11

  • B guoapp A short-drama client built with Flutter and Go, using Actions to automatically build Android, Windows and iOS versions. The multi-platform packaging is tidy, but the content source depends on third-party platforms that may fail at any time, and it treads in copyright gray areas. #Android#Windows#iOS#Video

    Hongguo Jian / Zhenguo Jian: a Flutter + Go multi-platform short-drama client, with GitHub Actions auto-building Android / Windows / iOS

    chenweitian423Dart★40

  • B functional-emotions-video A hand-drawn-style music video made with multi-agent division of labor, including a self-written GPU brush renderer and storyboards — interesting as a workflow showcase. But it's an artwork, not a reusable tool; the scripts and render pipeline have no standalone docs, so others can't take it away. #Video#GPU#Animation

    A painted music video for "Functional Emotions", made by Claude Opus 5.5: custom GPU brushstroke renderer, storyboard, and 7 parallel chapter agents.

    ledbetterljoshuaJavaScript★27

  • B MacOBlox Runs macOS Roblox on Linux via Darling — a rare approach, honestly stated in one sentence. But Roblox's client anti-cheat will almost certainly block it and Darling itself is incomplete, so it likely stops at the experimental stage: fine to watch, not to use as a tool. #Games#Linux#Game Porting#roblox

    Run the macOS Roblox client on Linux through Darling

    narezyObjective-C★9

COrdinary75
  • C jev-seo Input a homepage URL and it produces an SEO audit with PDF/XLSX/Markdown reports, automating routine audits and saving time for agencies. But only 66 stars in 2 days, and Screaming Frog, Ahrefs and Sitebulb are already mature, with the moat in crawlers and index data rather than models; commercialization is only a low-price alternative, unlikely to become a standard dependency. #Marketing#AI#LLMs

    Live SEO audit for any website from one homepage URL, judged by Jev. PDF, XLSX and Markdown reports.

    AgriciDanielPython★66

  • C SETS Claims to be a self-improving trading machine automating strategy search and execution, strongly appealing to retail traders; 195 stars in 1.1 days shows the topic brings its own traffic. But it lacks credible backtesting, risk control and performance disclosure, return attribution is very hard to verify, and it looks more like a marketing repo. Monetization can only be selling signals or subscriptions, facing mature platforms like QuantConnect and regulatory risk head-on, hard to become a standard dependency. #Crypto#Automation#Data Analysis

    Self Improving Trading Machine

    ShelpidJavaScript★195

  • C JIZURA A browser app that auto-generates text PVs from lyrics, a clear fit for fan and music creators by saving manual layout work. 163 stars in 0.8 days shows subculture distribution efficiency, but users pay little and the market is closed, while editors like CapCut are quickly building in similar effects—hard to monetize, hard to become a standard dependency, mostly one-off attention. #Video Editing#Animation#Plugins

    A browser app that automatically assembles text PVs from lyrics.

    852waJavaScript★163

  • C kryvora-node A reference client and validation worker for the Kryvora network—typical on-chain node software whose value depends entirely on whether the network has real usage and token incentives. 497 stars in 1.3 days looks more like airdrop/mining-expectation farming than developer demand. Only tokens and validation services can be monetized; homogeneous node implementations are plentiful, there's no lasting standard slot, and the risk is network shutdown or tighter regulation. #Crypto#Blockchain

    Reference client daemon and verification worker for Kryvora Network nodes.

    kryvora-networkGo★497

  • C warrenduffer Ranks the Nifty 50 with a model every 15 seconds and places orders via code, with a kill switch and intraday loss circuit breaker—engineering more solid than most retail bots. But excess returns can't be verified, and Indian broker APIs and SEBI regulatory risk are high. Monetization is subscription signals or proprietary trading; competitors are mature quant platforms, so becoming a standard dependency is unlikely. #Fintech#AI#Automation

    AI-driven intraday trading bot for Indian stocks. Jev ranks the Nifty 50 every 15s; code sizes each trade and places the stop; orders go live through Zerodha Kite or Kotak Neo. Day replay, kill switch…

    arimanyusTypeScript★76

  • C WarpLite A GPU-accelerated terminal and shell targeting the lag developers hit with large output, where switching willingness is high. But 222 stars in 0.5 days matching same-batch repos makes momentum dubious; Warp, Ghostty, Kitty and Alacritty already fill this slot, leaving little differentiation, and terminals are a category with strong free inertia, making team subscriptions hard. #Terminal#CLI#Performance

    Fast GPU-accelerated terminal and shell experience

    BufferHeraldLanguage unknown★222

  • C childstream Uses Windows child sessions plus Sunshine to run games on a second desktop and stream them, letting one machine game and work at the same time—a clever system-level hack. Valuable to users who want to save a machine, but setup is demanding, and GPU time-sharing may conflict with anti-cheat. No clear monetization, and the Moonlight/Sunshine ecosystem could build it in, so it's unlikely to become a dependency. #Streaming#Windows#Games#Performance

    Stream games from a second Windows desktop while you keep using your PC - via Windows child sessions + Sunshine

    mattxslvC#★54

  • C corz-client A Minecraft 1.21.11 client focused on building assistance, schematics and DonutSMP tools, sticky for that server's players. Its 268 stars reflect a highly active cheat/helper community rather than developer value; monetization is subscription or donation only, and it violates server ToS, risking bans and legal issues, so it cannot become a standard dependency. #Games#Game Cheats#minecraft

    A Minecraft 1.21.11 client focused on building tools, schematics, and DonutSMP utilities.

    zeldaboyzlixLanguage unknown★268

  • C clearwater Real-time photorealistic shallow-water rendering in a single HTML file, using WebGL2 with no dependencies or build, a useful reference for learning shaders and water simulation. 125 stars in 0.4 days is visual-virality growth, but it's only a demo with no library, API or maintenance promise, easily absorbed by the Three.js ecosystem and game engines, with essentially no commercialization path. #3D#Web Frontend#GPU

    Real-time, photoreal shallow water in a single HTML file. WebGL2, no libraries, no build step.

    AureliengmzHTML★125

  • C TauriKit Adds a modern UI component layer to Rust/Tauri desktop apps; the Tauri ecosystem has long lacked a mature component library, and if quality holds it could become a de facto standard monetized through enterprise components and support. But 223 stars in 0.5 days is highly similar to several same-batch repos with identical numbers and naming style, adoption evidence is insufficient, and official Tauri components plus egui and iced already squeeze the space. #UI Design#SDKs#Plugins

    Modern UI component kit for Rust desktop applications

    BridgeDruidCompressRust★223

  • C ghrandom It treats GitHub outages as a FreeBSD entropy source; the joke is clean and the implementation honest. But availability is near zero, making it a one-off novelty. #Operating Systems#Hardware#Networking

    GitHub's downtime as a FreeBSD entropy source.

    DominoTreeRust★10

  • C RepoLens It scores repositories purely from the frontend via the GitHub API with zero mocks, which is honest engineering. But such dashboards are already plentiful, and the scoring criteria are not explained. #Web Frontend#APIs#Data Analysis

    Live GitHub repository analyzer & scorer — real API data only, zero mock fallbacks, zero backend.

    MUdevelopsJavaScript★12

  • C Tesla_OS_On_QEMU Running the Tesla infotainment system inside QEMU is high-difficulty reverse engineering and virtualization research with unique value for automotive security researchers, riding the smart-car security话题; 62 stars in 1.8 days is niche-circle spread. But it can't be commercialized, sits in a clear legal gray zone, and Tesla can block it at any time via firmware updates, so it won't become a standard dependency. #Reverse Engineering#Embedded#Operating Systems#qemu

    Run Tesla OS On QEMU

    cn0xrootPython★62

  • C adobe-illustrator-menu A tool for research figure makers to assemble multi-panel figures in Illustrator; the need is real and fills a small pain point in journal submission. But the repo has only one sentence, not saying whether it's a JSX script or a CEP plugin, and no install steps, so usability can't be judged yet. #Plugins#Science#Images

    Adobe Illustrator tools for scientific multi-panel figure alignment, sizing, layout, and labels

    ephillips85-webLanguage unknown★9

  • C jev-auto-ime Asks a model whether the current input is Japanese or English to switch the macOS input method — an absurd idea, though the author honestly labels it personal use only. Using an LLM for something a few milliseconds of rules could handle makes latency and misjudgment hard to accept; more a fun experiment. #macOS#LLMs#Automation

    A tool that asks Jev whether the words you're typing are Japanese or English and switches the Mac input mode. Personal use only.

    kzkhykwPython★10

  • C cve-2026-87902-poc A PoC chaining WordPress path traversal to conditional RCE, with a pinned-version vulnerable environment, and the writeup is fairly standard. But public exploit code for a single CVE has a short shelf life and limited value to defenders. #Exploits#App Security#Network Security#wordpress

    PoC for CVE-2026-87902 — unauthenticated path traversal in WordPress page-template resolution (local PHP inclusion, conditional RCE) with a pinned vulnerable lab

    resslPython★20

  • C RustDeskPro Provides a self-hosted remote desktop client for teams, hitting enterprise needs for data staying on the intranet and per-seat licensing, with willingness to pay in the self-hosted community. But 221 stars in 0.5 days is suspiciously identical to repos in the same batch, casting doubt on real users; RustDesk itself is open source and widely self-hosted, and its free upstream version will keep suppressing such commercialization attempts. #Self-Hosted#Collaboration#Privacy

    Self-hosted remote desktop client for teams

    BehemothlyaoscillateRust★221

  • C CodexDesk A desktop companion for Codex-style coding agents that moves sessions, tasks and repo state from the terminal into a GUI; the need is real but the description is vague with no verifiable differentiation. 222 stars in 0.5 days matching same-batch repos suggests coordinated vote manipulation, and with no own model or protocol position it's most easily eclipsed by the official desktop app. #AI Agents#Productivity

    Desktop companion for AI coding agents and Codex workflows

    BinaryDelivererRust★222

  • C dingtalk-aicard Specs, skills and DWS integration for DingTalk AI cards, ready-made building blocks for developers inside the DingTalk ecosystem. But it is tied to a single platform, so people outside that ecosystem can barely use it. #AI#APIs#SDKs

    DingTalk AI Card specification, standalone Skill, and DWS integration.

    DingTalk-Real-AIPython★12

  • C elasticemail-examples Elastic Email's official API sample collection, covering multiple languages, webhooks and serverless. Anyone already using that provider can copy it directly, but it is essentially vendor documentation and offers nothing new. #Email#APIs#SDKs

    Elastic Email API examples: send transactional and bulk email, webhooks, inbound routing and contacts in 20 languages, frameworks and serverless platforms.

    ElasticEmailTypeScript★18

  • C Fyndra A Rust local file search for Linux, but fd, fzf, and plocate already exist; the description uses bold fancy text to stress "fast" without explaining the indexing approach or differences, and with 8 stars in two days there's no compelling reason to use it. #CLI#Linux#Productivity

    𝐅𝐲𝐧𝐝𝐫𝐚 — 𝐀 𝐟𝐚𝐬𝐭, 𝐥𝐨𝐜𝐚𝐥-𝐟𝐢𝐫𝐬𝐭 𝐟𝐢𝐥𝐞 𝐬𝐞𝐚𝐫𝐜𝐡 𝐭𝐨𝐨𝐥 𝐟𝐨𝐫 𝐋𝐢𝐧𝐮𝐱, 𝐛𝐮𝐢𝐥𝐭 𝐢𝐧 𝐑𝐮𝐬𝐭.

    0xPrince121Rust★8

  • C summer-cycle A Ghibli/Shinkai-style cycling mini-game built with Three.js, with scenes, textures and audio all generated in code. It's a likable visual demo, but the gameplay is thin, and new procedural scene demos like this appear every week. #Games#3D#Animation#Web Frontend

    A Studio Ghibli / Makoto Shinkai style cycling game in Three.js. Toon-shaded Japanese countryside with four times of day - every mesh, texture and sound generated in code.

    StarKnighttTypeScript★16

  • C dejevu A browser agent that looks at a page, calls a model once, and performs one action. The idea is simple, but the description builds its case by calling other demos 'slower', without explaining targeting accuracy or failure handling, which weakens credibility. #AI Agents#Browser Extensions#Automation

    Jev? Déjà vu. Browser agents that run on instinct, no Jev needed. One look at the page, one call to any open model, one action. Faster than the Jev demo on Google Flights.

    idovmamanePython★9

  • C t3craft Giving agents commands in Minecraft, getting pinged, and letting them build in your world is an interesting concept. But the description reads like a demo and doesn't say what to install or who it's for; it's a toy-level integration project, and ten minutes won't tell you whose problem it solves. #AI Agents#Games#minecraft

    T3 Code inside Minecraft: prompt agents, get pinged, and let them build in your world

    maxwellyoungJava★12

  • C riso-windowseat A single-file HTML project that makes procedural risograph-style short films, with a Claude Code skill and docs for replication, is a creative experiment and personal portfolio piece. It spread via visual impact, but has no productized interface or maintenance commitment, so its ceiling is low and the hype is one-off. #Animation#Agent Skills#Images

    Procedural risograph films in single HTML files (Window Seat, Roost and more), with the Claude Code skills, docs and tools to make your own.

    sevenevesaiHTML★143

  • C se3-livom-comfort Experimental LIVOM code with SE(3) constraints for the COMFORT localization benchmark, clearly a paper companion that only peers can reproduce. It has reference value for SLAM researchers, but ordinary developers can neither install nor use it, and there are no results or data notes. #Robotics#ML Research#Benchmarks

    SE(3)-LIVOM for the COMFORT Localization Benchmark

    gunheeShinC++★7

  • C livenerf Claims to be a benchmark tracking capability changes after model releases, but doesn't say which models it tests, what tasks it uses, or where results are. 30 stars in a day looks more like hype than substance, and the name doesn't match what it does. #Benchmarks#ML Research

    Benchmark for tracking model capability after release.

    ninjahawkPython★30

  • C blink Turns "one-pass typed decisions" into an embeddable C runtime with WASM support, an interesting idea that could save an LLM call when embedded in an app. But terms like "System One Model" are grandiose, and it doesn't say where the model comes from or how it is trained, so it is unusable today. #LLMs#WebAssembly#Embedded#AI

    An open-source, high-performance System One Model for one-pass typed decisions, with an embeddable C runtime and WebAssembly support.

    sqliteaiC★12

  • C REx-skill Turns the reverse-engineering workflow for binary vulnerability hunting into a series of agent skills, a clear direction. But it overlaps heavily with existing reverse-engineering automation tools, gives no hit rates or case studies, and at 10 stars it is unclear how it beats doing it by hand. #Reverse Engineering#AI Agents#Agent Skills#Exploits

    REx@Skill - Agentic Reverse Engineering eXecution Skill for binary vulnerability discovery

    tihanyinPython★10

  • C qwen-audio-studio A local web interface wrapped around qwen-audio speech synthesis, with the care taken to store keys in the macOS Keychain. But it is highly redundant with the batch of Qwen GUI projects, has a single function, and gives no reason it must be this one. #Audio#Speech#macOS

    Local web studio for qwen-audio-3.1-tts-next with secure macOS Keychain credentials and verified audio output

    tjxjTypeScript★7

  • C ClashDesk A cross-platform proxy workflow desktop client solving the cumbersome config and multi-subscription switching of Clash-family cores, with a large user base and mature payment habits. But 222 stars in 0.5 days is highly similar to repos in the same batch, suggesting vote manipulation; Clash Verge, Mihomo Party and FlClash already dominate, and homogenization is severe. Monetization relies on subscriptions and also carries compliance and distribution risk. #Networking#Windows

    Modern cross-platform desktop client for proxy workflows

    AutocratGirderRust★222

  • C ClaudePanel A desktop panel and toolbox for Claude and various AI assistants, trying to move model calls from the browser to an always-on desktop entry. Demand exists, but 221 stars in 0.5 days is identical to repos in the same batch with vague descriptions, making momentum credibility low. Monetization can only be an enhanced client subscription, while official desktop apps and many third-party wrappers are already extremely crowded, leaving no moat. #AI#Productivity

    Desktop panel and toolkit for Claude and AI assistants

    AvenueSnowStepRust★221

  • C ai-model-rankings It aggregates OpenRouter prices, speeds and benchmark scores into an ink-wash-style Streamlit dashboard. The data source is single and similar leaderboards are already plentiful. #Benchmarks#Data Analysis#LLMs

    Real-time AI model power rankings (水墨战力榜) — usage, benchmarks, price-performance & speed from OpenRouter, as an ink-wash Streamlit dashboard.

    Mocas-12JavaScript★9

  • C ml-lensvlm Paper code named LensVLM, about selective context expansion under compressed visual representations. But the organization name is easily confused with Apple's official one, and whether the code is the real official implementation is unverified, so don't treat it as authoritative yet. #ML Research#Multimodal#Images

    Official code for LensVLM: Selective Context Expansion for Compressed Visual Representation of Text.

    apple-aiml-researchPython★27

  • C spiralist This browser app turns photos into a continuous line drawing and records the drawing process, a visual toy that is zero-friction but far from developer needs. It spreads socially with no API or paid scenario, and risks being wiped out once design tools like Canva build in the same effect. #Images#Animation#Web Frontend

    Turn any photo into a drawing made of one continuous line, and film it being drawn. Runs in the browser.

    winchxyzJavaScript★52

  • C sys1 A Rust implementation of a System One-style decision model HTTP interface, and the engineering looks clean. But the "compatible API" points to a batch of newly appeared model names, so there is no ecosystem to be compatible with yet. #LLMs#APIs#SDKs

    System One compatible API for open decision models, e.g. Laya, written in Rust.

    alvarobarttRust★18

  • C game-apk-reverse-engineering Organizes the APK reverse-engineering workflow into AI skills and task templates, covering resource extraction, code recovery, and Unity asset import. It leans toward tutorials and process packaging, lacks directly runnable code, and its value depends on the toolchain you already use. #Reverse Engineering#Android#Agent Skills#Games

    A reusable workflow for APK reverse engineering, covering resource extraction, code recovery, data-table analysis, and Unity asset import, with AI skills and task templates.

    sayicLanguage unknown★9

  • C steam-live-charts A Streamlit dashboard that pulls live Steam player counts, top sellers and discounts in real time, simple and direct to implement. Its value depends on whether you want a self-built leaderboard; there is no new method. #Games#Data Analysis#Web Frontend

    Real-time Steam game leaderboards (最热游玩 · 热销 · 特惠 · 新品 · 免费游戏 · 限时免费) — live player counts, CNY prices & deals, auto-refresh every 60s.

    Mocas-12Python★10

  • C DIGI-F8748 Spanish-language steps for recovering the admin password on a DIGI ZTE F8748 router, read-only with no install. It's essentially a credential tool bypassing device access controls, with a very narrow audience and murky legal boundaries, so it's not recommended. #Hardware#Network Security#Reverse Engineering

    Recuperador de la contraseña de administrador para routers DIGI ZTE F8748 — sin instalar nada, solo lectura, paso a paso. macOS · Linux · Windows

    686f6c61Python★14

  • C tianzhi-core Uses classical texts as sources to turn the computable parts of Eastern divination into testable algorithms, more honest in attitude than most occult repos and acknowledging its own limits. But the output has almost no everyday use for developers; it is hobby research rather than a tool you can install and use. #Science#Learning#Knowledge Base

    Core algorithms of Eastern esoteric arts: based on classical texts, restoring the computable parts into testable algorithms

    zaoxu001Python★8

  • C SteamOS-ARM-SM8550 It ports SteamOS to SM8550 devices while calling itself the "official ARM version," wording that does not match the facts; 45 stars in 0.6 days is also abnormal, so check its authenticity before discussing usability. #Game Porting#Linux#Embedded#steam

    Official SteamOS ARM version, custom-adapted for SM8550 ARM devices.

    MaSieS4FunShell★45

  • C Intent-Router Compiles vague requests into a typed IntentSpec and hands it to a router or judgment model, which is an interesting idea. But it binds itself to the same-origin Jev/Laya repositories, making its independent value hard to verify. #AI Agents#LLMs#APIs

    Intent compiler for AI agents — converges vague requests into typed IntentSpec contracts (probe, ask, or halt before routing), the input layer for routers and typed-decision models like Jev & Laya

    angel291592Python★41

  • C snap It claims to turn unstructured state into typed decisions without generating any text, but the whole description is Jev/TypeSafe jargon, with no visible inputs, outputs or call method. Many mutually citing Jev repos appeared at the same time, looking more like joint promotion than an independent tool. #LLMs#CLI#APIs

    Typed decisions from unstructured state: one forward pass, zero generated text. Local, deterministic, Jev-compatible. Not affiliated with typesafe.ai.

    emnlmnRust★17

  • C jev_project_context A skill that gives coding agents long-term experimental memory, with plenty of concepts (evidence-first, decision-model layers) but no clear explanation of how to install it, where memory is stored, or how it fundamentally differs from ordinary context files. #AI Agents#Agent Skills#Knowledge Base

    Evidence-first long-term experiment memory skill for AI coding agents, with optional Jev decision-model layers

    poiuyjiePython★9

  • C laya-server A self-hosted API and web front end wrapped around a so-called Laya structured decision model, compatible with the TypeSafe Jev format; but it doesn't say what the model is or whose problem it solves, and team backing doesn't fill the information gap. #APIs#Self-Hosted#AI

    A self-hosted API and web interface for Laya’s structured decision models, compatible with the TypeSafe Jev API format.

    1Panel-devTypeScript★35

  • C jevk5 Claims to be an open-source weight replacement for TypeSafe Jev, giving typed judgments with probabilities in one forward pass. Without real weights and evaluation it's an empty promise, and it collides with dozens of same-named repositories from the same day, so credibility is doubtful. #LLMs#ML Research#Benchmarks

    JevK5: open-weight alternative to TypeSafe Jev. Typed decisions with probabilities in one forward pass; Apache-2.0 weights and code.

    allebeePython★42

  • C robo A self-described open-source agent for security work, running locally for reconnaissance and threat analysis, with chat and large-file reading. The feature list is broad but never clarifies the actual workflow or false-positive handling, the kind of 'does everything' description security agents most need to avoid. #AI Agents#App Security#Terminal

    Robo — an open-source AI agent for security work. Runs on your own machine: recon, threat analysis, and hands-on tasks in your terminal, desktop, or phone. Talks, listens, and reads files of any size.

    igniteenowPython★9

  • C ai-security-review Uses the BICO (behavior, identity, control, observability) framework to explain AI system security review. It reads more like a consulting template than a tool, with only a framework name and no executable checks or scripts, so it will likely be forgotten quickly. #App Security#LLMs#Monitoring

    Architecture-first AI security review using BICO: Behaviour, Identity, Controls, and Observability.

    Securing-AI-SystemsLanguage unknown★7

  • C solana-token-creator A page plus tutorial for issuing Solana tokens without writing code, with fairly specific parameters for Token-2022, permissions, and rent. But one-click token launchers are already everywhere and mostly tied to meme-coin projects, so it has limited reference value for serious developers. #Crypto#Blockchain#Learning

    Create a Solana token without code — SPL and Token-2022, authorities, metadata, rent, fees and liquidity. Practical guides.

    uni-launchHTML★24

  • C jev-judge-mcp It wraps a supposed Jev judgment model into MCP tools such as verify, screen and rerank. The name and API belong to a batch of Jev/Laya repositories that appeared within a day, with unclear origins; judgment capability, dependencies and cost are all unstated, so usability cannot be judged. #AI Agents#MCP#LLMs

    Typed judgment tools for MCP agents. TypeSafe's Jev model as verify, screen, find, classify, rerank, decide, compare, extract, review, gate, and score: the model judges, policy decides auto, review, o…

    PyModelPython★24

  • C genspark2api Wraps Genspark web sessions as an OpenAI-compatible API, with multi-account rotation and automatic registration. It is technically shallow, essentially bypassing the other side's account and payment limits, will break at any time, and should not be a production dependency. #Reverse Engineering#APIs#Automation

    OpenAI-compatible API bridge for the Genspark web session — multi-account round-robin, streaming, and automated signup toolkit. Unofficial; see DISCLAIMER.md.

    xinxinshuhao-createPython★11

  • C omarchy-dracula-reborn-theme It ports the Dracula color scheme into an Omarchy theme following the Base24 spec. It is a personal color port whose audience is limited to users of that distribution. #Linux#UI Design#Terminal

    Canonical Dracula colors interpreted into Base24 for Omarchy Quattro.

    OldJoboboCSS★11

  • C system1-agents Claims to use "System One decision models" such as Jev, Laya and Cua-S1 as an agent's brain for browser, computer and robot operation. The names and concepts all come from the same batch of short-lived repositories, with no weights, evaluation or reproduction steps. #AI Agents#LLMs#Robotics

    System 1 decision models (Jev, Laya, Cua-S1) as brain for agents: Browser use, computer use, games and robotics

    ThinkFlowLabPython★39

  • C CVE-2026-87902 Claims to be a universal PoC for an unauthenticated local file inclusion in WordPress core. The CVE number and year don't match, and the repository doesn't state affected versions, so it looks more like a demo script riding on a CVE number. #Exploits#App Security#Reverse Engineering

    CVE-2026-87902 - WordPress - WordPress Core - Critical 9.2 - Unauthenticated Local File Inclusion (conditional RCE)

    abraxasPython★24

  • C ponwrt The name suggests adding PON support to OpenWrt, but the README doesn't cover which ONTs are supported, how to flash it, or what the risks are. 61 stars in half a day looks more like reshare momentum than anyone actually flashing it successfully. #Networking#Embedded#Linux#openwrt

    pbs05C★61

  • C OmniJev Claims to be a multimodal limited-choice decision interface plus a MuJoCo workbench, with trajectory replay and a panel. The terminology is dense and the presentation demo-oriented, and after reading it you still don't know whose problem it solves. #Multimodal#AI#Robotics#Benchmarks

    OmniJev — multimodal finite-choice decision interface and MuJoCo embodied workbench: trajectory replays, decision probes, benchmark panels, 60s walkthrough.

    shapsiderPython★11

  • C DesignHub Promises logo variants, mockups, social images, a brand book PDF, fonts and icons, and design tokens all at once, trying to do everything. With no demo and no explanation of trade-offs, it is hard to believe all of it works a day after the repo was created; it looks more like a vision sketched against tools like Canva. #UI Design#Images#Documents

    Open-source, local-first design & brand toolkit. Build a brand once, then get logo variants, mockups, social assets, a brand book PDF and design tokens. Plus fonts, colors, icons, SVG & accessibility …

    yakew7TypeScript★10

  • C Valen It claims to train a Jev-style multimodal model, "System One plus vision," but gives no training details or results; the terminology comes first. #Multimodal#LLMs#ML Research

    Train a Jev-like multimodal model by yourself. System One Model, now with vision.

    Liuziyu77Python★19

  • C jev-harness A coding harness built around the so-called TypeSafe Jev model: the model proposes, Jev answers narrow questions, and code adjudicates and leaves a record. The division of labor is readable, but it depends on an unclear model and won't run without it. #AI Agents#LLMs#Code Quality

    A custom coding harness for TypeSafe AI's Jev: an LLM proposes, Jev answers narrow questions, code decides, every step leaves a receipt.

    TypeSafeAITypeScript★8

  • C LinuxScripts A collection of small Bash tools for Linux operations, self-described as safe and modular. Almost everyone has a set of scripts like this, and there's no reason to replace your existing toolbox. #Linux#Automation#CLI#DevOps

    A collection of practical, safe, and modular Bash utilities for Linux system administration, development, networking, storage, maintenance, hardware diagnostics, SSH, and everyday automation.

    ashishsinghboraShell★11

  • C personal-agent-open-source A single-user personal agent emphasizing managed tools and human-in-the-loop development, a reasonable direction but the description stays at the vision level. Personal agent frameworks have clustered in the past six months, and without saying which specific obstacle it solves, it's hard to tell when it will be usable. #AI Agents#Automation#Self-Hosted

    A single-user personal agent with governed tools and a human-in-the-loop development workflow.

    hensonzyw-gitPython★8

  • C podcast-deep-dive It turns a podcast episode into an in-depth document, with a one-sentence description that doesn't explain transcription, segment selection or citation handling. It sounds like a prompt workflow rather than a tool, and similar uses are easy to do with general models. #LLMs#Audio#Content Writing

    Turn a podcast episode into a deep-dive document

    imvanessaliLanguage unknown★7

  • C qwen21-fast-comfyui Japanese notes and workflows for running Qwen-Image on 12GB of VRAM, useful as a reference for users with the same setup, but it's really notes plus config rather than a tool. A content repo whose value depends on whether the author's environment can be reproduced. #Image Generation#Learning#comfyui

    A memo on running Qwen-Image-2.1 on an RTX 4070 12GB / with a DeepSeek-chan workflow included 🐳

    kuraneko1Python★7

  • C notion-skills A collection of 'skills' for Korean Notion users, mostly prompts and usage notes with no executable engineering content, and a narrow audience. A content repo, not a tool. #Knowledge Base#Productivity#Prompting

    알잘딱깔센 skills for Notion Korean users

    oswarldTypeScript★11

  • C OpusSkate A C++ skateboarding mini-game generated by Opus 5.5 that explains nothing beyond "AI can write games": no gameplay, controls or build instructions. Like the Three.js scenes from the same Opus batch, it is more a showcase than something worth ten minutes. #Games#3D#Game Dev

    C++ Skate Game by Opus 5.5

    OminousIndustriesC++★34

  • C Huawei-Cup-Mathematical-Modeling-Skill Turns the Huawei Cup modeling competition workflow—modeling, validation, writing, defense—into an agent skill, with an award-winning paper attached. It has some template value for contestants, but it is essentially a competition guide that is useless outside the contest. #Learning#AI#Science#Agent Skills

    Huawei Cup mathematical modeling skill: modeling, validation, writing and defense assistance based on 2024 Mathematical Modeling Star papers.

    Taron0323Language unknown★14

  • C laya-pilot Uses Playwright to generate replayable Excel test cases from actual pages, with judgment delegated to a local Laya. The idea is practical, but putting test assets in Excel and relying on an obscure decision API makes adoption costly. #Testing#AI Agents#Spreadsheets

    Generate replayable Excel test cases from live browser pages, or run existing Excel cases, using Playwright and local Laya or a compatible decision API.

    anthonyliJavaScript★18

  • C laya-jev-GraphRAG Puts Laya/Jev decision models into a four-stage GraphRAG pipeline, connecting four graph databases plus A* traversal. The stacking is obvious, but with no benchmark results it's hard to believe it runs end to end. #RAG#AI Agents#Databases

    Agentic GraphRAG engine using swappable System One decision models (local Laya / cloud Jev). Features a complete 4-phase pipeline (Ingestion, Pre-Retrieval, Traversal, Post-Retrieval) and evaluation a…

    bodepudimuneendra-netizenPython★22

  • C opus-5.5-system-prompt-components Breaks the leaked Opus 5.5 system prompt into readable modules. It has some reference value for people studying agent behavior, but it is second-hand leaked content whose source and completeness can't be guaranteed. #Prompting#LLMs#Knowledge Base

    Claude Opus 5.5 system prompt leak split into readable components (tools, behavior, agentic rules, memory). Reorganized from a public extract for study.

    cbrunnerLanguage unknown★30

  • C system-one A self-described provider-neutral System One runtime whose description never says what it schedules or how it connects to models, only conceptual terms. It shares origins with a batch of mutually citing Jev repos from the same period, and the star growth doesn't match the content. #LLMs#AI Agents#APIs

    Provider-neutral System One runtime for TypeScript and Pi

    iamaamirTypeScript★45

  • C Linux- A Russian-language Linux command cheat sheet, neatly organized but a pure content repo with no tools or scripts. It's fine for Russian-speaking beginners to bookmark, but adds no new information for developers who read the charts daily. #Linux#Learning#CLI

    Linux полная шпаргалка по командам на русском

    justxorLanguage unknown★15

  • C jev-guard A gateway for real-time risk triage of exchange deposits and withdrawals, using Jev for judgment and deterministic code for decisions — the split is reasonable. But the description revolves around Jev terminology without covering rule sources or false-positive handling, and alongside the author's other repos it reads like a coordinated promotion push. #Fintech#LLMs#APIs

    Real-time risk triage gateway for exchange deposits and withdrawals — Jev (TypeSafe System One) handles triage only; adjudication stays in deterministic code.

    klauswgJava★35

  • C openwrt-pon-drivers From the name, an OpenWrt PON driver, but the README doesn't say which optical modules and chips it supports or how to build it into firmware. Possibly useful for people tinkering with ONTs, but there isn't enough information to act on yet. #Networking#Embedded#Linux#openwrt

    pbs05C★15

  • C openwrt-pon-userspace The same author's userspace PON implementation, also without a description, so its boundary with the driver repo can't be determined. Four repos committed separately without explaining their relationship leaves readers guessing. #Networking#Embedded#Linux#openwrt

    pbs05Rust★18

  • C pix-golpe A demo-style comparison of "AI detecting Pix scams," running a Rust approach and a Python approach side by side. The output is more video material than a reusable detection tool, and you leave with nothing you can take away. #Scams#AI#Fintech

    Demo: IA detecta golpe do Pix e rastreia a quadrilha. Rust + Jev vs Python + DeepSeek em tela dividida.

    sandecoPython★13

DSkip103
  • D DockForge A dock and multi-panel layout system for desktop apps, solving the repeated reinvention of cross-platform window docking and panel arrangement. But 222 stars in 0.5 days and naming style appearing in groups with same-batch repos raise authenticity doubts, and Qt, Avalonia and Electron ecosystems already have mature layout solutions. Monetization can only be commercial licensing or enterprise customization, hard to settle into a standard dependency. #UI Design#SDKs

    Flexible dock and panel layout system for desktop apps

    BricklayerSurmountRust★222

  • D jev-engineering-zh This translates The Jev Engineering into Chinese, giving Chinese coding agent users a systematic methodology as a documentation-style传播 asset. It is not a product and can only funnel to courses or consulting; translation copyright, source changes and replacement by more authoritative texts are risks, so it cannot become a dependency. #AI Agents#Documents#LLMs

    Complete Chinese translation of "Jev Engineering: Made for Coding Agents" — preserves the original structure and 7 illustrations

    yibieLanguage unknown★99

  • D HelixEdit It claims a Rust-written modal editor with LSP support, a need already proven by Helix, Neovim and Zed, but the description nearly duplicates existing projects with no visible differentiation. 221 stars in 0.5 days matching same-batch repos makes momentum untrustworthy; as a tool it's hard to charge for, as a standard position there's no room, and the biggest risk is being seen as a Helix clone. #Editors#CLI

    Modal code editor with LSP support written in Rust

    CassowaryDevelopRust★221

  • D ZedLite A Rust code editor billed as lightweight and high-performance—performance and memory are real pain points, but this space already has strong projects like Zed and Helix, and the one-sentence description gives no way to judge whether there's a real core. 221 stars in 0.5 days matches the numbers of sibling repos, so momentum is likely fake. Editor distribution costs are extremely high, making both commercialization and standardization hard. #Editors#Performance

    Lightweight high-performance code editor built with Rust

    centralcashierboostLanguage unknown★221

  • D claude-opus-5-5-demo The repo has no README or description and is essentially demo code around Claude Opus 5.5, with no identifiable problem solved and no reusable API or signs of maintenance. Its 547 stars come from model-topic traffic rather than product capability; it can't be commercialized or depended on, and the risk is that value drops to zero once the model vendor publishes an official example. #Games#3D#Prompting

    claude-opus-5-5-demo

    riba2534JavaScript★547

  • D PDoomVideo It's just the source for one AI-generated music video, showcasing Claude Opus 5.5's generation ability, solving no developer problem with no API or reusable library. Its 345 stars come from model hype and visual virality, making it a one-off content asset that can't be commercialized or depended on, with attention likely to drop to zero as the model topic fades. #Video#Animation

    Source code for the Claude Opus 5.5 music video for I'm Upping My P(doom)

    JohnHeibelJavaScript★345

  • D turbo-kart-rally A Three.js kart game generated by five Opus sub-agents from a single prompt, with all assets procedural. It's interesting as a sample of "one-shot AI generation", but as a game it has no playable depth. #Games#3D#Web Frontend#Animation

    Arcade kart racer in the spirit of Mario Kart, built entirely with Three.js. 100% procedural assets, no build step. Built by five Claude Opus 5.5 sub-agents from one prompt.

    bridge-mindJavaScript★7

  • D decidr It gets typed decisions from a local model in a single forward pass via Ollama, skipping the generation loop and JSON parsing. The direction is right, but the wording matches other 'decision model' repos from the same batch, with no comparison data. #LLMs#Performance#SDKs

    Typed decisions from local LLMs in one forward pass, via Ollama. No generation loop, no JSON parsing.

    devanmolsharmaPython★7

  • D Swarmmy Claims to be a zero-dependency agent orchestration framework, but the description cuts off at 'intelligent' and never explains the core abstractions. Multi-agent frameworks are already oversupplied; without concrete scenarios and examples, few will try it. #AI Agents#Automation#LLMs

    A high-performance, zero-dependency framework for orchestrating intelligent

    loayabdalslamPython★12

  • D system-design-notes-zh A Chinese translation of English system design interview notes — pure content reposting with no tool or new organization. Convenient for interviewees, but the repo has no translation progress or licensing info, so its value depends entirely on the upstream. #Learning#Knowledge Base

    Chinese translation of System Design Interview notes (Chinese translation of liquidslr/system-design-notes)

    lukeTheNeuromancerLanguage unknown★15

  • D fiberhome-factory An undescribed Rust repo; the same author opened four PON/OpenWrt-related repos in one day with no indication of how they divide work or how to use them. Stars are rising fast but concentrated in the same batch, so check the flagship one first. #Networking#Embedded#Hardware#Linux

    pbs05Rust★12

  • D SDD-Agents Yet another spec-driven agent framework that writes specs before the agent executes. Kiro and spec-kit already cover this space, and the repo offers no differentiation or runnable entry point. #AI Agents#LLMs#Prompting

    Spec-Driven Development for Agentic App

    zekiriabdPython★8

  • D jevper Wraps a classification layer around OpenAI-compatible clients, returning probabilities and confidence instead of full text. The idea isn't bad, but 'Jev-shaped / TypeSafe System One' is invented jargon with no visible difference from ordinary structured output, and it shares origins with several same-named Jev repos. #LLMs#APIs#SDKs

    Jev-shaped (TypeSafe System One) classification wrapper over OpenAI-like clients: probabilities and confidence instead of prose

    zhulinchngPython★10

  • D lejudge-jev-jepa Aims to use a decision model instead of an LLM to judge natural-language constraints in JEPA world-model planning, which is an interesting direction; but the repo only has a conceptual description, with no data or examples, so it's impossible to tell whether it works. #ML Research#LLMs#AI Agents

    Natural-language constraints for JEPA world-model planning, judged by a decision model instead of an LLM.

    AbdelStarkPython★7

  • D Grounded-Action-Model A research project on 3D grounding for robots, with the text saying code will be released later. For now there is only a vision and nothing verifiable. #Robotics#ML Research#3D

    Grounded Action Model: 3D Grounding as a Foundation for Robotics (code coming soon)

    GehaoZhang6Language unknown★20

  • D tidewater A seaside town Three.js scene generated by Opus 5.5, with no gameplay or purpose. It's a one-off generation demo, and after watching it you still don't know what to do with it. #3D#Web Frontend#Game Dev

    Coastal town built with Opus 5.5

    dgreenheckJavaScript★18

  • D jev-suite A collection of four decision tools the same author piled onto Jev, none of which can be explained in one sentence. The scope is loose and depends on immature upstream concepts — a textbook platform-shaped shell. #LLMs#Fintech#APIs

    Four decision-quality tools on Jev (TypeSafe System One): Jev answers structured questions, deterministic code keeps the final say.

    klauswgJava★36

  • D rkn-guard The README holds only jokes like "star it or I'll bite your finger," with no visible functionality; the name hints at bypassing RKN blocking, and both the information and responsibility are unclear. #Network Security#Privacy#Networking

    Star it if you used it, or I'll bite your finger off.

    FlecksisGo★17

  • D research-pass It claims to be a community-validated AI-assisted research protocol, but there is no protocol text and no code, only a one-line positioning statement. #AI#Learning#Documents

    An open, community-tested protocol for rigorous AI-assisted research.

    MagicalDealerLanguage unknown★15

  • D stuntd A local proxy that learns the app's LLM judgments and answers with a Laya head model, compatible with Jev and OpenAI. The concept is convoluted, and it depends on the same batch of model repositories of unknown origin. #LLMs#Networking#AI Agents

    Local proxy that learns your app's typed LLM decisions and answers them with a Laya head. Jev and OpenAI compatible.

    bladedevoffPython★19

  • D avash An assembly project with no description, so it's unclear whether it's a shell, an assembler or an experiment. The author's reputation in the JS community may explain the initial stars, but the content itself offers nothing to evaluate. #CLI#Terminal

    keithamusAssembly★16

  • D vgi-autodev Exposes so-called Dev-0.4b typed decisions to DuckDB workers, but the description is all undefined terms with no query scenarios or return structures. It shares origins with the same batch of Jev repos and can't be evaluated yet. #LLMs#Databases#APIs

    VGI worker exposing Dev-0.4b typed decisions to DuckDB

    lmanganiPython★8

  • D venus-uni-comfyui A school's ComfyUI workflows uploaded as a package, of only marginal reference value to ComfyUI users. It is essentially a content collection rather than a tool, with no install instructions or maintenance commitment, no different from the workflow shares found everywhere in the community. #Image Generation#Images#Multimodal

    ComfyUI workflows from Venus University

    venus-uni-devLanguage unknown★13

  • D support-fins-zh A Simplified Chinese localization of an upstream project's interface, adding nothing beyond translation, and claiming to preserve the full upstream history, so it is more like a mirror. Users are better off installing upstream and waiting for official i18n; maintaining it separately risks drifting from upstream. #Documents#Self-Hosted#Collaboration

    Support Fins Simplified Chinese localization — zh-CN UI translation of upstream gittrahan/support-fins (MIT, includes the full upstream commit history)

    xwnxzzJavaScript★10

  • D wechat-jev-assistant Reads Windows WeChat records locally, redacts and classifies them; the privacy angle is somewhat interesting, but "TypeSafe Jev" is a made-up concept with no stated basis for its judgments. Several repos with the same Jev name have clustered in a short time, looking more like coordinated star-farming, and should not be taken seriously. #Privacy#LLMs#Data Analysis

    Windows WeChat conversation analysis assistant: local reading, anonymization, TypeSafe Jev judgment, and encrypted history

    yushen100Python★13

  • D jev-tree Claims to be a Jev-native probabilistic tree/graph runtime for verifiable multi-step decisions; beyond one positioning line there's no usage or examples, and 34 stars in a day and a half seems fast, looking more like bandwagon naming. #LLMs#ML Research#AI Agents

    Jev-native probability tree and graph runtime for verifiable multi-step decision making.

    Chuf-HPython★34

  • D jev-recipes Over 80 TypeScript recipes around Jev, covering retrieval, answer validation and dialogue flow. It belongs to the same batch of Jev/Laya ecosystem repositories, and without that model it is just empty-shell examples. #AI Agents#LLMs#RAG

    80+ composable TypeScript recipes powered by Jev for AI agents, retrieval, answer verification, and conversation workflows.

    agencyenterpriseTypeScript★10

  • D DLSS5-composite-node-for-blender The description is a single line, 'one node to use dlls5', with no explanation of how to hook it into Blender compositor nodes, which GPUs and versions are supported, and no usage notes for the C++ source. DLSS depends on a proprietary runtime, so an open-source node is hard to stand alone and looks more like a placeholder repo. #3D#GPU#blender

    one node to use dlls5

    dutou520C++★8

  • D ironshield-analysis It claims static analysis of the Ironshield anti-cheat, but the repo has only a title, with no method, output or conclusion, and 33 stars in half a day looks like riding the anti-cheat topic. Such content is also often used for detection evasion, so it isn't worth the time. #Anti-Cheat#Reverse Engineering#Game Cheats

    Static analysis of Ironshield anti-cheat

    hseoaC★33

  • D jevbot A bot that lets Jev talk in Discord, described in one sentence with no command list or deployment instructions. A peripheral little project under the Jev topic whose usefulness depends on whether the upstream service itself holds up. #LLMs#Social#discord

    discord bot for jev that lets it talk

    lyramakesmusicPython★12

  • D jev-agent-design-with-topk-logits-choices Stacks terms like Jev, top-k logits, hierarchical memory, and dependency replanning into a "research design" with no runnable instructions or results, so it's impossible to tell what it actually accomplished. #AI Agents#LLMs#ML Research

    Research design for a Jev-native agent system: tool integration, speculative parameter proposals, external helper logits Top-k proposals with Jev-controlled fallback ,decision-aware hierarchical memor…

    6Mikao9Python★20

  • D domaincraft It claims to be domain-driven design anchored in code and tests, yet offers only a one-line description and no sample project, so it is unclear whether it is a methodology note or a scaffold. #Code Quality#Testing#LLMs

    Domain-driven design, grounded in code and tests.

    DANNNNNJUANPython★7

  • D AdaHVLA-Adaptive_Harness_VLA It aims to connect task reasoning and VLA execution through an adaptive harness, but both the paper and real-robot deployment are marked "coming soon," leaving only a framework description. #Robotics#ML Research#AI

    AdaHVLA connects task-level reasoning and memory with vision-language-action (VLA) execution through an adaptive harness. Preprint version of our paper is coming soon. The real-world deployement part …

    HaaareallyPython★8

  • D kunying-cms A PHP film and TV CMS built around scraping aggregation, VIP membership, and encrypted paid themes — essentially a complete kit for building piracy streaming sites. Technically unremarkable with clear legal risk; not recommended. #Piracy#Video#Self-Hosted

    KunYing CMS - free and open-source video CMS | one-click install · collection aggregation · member VIP · multiple themes (including 2 encrypted paid themes) | KunYing Video CMS

    kungecPHP★15

  • D laya-goish Laya's GGUF inference and HTTP service, implemented in "Goish Rust". The name is a parody and the positioning is vague, and with several similar services in the same batch, it's hard to distinguish without performance data. #LLMs#APIs#Performance

    Laya GGUF inference and HTTP server in Goish Rust

    cogentica-aiRust★13

  • D ComfyUI-Laya A ComfyUI custom node whose name suggests a Laya connection, but the description is empty, with no node list, parameters or example workflow. There's no telling what installing it gets you, so don't spend time on it for now. #Image Generation#Plugins#comfyui

    jtydhr88Python★9

  • D muse-jev-playbook Wants to add a "Jev decision" layer in front of Muse as a cheap gate, configuring confidence policies and routing; it's concept-first with no testing, and it doesn't explain what Jev itself is. #AI Agents#LLMs#ML Research

    Jev decision layer for Muse: a fast, cheap TypeSafe AI gate before expensive agent work — confidence policy, recipes, reference router, honest measurement.

    Bodila51Python★14

  • D UltimateLemmings-releases Only a signed and notarized macOS installer, with private source. It is a distribution channel, not an open-source project that can be evaluated. #Games#macOS#Downloaders

    Signed, notarized macOS release downloads for Ultimate Lemmings (source is private)

    ErikVelandLanguage unknown★9

  • D mojev "Typed calibration decisions in a single forward pass" is dense with terminology but offers no problem definition, interface or experiments, making it hard to verify and harder to use. #LLMs#ML Research#AI Agents

    MoJev: typed, calibrated decisions in one forward pass.

    MoLeMo-LabPython★16

  • D alyx-jailbreak-releases A release page for a jailbroken Alyx APK for Quest 3, with only binaries and Russian/English docs and no application source. It is cracked distribution, and the payload can't be audited. #Jailbreak#Reverse Engineering#Android

    Alyx Jailbreak for Quest 3 — APK releases and RU/EN documentation, no application source code.

    alferikoLanguage unknown★33

  • D chat-nojev A derivative of jev-chat that claims to merge judgment, ranking and candidate replies into a single call with just one key. The changes are minor, and the parent project itself is of questionable origin. #LLMs#APIs#AI Agents

    A derivative of the jev-chat conversation copilot: judgment, ranking, and candidate replies are all produced by the same model in a single call, so you only need one key. Three builds for Windows / ma…

    chy4proPython★11

  • D qwen-image-2.1-comfyui-workflow The repo is named Qwen-Image ComfyUI workflow, but the description is empty, with no screenshots, model version or node dependency notes. It's one of many config repos riding the model's release hype, and whether the results can be reproduced is unclear. #Image Generation#Images#comfyui

    foprcPython★8

  • D travel-agent A Spanish-language practice exercise combining MCP and ADK for a travel agent — essentially a sample project following a tutorial, with no reusable components and no problem it solves that others haven't. #AI Agents#MCP#Learning

    Crea un agente de viajes con el kit de herramientas de MCP para bases de datos y el Kit de desarrollo de agentes (ADK)

    nbellocamPython★8

  • D cmd-mod-jev-nudge A small Command Code plugin that uses Jev to decide when an agent should keep working. Its scope is narrow and it depends on the unpublished Jev, so outside users cannot verify its value. #AI Agents#Plugins#LLMs

    Command Code mod: nudges the agent to keep going when it stops with work left, judged by Jev

    CommandCodeAITypeScript★12

  • D Stremio_CLI A command-line viewer and downloader for the Stremio ecosystem that actually points to pirated resource streams. It isn't technically hard, but the legal and account risks fall entirely on the user. #Piracy#Downloaders#CLI#Streaming

    CLI Tool for watching/downloading movie/series/anime

    Uchiha01ShivamPython★9

  • D AI-Painter-Harness The description is just one line, "AI draws what you want", with no information about the model, interface or how to call it. A typical empty-vision repository. #Image Generation#AI#Images

    AI Paints Whatever You Want!

    alby13Python★15

  • D palworld-manager-mod It uses the same template as several mod menu repos from the same batch, claiming files are processed only in the browser, yet there's no code or usage instructions. Palworld has no such troubleshooting need, and the account got stars a day after creation, suggesting batch boosting. #Games#Game Cheats#Windows

    Local Palworld mod troubleshooting tool. Files stay in your browser.

    e-stevens8901Language unknown★10

  • D facebook-auto-reporter A cloud management panel for bulk-submitting Facebook reports, dressed up as a 'compliance review process' but essentially a tool for amplifying report volume. No technical details, and its purpose is harassment and account suppression; not worth touching. #Scams#Social#Automation

    Cloud-based Facebook report management & moderation dashboard for organizing legitimate content reports and moderation workflows.

    masonrcodesLanguage unknown★25

  • D Claude-Opus-5.5-100-HTML-Files A gallery of single-page HTML generated by 100 models plus prompts. It is a content collection rather than a tool. #AI#Prompting#Web Frontend

    100 self-contained HTML pages made with Claude Opus 5.5 — gallery with screenshots and the exact prompts

    MiaAI-LabHTML★34

  • D OpenAI-Interview-Notes OpenAI interview questions and answers, a pure content repository. Neither the source nor the timeliness of the questions is explained, so for interview candidates it is just another scattered list. #Learning#Jobs#AI

    OpenAI Interview Questions and Answers

    SchuturePython★21

  • D sasuke A Rust project called sasuke desktop with a two-word description, so it's unclear whether it's a desktop environment, a tool, or a game. It appeared alongside the author's other Jev repos and carries zero information. #CLI#Terminal

    Sasuke desktop

    klauswgRust★32

  • D one-person-one-site A navigation list of personal websites and independent blogs, a resource collection with no tool value and no verifiable inclusion criteria. Lists like this are everywhere. #Content Writing#Knowledge Base#Learning

    One person, one site · a collection of personal websites, independent blogs, and digital gardens worth following / A curated directory of personal websites, blogs & digital gardens

    realchendahuangPython★7

  • D yipiantian A GDScript project with an empty description, so it's unclear even whether it's a game or a plugin. A repo with no stated purpose can't be evaluated, so skip it. #Game Dev#Games

    henjiccGDScript★8

  • D Graveyard-Keeper-2-Trainer A trainer for Grave Keeper 2 with 55 cheat options and save repair. It's a game cheat and outside the scope of what we recommend. #Game Cheats#Games#Windows

    Graveyard Keeper 2 trainer for PC — infinite energy and resources, instant crafting, perfect corpse quality, unlimited tech points, tireless zombie workers, invincible undead army. Plus a borderless w…

    BeigewithinpackC++★30

  • D Dressmaker-Trainer A trainer for a clothing-store simulation game, with 45 cheats plus a so-called precision aid. It is a game cheat, so skip it. #Game Cheats#Games#Windows

    Dressmaker Trainer for PC — infinite money and fabric, no cutting waste, forgiving sewing line, unlock every pattern, trim and dye, freeze the commission timer. 45+ options with precision assists sepa…

    GlueoevoyageC++★30

  • D Happy-Wheels-Cheats A cheat tool for Happy Wheels that claims to block score submission by default. It is still a game cheat. #Game Cheats#Games#Windows

    Happy Wheels cheats for PC — invincibility, limbs that stay attached, gore toggle, reveal the real win condition, unlock all 11 characters, gravity and ragdoll physics sliders. 45+ options. Stat submi…

    GravelLizardSlashC++★30

  • D akproxy No description, only TypeScript and the name proxy, so it's unclear whether it's a proxy service, a packet capture tool or something else. It can't be evaluated. #Networking#Self-Hosted#DevOps

    aghub-appTypeScript★14

  • D decision-tools No description, and the name doesn't reveal whether it's a decision library, a CLI or a set of examples. It can't be evaluated. #LLMs#APIs#SDKs

    callebtcPython★12

  • D slither-io-mod-menu A slither.io cheat menu whose repo has only a title and year, with no implementation notes. Such projects usually bundle unknown downloads and break multiplayer fairness, so regardless of quality it isn't worth a look. #Game Cheats#Games#Scams

    Slitherio Mod Menu 2026

    edwardsa9942Language unknown★10

  • D Visual-Jev The description is empty and the repo name follows the naming of a batch of Jev projects from the same period, with no feature notes. There's no way to tell what it does, so per the rules, skip it. #LLMs#AI Agents

    jiangxiluningPython★8

  • D LocalRoom A TypeScript project with only a repository name and no description, so it is impossible to tell whether it is a local collaboration room or something else. Not enough information to evaluate. #Collaboration#Self-Hosted#Web Frontend

    shxhxnTypeScript★7

  • D qwen38-27B-dual-rtx5060 No description; from the name it is a configuration record for running Qwen on dual RTX 5060s, essentially personal notes. The hardware combination is also too unusual to be of much reference to others. #LLMs#GPU#Hardware

    skrodahlLanguage unknown★11

  • D CVE-2026-87902 Claims to be a PoC for an unauthenticated WordPress RCE, but the identifier doesn't match the description, and there is no affected scope, version, or reproduction steps. Repos like this are either placeholders or riding a CVE name, and security researchers who open it just waste time. #Exploits#App Security#Network Security

    Unauthenticated RCE on Wordpress

    vulpecunaPython★8

  • D AX It only calls itself a "fast, lightweight CLI," without saying what commands it offers or who it's for, so the information content is near zero. #CLI#Terminal#Productivity

    A fast, lightweight Cli.

    Axium-LabsRust★9

  • D my-agents A dumping ground for personal agent instructions, roles and templates, with no stated use case and no reusable engineering structure. Prompt and agent template repositories are already flooded, and this one is not needed. #AI Agents#Prompting#Learning

    My custom agent instructions, specialized roles, skills, and reusable templates.

    SandeepravillubuildsLanguage unknown★8

  • D RC4Decryption Only a title, "RC4 decryption", with no explanation of input format, usage or origin. However short the code is, its usability can't be judged, and you finish reading no wiser. #Reverse Engineering#Privacy#Debugging

    WeedHashPeddlerC★23

  • D Medusa-Active-Matter A cheat with aimbot, ESP and wallhack all included, with no technical explanation. It ruins other people's gaming experience, so skip it. #Game Cheats#Anti-Cheat#Games

    Aimbot, ESP, Wallhack, and much more

    artem-smithops1990a5Language unknown★30

  • D MooMoo-menu-mod A MooMoo.io script menu repo whose description is just the word 'Unpatched', with no features, installation or risk notes. One of several same-template repos in the batch, it's a typical game-cheat distribution page. #Game Cheats#Games

    MooModScript Unpatched -2026

    elenfisher73Language unknown★7

  • D point-blank-mod A Point Blank cheat menu repo whose description is just 'NEWEST mod Menu', with no technical content. One of several batch-registered cheat distribution accounts, it's high-risk and worthless. #Game Cheats#Games#Scams

    Point Blank NEWEST mod Menu

    evelyng900Language unknown★10

  • D asuka4scape-source-vulnerabilities No description; judging by the repo name, it's a collection of exploits for some private server's source code, with neither reproduction steps nor impact notes. Nobody should run something of unknown origin directly. #Exploits#App Security#Games

    murk-susSwift★10

  • D Jev_Star No description, Python, 31 stars in a day and a half, but no way to tell what it does. The name resembles a batch of recent repos with the same prefix, so it looks more like riding the hype to grab a spot; not worth your time. #AI#LLMs#Learning

    sc2musaPython★31

  • D Dogicat_DMS It calls itself "a new database management service" without saying which databases it supports or which operational pain it solves, which amounts to no description. #Databases#DevOps#Self-Hosted

    new database management service

    HeartingUPython★11

  • D openNR No description, only C++ code. The name hints at noise reduction, but its purpose can't be confirmed at all, and there's no way to run it. #Hardware#Embedded#Reverse Engineering

    clshortfuseC++★14

  • D tiktok-reporting-bot An automation tool for bulk-submitting TikTok reports and appeals, with a keyword-stuffed description and no implementation details. Its purpose is to amplify report volume and pressure accounts, which is platform abuse, and no one should touch it. #Scams#Automation#Social

    Automated reporting assistants are programmed software mechanisms designed to streamline and aggregate the submission of appeals or violation notices regarding TikTok profile, acc, or reports report b…

    freeswordbaitLanguage unknown★7

  • D Zeta5 An empty-description Lean repo that says neither what propositions it formalizes nor who it's for, making its value impossible to judge. Ten stars at this scale is no evidence of anything. #Science#Learning#ML Research

    mo271Lean★10

  • D jailbreak-archive The title points to a jailbreak prompt archive, but the README is empty, with no purpose statement or signs of maintenance, so it's just a materials repo. #Jailbreak#LLMs#Prompting

    AIwolfieLanguage unknown★11

  • D AirCard A C# project with no description or explanation, making it impossible to tell what it does or who it is for. #Productivity#UI Design#Windows

    FeryquenceC#★9

  • D Interpark- An Interpark ticket-scalping script whose README disclaims responsibility if you profit from it and run into trouble, leaving features and usage to be figured out yourself. Tools like this are short-lived and high-risk, with nothing worth borrowing for everyday development. #Automation#E-commerce#Scams

    Open source and free for everyone to use; the tutorial is in the readme, figure it out yourself. If you use it for profit and run into problems, I'm not responsible.

    michelle11032424-jpgPython★8

  • D Riot-Account-Checker-Manager A batch Riot account checker — a typical tool in the account-farming gray market, emphasizing low proxy consumption and bulk capability. It also opens a release mirror repo for exposure, with no security or privacy notes; not something to recommend. #Scams#Games#Game Cheats#valorant

    A high-performance toolkit for automated Riot Games account checking, sorting, and management. Built for speed, flexibility, and low proxy consumption — suitable for both small-scale and bulk operatio…

    plumenodeblaze7HTML★12

  • D JevAny "Calibration-aware reinforcement learning for adaptive decision systems" stacks four abstract terms in one sentence, with no visible task, dataset, or usage. With no verifiable problem statement, there is no way to tell what installing it would do. #LLMs#ML Research#AI Agents

    Calibration-aware reinforcement learning for adaptive decision systems

    weitianxinPython★8

  • D deepseek_wale_girl A pack of 104 DeepSeek whale-girl animated stickers in GIF and WebM. It is a pure asset repository and does not constitute a tool. #Images#Animation#Social

    104 DeepSeek whale-girl fan-made animated stickers, each provided as GIF and Telegram WebM files.

    DejavuMoeLanguage unknown★11

  • D RoboDawn A robotics project with an empty description and no explanation or code clues to evaluate. #Robotics#AI#ML Research

    Hugo-AGIPython★10

  • D Free-Demonstration-Guide A so-called "Flash USDT sender" that claims to show as confirmed on TronScan and Etherscan; no such effect exists on-chain, and this is typical fake-token scam language. Block it outright. #Scams#Crypto#Blockchain

    Send Flash USDT Sender on TRON, SOLANA, ETH Network. Appear fully confirmed on major block explorers on Tronscan, Solscan, Etherscan e.t.c...

    Oryc11Language unknown★19

  • D flux-action No description, and it borrows the black-forest-labs organization name. Not only is it unclear what it does, but it also looks like impersonation. #Image Generation#AI#SDKs

    black-forest-labsPython★8

  • D robux-generator A so-called Robux generator, a classic account-scam lure: platform currency can't be generated client-side. The repo has no code notes, only a bait title, and anyone should steer clear. #Scams#Games#Social

    NEWEST Robux Generator

    emilherrera319Language unknown★10

  • D cryptoptic The description is just the repo name itself; a CodeQL project that doesn't say what rules it checks or what kind of code it targets. Can't tell if it's a security scanner or an empty shell, and the org name looks like a placeholder. #App Security#Code Quality#Reverse Engineering

    cryptoptic

    nationwide-group-ossCodeQL★12

  • D librenlp The description is just 'pogo lol,' so there's no telling what it does, and being written in Rust explains nothing. A repo whose purpose can't be determined is an immediate skip. #Learning#AI#ML Research

    pogo lol

    notmarekRust★7

  • D SaintsRowPC No description, C++ code paired with a game name, with no way to tell whether it is a port, a tool, or an asset pack. Nothing to evaluate and no credible signs of use. #Games#Game Porting#Windows

    whompayC++★15

  • D tilian The repository has no description at all, just a name and 29 stars accumulated in a day, with no way to tell whether it is a tool, a script, or something else. No description plus abnormal growth leaves no basis for evaluation. #Automation#CLI#Productivity

    yb1320057104Python★29

  • D nordvpn An empty repo named nordvpn, with no description and no code, likely a name squat or mirror, offering no usable information. #Network Security#Privacy#Networking

    AttoServantViewLanguage unknown★13

  • D valogalo-latest-version The description is incoherent English and the repository name does not match the content, suggesting a junk or placeholder repository. #Scams#Downloaders#Networking

    Hi, this is repo of gologaloligalogo! We are tryna to get some storage a cloud access on a github distance machime building

    DrakeilsheaveRust★9

  • D harmony-release The repository has no description, and neither the name nor the Python language reveals what it publishes. Its purpose can't be determined, so skip it. #Build#SDKs#Automation

    ShufanSunPython★8

  • D subtitle-edit Shares its name with the well-known subtitle tool Subtitle Edit but has no description and an unknown language — likely a squatted or mirrored repo. If you want a subtitle tool, go straight to the original project and don't download from this entry point. #Video Editing#Video#Downloaders

    nucleusmotormancoilLanguage unknown★13

  • D Riot-Account-Checker-Manager-release A release mirror with the same name and description as the previous one, created at the same time just to occupy two slots. The content is still an account checker, a gray-market tool with nothing worth discussing. #Scams#Games#Game Cheats#valorant

    A high-performance toolkit for automated Riot Games account checking, sorting, and management. Built for speed, flexibility, and low proxy consumption — suitable for both small-scale and bulk operatio…

    plumenodeblaze7Language unknown★12

  • D CS-2-External No description; from the name it is an external cheat for CS2, i.e. a game hack. Not worth recommending, and it carries ban and malicious-code risks. #Game Cheats#Games#Exploits

    titangeckogatherLanguage unknown★7

  • D github-profile-guide A guide to decorating a GitHub personal profile, with no description and no substantive code. It is a take-it-or-leave-it tutorial note. #Learning#Documents#Content Writing

    WhiteSnake-MrBBooPython★9

  • D Valorant-External An external cheat program with no description, identifiable only from its name as a Valorant cheat. #Game Cheats#Exploits#Games

    Maxiforcool18Language unknown★8

  • D tala536b-2026 No description and an unknown language, so there's no way to tell what it does. A repo like this isn't worth opening no matter how many stars it has. #Learning#Science#ML Research

    pmagistryLanguage unknown★9

  • D vallennextproject No description at all, just a repository name and a Python skeleton, with no way to tell what it does, who it is for, or how to start. The 9 stars it got within a day of creation show no corresponding usage, so it is not worth clicking. #Automation#CLI#Productivity

    vallenmulia99Python★9

  • D mtechbro94-chine-ivoire-import A China–Côte d'Ivoire import-related repository with no description or documentation, so it's impossible to tell whether it's a business system or a page prototype. #E-commerce#Web Frontend#Marketing

    danielkouame091-sourceCSS★8

  • D CSE-326 A coursework project using a course number as the repository name, with no external value and no reusable implementation details. #Learning#Web Frontend#Documents

    tniishhJavaScript★8

  • D app.py No description, and the file name is the repository name, so no purpose can be discerned. #Learning#Documents#AI

    danielkouame091-sourceLanguage unknown★9

  • D alexharonyandega-dev A personal profile repository with the same name, with no description and no content. It doesn't constitute a project. #Learning#Documents#AI

    alexharonyandega-devLanguage unknown★12

Show HN 3

A

Show HN: I built a post-mortem debugger for native Windows x64/x86 crashes

Show HN

Post-mortem debugging for Windows x64/x86 crashes, moving WinDbg's analysis and cross-correlation into a modern GUI and using MCP so agents reason over structured data with fewer hallucinations—hitting a long-standing gap in native Windows troubleshooting. 20 points on HN and beta licenses already sold show willingness to pay. Competitors are WinDbg, Visual Studio and TTD; the risk is a closed-source small team struggling to shake the Microsoft ecosystem.

Hello HN! I've spent years debugging Windows crashes with tools that were either friendly but limited (e.g. Visual Studio) or powerful but archaic (e.g. WinDbg). I developed patterns and methods for u…

#Debugging#Reverse Engineering#Windows#MCP

▲20

A

Show HN: RxFilm Studio–Create and edit your product videos with AI agent

Show HN

This hands the entire product marketing video pipeline—script, editing, effects, voiceover, subtitles—to AI agents, targeting indie developers and small teams producing ad assets, a real need with budget. Monetization via subscription and credits is clear, but Runway, Descript, InVideo and Sora-like tools are extremely dense, making differentiation and asset copyright the biggest risks.

Hi HN, I built RxFilm Studio, a video workspace that lets you create and revise videos with AI agent. The reason I started this project is simple: I’ve noticed that many independent developers and sma…

#Video Editing#AI Agents#Marketing#Video

▲19

B

Show HN: Ive Sent It – online courier for files, with signed proof of delivery

Show HN

Turns large-file delivery into a "courier" with signed delivery receipts, solving the proof problem in legal and finance scenarios where the recipient denies receiving it—more compliance-oriented than WeTransfer. 18 points on HN shows the pain point is understood, and SaaS subscription plus enterprise compliance is a clear monetization path. Risks: email and cloud storage vendors adding native sign-off, and compliance qualification barriers.

#Documents#Privacy#Productivity

▲18

Hot New Repos

  1. SewCabinSpout/cleanupper★723 🧹 Free, open-source macOS disk cleanup CLI. Scan & safely remove caches, logs, Xcode DerivedData, npm/Homebrew/pip junk and stale node_modules from your terminal. Trash-first, zero telemetry the terminal-native CleanMyMac alternative.
  2. riba2534/claude-opus-5-5-demo★547 claude-opus-5-5-demo
  3. kryvora-network/kryvora-node★497 Reference client daemon and verification worker for Kryvora Network nodes.
  4. rgem227/knoweldge-base★406 Personal or team knowledge base, supports MCP API calls.
  5. anishfn/shapeshift★404 An input that becomes what you mean: one text box that morphs into the right UI as you type. Powered by TypeSafe Jev, works offline.
  6. JohnHeibel/PDoomVideo★345 Source code for the Claude Opus 5.5 music video for I'm Upping My P(doom)
  7. zeldaboyzlix/corz-client★268 A Minecraft 1.21.11 client focused on building tools, schematics, and DonutSMP utilities.
  8. miuuyy/Astra-Ares★238 Adaptive reasoning effort for GPT-6 during Codex tasks, powered by Jev to reduce token usage.
  9. BridgeDruidCompress/TauriKit★223 Modern UI component kit for Rust desktop applications
  10. BinaryDeliverer/CodexDesk★222 Desktop companion for AI coding agents and Codex workflows

Star gain 1d

  1. NandhaKishorM/laya+3915★20381
  2. jaredpalmer/kev+1776★5976
  3. Finderchangchang/jev-chat-JARVIS+1498★5222
  4. jev-chat/jev-chat-jarvis+1498★5222
  5. browser-use/jev-ultrafast+1011★19055
  6. debpalash/VoiceStudio+843★34959
  7. unreallabsai/unreal-agent+809★1706
  8. affaan-m/ECC+706★264510
  9. eternity4719/HowToLiveBetter+703★13011
  10. Mak5er/AirCard+696★3468