tierlist.tech
中EN

09.20 Today's Picks

299 new projects found today, 228 rated.

Created 2026-09-18 – 2026-09-20Fetched 600Filtered 50Rated 33

A · Worth watching

A

nimble

bespokelabsaiPython

Bespoke Labs extends its synthetic data pipeline into a combination of local typed decisions, contrastive data filtering and model evaluation, hitting the real pain of fine-tuning data quality; 421 stars in a day and a half plus an existing Curator user base support a hosted data and evaluation platform, with the risk that big players like Databricks push similar capabilities down-market.

Local typed decisions, contrastive data curation, and model evaluation.

#LLMs#Benchmarks#Data Analysis

★421

A

simple-jev

featherless-aiPython

Wraps any open-source model into a classifier or jev endpoint, capturing migration demand for the new typed decision paradigm, with 282 stars in a day and a half showing a clear gap in the ecosystem; it could charge for a gateway service or SDK, but the risk is that once Jev officially supports this natively, such thin wrappers get absorbed upstream.

Turn any open model into a classifier/jev endpoint

#LLMs#APIs#Self-Hosted

★282

A

crayfish

skydovesKotlin

A Compose Multiplatform image cropping component covering Android/iOS/desktop/Web from one codebase, by skydoves, so maintenance looks stable. You can add the dependency today; cropping isn't new, but full four-platform support is genuinely rare.

🦞 Compose Multiplatform image cropper for Android, iOS, desktop, and the web.

#Android#iOS#Images#SDKs

★20

A

glance-linux

ayandexyzPython

Face unlock for Linux with liveness detection: a PAM module plus daemon, also wired into the Omarchy lock screen indicator. Since Howdy there's been no decent replacement, and it states its boundaries clearly; the limits are hardware compatibility and security auditing, which need time.

Face unlock for Linux with real liveness detection: a PAM module, a daemon, and an Omarchy lock screen indicator

#Linux#Privacy#App Security#Self-Hosted

★35

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

BHas a spark41
  • B abide Makes coding agents follow project rules (naming, directories, commit conventions), addressing the common friction of agents that write code but violate team conventions, with 156 stars showing real demand from engineering teams; it could charge per seat as a CI or editor plugin, but Cursor and Claude Code are natively absorbing rule files, so the plugin slot is easily swallowed. #AI Agents#Code Quality#CLI

    Make your coding agent abide by all your project rules

    coldteadotaiTypeScript★156

  • B jeff A self-hostable decision service positioned as a TypeSafe jev alternative, initiated by LangChain's author, showing typed decision has been noticed by the mainstream ecosystem; it is clearly valuable to teams unwilling to hand decision data to third parties, and 108 stars is a decent start, but it must match official accuracy while facing pressure from openjev and simple-jev. #LLMs#Self-Hosted#APIs

    A self-hosted drop-in replacement for TypeSafe's jev, powered by GliFormer.

    logan-markewichPython★108

  • B guizang-product-video-skill Reuses real product components and design language to generate software update promo videos from code, chaining storyboard, music, motion sound effects and rendering into a Claude Code/Codex skill, solving the cost of product videos for small teams; the author brings traffic, so a template marketplace or done-for-you service is viable, with the risk that end-to-end video generation models produce the video directly. #Video#Agent Skills#Content Writing#claude-code

    Guizang product video skill: reuse real product components and design language to make software update promo videos with code. Includes storyboard copy, original music, motion sound effects, and video…

    op7418Python★148

  • B splash A local inference engine for Apple Silicon that schedules around the model itself, targeting the memory and throughput pain of running local LLMs on Macs; 293 stars in a day show people still seek easier options beyond MLX and llama.cpp, but those are already de facto standards, so a self-built engine can hardly charge license fees and must rely on hosting or enterprise support. #LLMs#macOS#Performance

    A local inference engine for Apple silicon, built around the model.

    incoaiPython★293

  • B jev-align Uses Jev plus GEPA to train calibrated classifiers from human feedback, turning 'label a few examples and ship a small model' into a reusable process, practically valuable for enterprise moderation and ticket classification; it could sell MLOps services, but the risk is that platforms like OpenAI absorb fine-tuning and calibration as hosted capabilities. #LLMs#ML Research

    Build calibrated AI Functions from human feedback using Jev and GEPA.

    sutro-shPython★109

  • B traceclause A local-first document requirements and evidence-checking workbench that outputs traceable citations and keeps human review, hitting the hard need for verifiable RAG in compliance, bidding and audit scenarios. Its 60 stars are small but the direction is right; it could be enterprise compliance software, with the risk of legal AI and general RAG platforms pressing down simultaneously. #Documents#RAG#Privacy#AI

    Local-first document requirements and evidence review workbench with traceable citations and human review.

    LingxiangXuPython★60

  • B geo-sleuth It infers a photo's shooting location from OSM geometry, skyline elevation, satellite imagery and street view and shows the derivation, turning the gray skill of geolocation into a compliant agent skill. Its 108 stars show real OSINT and news-verification demand; risks are privacy controversy and map API licensing costs, with monetization via investigation, insurance and other verticals. #Agent Skills#Images#Privacy

    An agent skill that finds where a photo was taken — OpenStreetMap geometry, elevation skylines, satellite imagery and street view — and shows its work. Works with Claude Code, Codex, Cursor, Gemini CL…

    OldcirclePython★108

  • B openjev Implements a Jev-compatible open-source System One decision service on DiffusionGemma, replacing the commercial API with self-hostable inference, directly valuable to teams wanting typed decision in private environments; 123 stars show the ecosystem seeking open alternatives, but it must match official model quality and faces competition from jeff and simple-jev. #LLMs#AI Agents#Self-Hosted#jev

    Open, Jev-compatible System One decision server on DiffusionGemma

    razorback16Python★123

  • B jev-pruner Uses Jev to decide which lines to keep before long Bash output enters the model context, saving tokens and improving agent judgment, a pain-point plugin Claude Code users can install immediately; 114 stars come from plugin marketplace spread, monetization can only be team editions or subscriptions, and the risk is agent platforms building output trimming in natively. #LLMs#Plugins#CLI#Scams

    Claude Code plugin: trim long Bash output with TypeSafe Jev before the model sees it

    tamaratranTypeScript★114

  • B omarchy-iphone-mirror Projects an iPhone onto the desktop on Omarchy (Arch + Hyprland) and takes over input separately over USB or Wi-Fi, filling a gap in Linux phone integration. Omarchy users can try it directly, but it's tied to one distro config and the underlying implementation is unclear, so it's not portable. #Linux#iOS#Streaming#Productivity

    On-demand iPhone mirroring and focused input for Omarchy over USB or Wi-Fi.

    DanielLemkyPython★50

  • B mac-shortcuts-mcp Uses Apple Shortcuts as a bridge to give any agent read-only access to Mac calendars, reminders, notes, and clipboard, plus a catalog of 539 Shortcuts action parameters. The boundaries are honestly stated (read-only, local, no account), a rare restraint among MCP projects; the limit is macOS-only and requires building the shortcuts yourself. #MCP#macOS#AI Agents#Productivity

    Let any AI agent read your Mac — Calendar, Reminders, Notes, Contacts, Clipboard, Location, Weather — through Apple Shortcuts. Read-only, local, no account. Includes a harvested catalog of 539 Shortcu…

    Dominic-DKTypeScript★8

  • B Tcalm A zero-dependency CLI that tracks Islamic prayer times in the background on Linux/macOS, auto-pausing playback and popping a notification at each time. The scope is tight and usage clear, ready to use for developers who pray; it's single-purpose and doesn't aim to be more. #CLI#Linux#macOS#Productivity

    A lightweight, zero-dependency command-line utility for Linux and macOS that monitors Islamic prayer times (Adhan) in the background, automatically pausing media playback during prayer times and deliv…

    abod8639Python★11

  • B EdgeSSH A lightweight WebSSH workbench on Cloudflare Workers with host management, encrypted storage, SFTP, and process monitoring. Low deployment cost is the main selling point, suited to managing a few personal machines; Workers' limits on long connections and SFTP need to be spelled out. #Self-Hosted#Networking#DevOps#Terminal#cloudflare

    A lightweight WebSSH workspace on Cloudflare Workers with host management, encrypted storage, SFTP, process monitoring and globe visualization.

    aozoraeTypeScript★32

  • B iphone-tailnet-bridge It proxies a paired iPhone's CoreDevice connection over Tailscale using Bonjour and socat, letting remote machines debug real devices. The scenario is specific and the approach clean, immediately useful to remote iOS developers; the limits are that it covers only one link and the docs are thin. #iOS#Networking#macOS#tailscale

    Proxy a paired iPhone CoreDevice connection over Tailscale using Bonjour and socat.

    ViaaaronShell★13

  • B cli-faq-shortcuts Mines your own Claude Code and Codex history for repeated prompts and turns them into short commands installed as Agent skills across three major CLIs. The angle is concrete and targets real repetitive work, but the implementation is likely just prompts plus scripts, so you'll have to test the speedup yourself. #Agent Skills#CLI#Productivity#claude-code

    An Agent Skill that turns the asks you keep typing into short commands, mined from your own Claude Code and Codex history. Works in Claude Code, Codex and Cursor.

    kishormorolPython★48

  • B apk-reverse A one-stop script collection for Android APK reverse engineering that chains decompilation, static scanning and other common steps, plug-and-play for security practitioners and risk-control research, with 197 stars in 0.4 days from the Chinese reverse-engineering community; but jadx, apktool and MobSF already cover most needs, making commercialization hard and legal compliance the ceiling. #Reverse Engineering#Android#App Security

    Suitable for Android APK reverse engineering analysis

    newliver666Python★197

  • B skyline-speeder It rewrites congestion control with eBPF struct_ops and a Rust control plane, speeding up satellite and cross-border links with non-congestion random loss—a real, under-served networking pain point, with 74 stars showing research interest. But the kernel 6.12+ requirement limits deployment, so commercially it's only dedicated lines or cloud vendor customization. #Network Security#Networking#Linux#Performance

    Sender-side TCP acceleration: eBPF struct_ops congestion control with a Rust control plane. For links with non-congestive random loss. Kernel 6.12 LTS+.

    CYBERVERSE-ResearchPython★74

  • B Awwards-mcp Gives coding agents retrieval, screenshots, and design-element extraction from award-winning sites, pitched as a free Mobbin alternative, usable by frontend and UI people once installed. Concerns are its reliance on scraping, unclear long-term availability, and the recent cluster of "design inspiration MCP" projects. #MCP#UI Design#AI Agents#Open-Source Alternatives#mobbin

    Free, open-source MCP server that gives AI coding agents (Claude Code, Codex, Cursor, OpenCode, ZCode…) design inspiration from award-winning websites — search with real screenshots, pull design DNA, …

    INSANE0777TypeScript★25

  • B coastal-simulation A coastal simulation solving the shallow water equations in real numbers with Three.js, featuring breaking waves, foam, wet sand and a shoreline walking view. Rare among WebGL water demos for explaining its solution clearly, it's good code to read for graphics people; the limitation is that it's a demo, not a reusable library, with no docs yet. #3D#Web Frontend#Animation#Science

    A real-time coastal simulation in Three.js, pairing a shallow-water solver with breaking swells, whitewater, wet sand, spray and shoreline exploration

    iamtechartistJavaScript★13

  • B react-native-nitro-rolling-number A React Native rolling-number (odometer) component with a native implementation via Nitro Modules, by the author of react-native-zoom-toolkit, so quality has a baseline. Scope is restrained and it's installable today; the downside is many similar components, so only swap when you need native performance. #Animation#iOS#Android#SDKs

    Native rolling number (odometer / ticker) view for React Native, built with Nitro Modules

    ronickgTypeScript★10

  • B jev-semgrep Uses Jev to score each line of code semantically, supporting AND/OR/NOT combinations and cross-language, Japanese-English mutual search, making 'grep by meaning' a usable CLI; 74 stars show real demand for semantic search, but it overlaps with vector search and IDE semantic search, making it hard to charge independently or become a standard tool. #CLI#LLMs#Code Quality

    grep by meaning, across languages. TypeSafe Jev scores every line against a meaning; combine meanings with AND/OR/NOT. A meaning-based grep. Search English in Japanese and Japanese in English

    uehajJavaScript★74

  • B Pixel-Camera It backports the Pixel 11 camera Look, 5x portrait, and 10x zoom toggle to Pixel 6–10, with clear scope and device list, so older Pixel users can decide in ten minutes. The catch is it requires root/Magisk, and such camera feature ports have precedent, so innovation is limited. #Android#Images#Hardware

    Unlock newer Google Pixel camera features on older Pixel devices. This project brings Pixel 11's next-generation Camera Looks to Pixel 6–10, enables 5x Portrait Mode on Pixel 6 Pro to 10 Pro, and adds…

    AkshayykadamKotlin★29

  • B 2ship2harkinian A native PC port of Majora's Mask via recompiling the N64 binary into executable code, a real contribution to the porting scene that's playable once built. The limit is you must supply your own ROM, the audience is narrow, and the repo itself has little info beyond a website link. #Games#Game Porting#3D#n64

    https://www.harbourmasters.org/game/2ship2harkinian

    HarbourMastersC★10

  • B ComfyUI-Hyperflow It wires Video Rebirth's 8-step LoRA into ComfyUI's native MiniMax-H3 without touching core files or needing a custom sampler, and the wording stays restrained. Useful for ComfyUI users chasing new video models, but it depends on a not-yet-common weight, so the ceiling is clear. #Video#AI#Image Generation#comfyui

    ComfyUI node pack for HyperFlow [Video Rebirth's] 8-step LoRA for MiniMax-H3 (video + audio), ported onto ComfyUI's native MiniMax-H3 model. No ComfyUI core files are modified; no custom sampler is ne…

    Saganaki22Python★31

  • B ddc A Rust rewrite of DEX-to-Java decompilation featuring progressive analysis and a bilingual CLI, where faster startup and lower memory matter for batch sample analysis; 115 stars show Android reverse engineering still lacks modern tools, but jadx has a deep ecosystem and accuracy is the hard metric, making direct charging difficult and leaving only enterprise security suite licensing. #Reverse Engineering#Android#CLI

    DEX → Java decompiler in Rust — fast, progressive analysis, bilingual CLI

    ejfkdevRust★115

  • B hype A local Markdown presentation tool in C++ with a visual slide editor, taking a native route opposite Marp/Slidev. The one-liner is clear and suits frequent tech sharers tired of heavy web tools—check the demo; limits are just 11 stars in a day and no platform or setup details. #Presentations#Documents#Editors#Productivity

    Simple Markdown presentations with a visual slide editor

    omacomC++★11

  • B Audio_Bridge Turns Windows into a Bluetooth A2DP/LE Audio receiver so phone audio plays through the PC speakers, with a tray app and auto-reconnect; the need is real, the scope restrained, and the bilingual description not exaggerated. Novelty is limited, as store apps like Bluetooth Audio Receiver and AudioRelay already cover most of this. #Windows#Audio#Hardware#Productivity

    Lightweight Bluetooth audio receiver (A2DP Sink / LE Audio) for Windows. Connect your phone or tablet to your PC and play its audio through Windows speakers. Tray app with auto-reconnect, performance …

    DylanUncleC++★11

  • B mpc2000-emulator An experimental emulator running the Akai MPC2000 sampler on the MAME engine, aimed at old-school beatmakers who can't afford the hardware. The direction is specific and the description restrained, but it calls itself experimental and has only 8 stars after a day, meaning nobody has really gotten it working; similar needs are currently met with existing MAME drivers and sampler software. #Audio#Hardware#Reverse Engineering#mpc2000

    Experimental MPC2000 Classic emulator with a MAME-based engine.

    dariuszchynekPython★8

  • B zia A Sine plugin that makes Zen Browser look and behave like Dia. Its 28 stars show fellow fans, but this is pure appearance mimicry while functionality stays with Zen, a typical 'yet another' that needs ten minutes and no bookmark. #Plugins#Browser Extensions#UI Design#Web Frontend

    A Sine mod that makes Zen Browser look and work like Dia.

    z1n-kJavaScript★28

  • B ghost A full backfill and read-only query API for ERC-20 fund flows on Robinhood Chain, clearly stating it pulls 1.4 billion transfers via HyperSync and stores them in ClickHouse; the scale and boundaries are honestly described. On-chain data analysts can pick it up directly, but it serves only one new chain, so its value rises and falls with that chain. #Blockchain#Data Analysis#Databases#APIs

    Fund-flow pipeline for Robinhood Chain (chain 4663): full ERC-20 Transfer backfill via HyperSync, 1.4B transfers in ClickHouse, read-only JSON API.

    0xmikadzykiPython★17

  • B vnpresence Syncs the cover, title, and reading time of the visual novel you're reading to Discord status; a small thing done cleanly, useful to VN players once installed, and the description isn't exaggerated. The limit is that many similar Rich Presence scripts exist and the technical depth is low. #Games#Social#Productivity

    Show the visual novel you're reading on Discord - cover art, title, and elapsed time, automatically.

    BasedYukiPython★25

  • B should-i-submit A pre-submission local self-check for bounty reports that helps authors avoid triage rejections. The audience is clear and the pain real, but it doesn't say where checks come from or whether there's a rule set, so results depend on author experience and it isn't worth a dedicated look yet. #App Security#Code Quality#Testing

    A local, safety-first pre-submission reviewer for bug bounty reports.

    mdpsecLanguage unknown★16

  • B everything-search-skill Wraps Voidtools Everything's local index search as an agent skill, a piece Windows users genuinely lack, and 42 stars in 0.7 days shows the demand. But the implementation is a thin wrapper, and the description is stuffed with ⚡ and tool-name stacking, showing little craft. #Agent Skills#Windows#Productivity

    ⚡ Ultra-fast, index-powered local file search skill for AI agents (Cursor, Codex, PI-Desktop) using Voidtools Everything

    Mayuqi-cryptoPowerShell★42

  • B bend-sha256 Implements SHA-256 in Bend with a machine-checked correctness proof against the independent FIPS 180-4 spec, the most novel concept in this batch. Unfortunately its audience is only formal-methods and parallel-language enthusiasts, its practical usability is near zero, and it's better as reading than as a tool. #ML Research#Benchmarks#Learning

    SHA-256 in Bend with a machine-checked functional correctness proof against a separate FIPS 180-4 specification.

    Giulio2002Python★15

  • B NobodyTdxTQ Generates A-share quant strategies from natural language and backtests them on backtrader, using the Tongdaxin TQ interface for data, with a web frontend that produces reports in one click, lowering the bar for beginners. The limits are the recent flood of similar natural-language stock-picking plus backtesting tools and compliance and stability questions around the TQ data source. #Fintech#LLMs#Data Analysis

    Generate quantitative strategies from natural language and backtest them, an A-share quantitative backtesting system based on [backtrader], with data from the Tongdaxin TQ API. Built-in web frontend: …

    NobodySniperPython★13

  • B live-jev Controls Ableton Live with one sentence in Japanese or English, summoned with ⌘⇧Space and executed on speech. Among the jev-prefix batch this has the most real need, but it depends on TypeSafe interfaces and doesn't say how broad the command set is or what happens on failure, so wait for it to grow. #Audio#Speech#macOS#AI Agents

    Control Ableton Live with one short sentence (Japanese / English). Summon with ⌘⇧Space, type or dictate, done.

    okinaaudioPython★25

  • B Glyd A Rust compression library claiming smaller output than zstd -3 and decoding twice as fast as LZ4. The specific numbers are a plus, but 10 stars in a day and a 'world's fastest decoder' claim need your own benchmark; with so many compression options, unreproducible data means nothing. #Performance#CLI#SDKs

    The world's fastest-decoding open-source compression: fewer bytes than zstd -3, 1.3x its decode speed, 2x LZ4. Rust + C ABI + CLI.

    surya-koritalaRust★10

  • B iPhone-Duo-App-Patcher Patches Mach-O version numbers, UIDeviceFamily, and plist so the iOS simulator renders a foldable dual-screen layout. It saves work for iOS developers targeting foldables and states its scope clearly, but it is a binary-patching hack with a tiny audience; the alternative is configuring the simulator yourself. #iOS#Debugging#Reverse Engineering#Hardware

    Patches Mach-O build versions, UIDeviceFamily, and plists to unlock dual-screen foldable layouts in iOS Simulator.

    thilojaeggiShell★14

  • B Frida0xKit It bundles common Frida scripts into a modular toolbox for Android runtime instrumentation and reverse engineering. It saves digging through snippets, but overlaps heavily with Frida's official examples and objection, and no killer feature is apparent yet. #Reverse Engineering#Android#Debugging#frida

    A modular toolkit for Android runtime instrumentation, reverse engineering, and Frida-based analysis.

    allellooJavaScript★20

  • B jev-reviewer A browser tool for systematic review data extraction with verbatim citations, pointing each item in RoB 2, QUADAS-2 and similar templates to the source line; this is the most solid angle in the Jev series. Unfortunately the batch of repos appearing together drags down credibility. #LLMs#Science#Documents#Browser Extensions

    Data extraction for systematic reviews, quoted from the papers. Ask a trial report and its supplements your extraction form or a RoB 2, ROBINS-I, QUADAS-2 or TIDieR template; Jev points at the lines, …

    choxosJavaScript★18

  • B pi-prefix-cache-compaction Context compression for Pi that reuses prefix caches from vLLM/SGLang/llama.cpp to avoid cold-start recomputation. The direction targets agents' real costs, but it's tightly bound to one specific client with no docs or examples, making it hard for outsiders to adopt directly. #LLMs#Performance#AI Agents

    Pi compaction that reuses the server's prefix cache (vLLM / SGLang / llama.cpp Anthropic endpoints): no cold re-prefill, plus a warm-up so the next turn starts from cache

    jagdeepsinghdevTypeScript★10

COrdinary60
  • C JevRouter It uses Jev for lightweight routing across models, tools and sub-agents, replacing some prompt-based judgment. The idea is sound but model routing is already a red ocean—OpenRouter, LiteLLM and agent frameworks all have similar built-in capabilities; its 86 stars mainly ride ecosystem hype, making independent revenue or standard-dependency status unlikely. #LLMs#AI Agents#APIs

    A lightweight Jev-powered router for models, tools, and subagents

    BillionsBobbyTypeScript★86

  • C qiaomu-download A declarative video-download skill that handles YouTube, Bilibili, X and other sites uniformly with audio extraction and subtitles, ready to use for content workers; 123 stars in 0.6 days ride the skill ecosystem boom, but it is essentially a prompt wrapper over yt-dlp with no technical moat, carries platform anti-scraping and copyright risks, and is hard to commercialize. #Downloaders#Video#Agent Skills

    A general-purpose online video download and media extraction skill. Use when the user provides a YouTube/youtu.be, Bilibili/bilibili/b23.tv, X/Twitter, Vimeo, TikTok, Instagram, Facebook, Twitch, Redd…

    joeseesunPython★123

  • C NullMotion It outputs motion-graphics ad finals side by side with black-and-white draft frames and supports frame-by-frame MP4 export, suited to concept previews and pitch decks, with 100 stars from visual impact. But it's more a portfolio demo than a toolchain, replaced by After Effects templates and AI video generation, with limited commercialization space. #Video#Animation#Video Editing

    Launch-film preview: a finished motion-graphics ad over the black-and-white HyperFrames drafts it grew from, with frame-accurate MP4 export.

    blixvipHTML★100

  • C expo-dynamic-notifications Provides Dynamic Island-style in-app notifications for React Native, filling a gap in iOS top-of-screen interaction components, with 110 stars showing real RN developer demand; but the scope is too narrow and heavily dependent on platform API changes, and Expo or a major community library could build in the same capability at any time, making independent commercialization nearly impossible. #iOS#UI Design#Animation#react-native

    💧 Dynamic Island–style in-app notifications for React Native.

    rit3zhTypeScript★110

  • C editaplot2026 Drives Origin/OriginPro via a skill to generate editable scientific charts with 2026 version support, bridging the gap between scripting and manual fine-tuning in research plotting; 85 stars show a real niche, but proprietary formats and version compatibility are a long-term burden, limiting it to a small tool or tutorial monetization. #Science#Agent Skills#Data Analysis

    EditaPlot — editable Origin/OriginPro scientific plotting skill, with Origin 2026 (10.30) compatibility fixes.

    liushunqi8-hashPython★85

  • C djev-spark A container recipe for running DiffusionGemma structured decisions with NVFP4 on DGX Spark, a hardware-adaptation experiment that saves debugging time for teams that already bought a Spark; but it is a single-point recipe with no maintenance commitment, and its value will quickly fade as official quantization support matures. #LLMs#GPU#jev

    DiffusionGemma NVFP4 structured decisions on a DGX Spark: container recipe

    mmastracHTML★84

  • C weilanx-cover-skill A Chinese social-cover generation skill for agents, focused on multi-ratio layout and character consistency; the need is real but the description is thin, not saying which image model it calls or how consistency is achieved. Useful for Xiaohongshu or WeChat cover makers; alternatives are template sites. #Image Generation#Agent Skills#Content Writing

    weilanx Cover Skill: generate Chinese social covers with multi-ratio layouts and consistent characters.

    Azure12355Language unknown★22

  • C clarify-intent A skill pack teaching agents when to ask follow-up questions and when to just act, addressing a real pain point since coding agents today either talk too much or guess wildly. But it constrains behavior only through prompts, depends heavily on the host model, and has no evaluation backing it. #AI Agents#Agent Skills#Prompting

    An AI Skill that teaches agents when to ask and when to just do it — minimal questions, no requirements interviews. Beginner-friendly.

    Hugo-DDTLanguage unknown★25

  • C lorenzini Makes third-party review replies from Claude Code and others actionable and judges whether a "pass" is genuine or just polite, a small but real niche. The problem is hardcoded rules that only fit a specific review flow and break with a different bot. #AI Agents#Agent Skills#Code Quality

    Claude Code skills that wait for a third-party PR reviewer and decide whether its verdict actually means pass

    Nanako0129Shell★12

  • C is-malicious Scan a repo for suspicious signs before running someone else's code, a need anyone cloning unfamiliar projects understands. But there's only one line of description, no word on whether detection is rules or a model, and false positives are costly, so it trails the existing sandbox-plus-dependency-audit combo. #App Security#Code Quality#Testing

    A codebase scanner that helps you not run malicous code

    luantakTypeScript★13

  • C thor-pathfinder Adds app swapping and button shortcuts for the AYN Thor dual-screen handheld, a very specific problem the author clearly uses himself. But the audience is only owners of this niche device, the feature set is narrow, and outside readers can barely use it. #Android#Hardware#Productivity

    Swap apps between the AYN Thor's two screens and put shortcuts on its buttons. Free and open source (GPL-3.0).

    KaitonGxxKotlin★8

  • C ms-sam-mike-mary-decomp Rewrites SAPI 5's Microsoft Sam/Mike/Mary voices in portable C to sample-level accuracy and restores SAPI 4 effects like Hall and Stadium, a serious reverse-engineering recreation. Unfortunately its use cases are mostly nostalgia and memes, so its engineering value outweighs its practical value. #Speech#Reverse Engineering#Audio#Windows

    Microsoft Sam, Mike and Mary (SAPI 5) rebuilt in portable C, sample-accurate, plus the SAPI 4 voice effects (Hall, Stadium, Space, RoboSoft, Whisper)

    KamiKitsune420C★10

  • C SPC970-MechaLIBerator A firmware dump tool for the SPC970 chip, a narrow but practical item in hardware reverse engineering that those who need it will really need. But the description is one line, with no supported hardware versions, wiring, or risks, making it hard for outsiders to get started. #Reverse Engineering#Hardware#Embedded

    SPC970 Mechacon dumper

    LibbersC★26

  • C grokbot-field-notes Turns the rules, roles and failure logs from the Grok Bot team's 72-hour livestream build into a copyable AGENTS.md template, which is experience distillation rather than a tool, helping people avoid pitfalls but with no technical moat since anyone can copy it; it has no product vehicle and can only drive consulting or course traffic, hard to sustain after the hype fades. #Prompting#AI Agents#LLMs

    Rules, playbooks, bot roles and a failure log from the xAI Grok Bot team's 72-hour live build. Drop AGENTS.md into your repo and go.

    unicodef1wnLanguage unknown★195

  • C zcode-prompt-toolkit It lets you replace a harness's first system message and rewrite built-in prompts and guardrails, a practical patch for those wanting to deeply modify agent behavior. But it's tied to a specific version of one tool and breaks on upgrade, and similar prompt-override scripts are plentiful. #Prompting#LLMs#AI Agents

    Take full control of your ZCode harness: inject your own first system message and edit every built-in prompt, guardrail, and tool description (verified on ZCode 3.14.1)

    2BE-opsPython★23

  • C pkg-sender Pushes PKG packages over the LAN to install on PS4/PS5, a daily necessity for jailbroken console players, with a decent description of the interface and flow. But the community already has several stable senders, so this is just a rewrite in another language. #Games#Networking#CLI

    PKG Sender - install PS5/PS4 packages over LAN

    LoopayehC#★10

  • C AirCard-iOS It skins Apple Wallet cards and the lock-screen passcode screen for personalization-minded users, a shallow beautification relying on non-public system capabilities, with 73 stars from showmanship. It breaks with any iOS version change, has no commercial moat, and may trigger platform policy risk. #iOS#UI Design

    Apple Wallet card skins and lock screen passcode themes on iOS 27.

    Mak5erSwift★73

  • C kisakcod-port Runs the community-rewritten CoD4 kisakcod in the browser, which sounds interesting. But the repo has only a one-line description, doesn't say if it's actually playable or what approach it uses, and isn't an original implementation—you'd have to wrestle with the build to verify. #Games#Game Porting#Web Frontend#WebAssembly

    port of kisakcod (cod 4 2007 reimpl) to the web

    slqntdevssHTML★22

  • C pi-skill-orchestrator An extension for the Pi editor that reduces how much the Agent Skills directory eats into the system prompt—a real but very niche pain point, tied to one editor ecosystem. It also gives no measured token-savings data. #Agent Skills#LLMs#Prompting

    Pi Skill Orchestrator is a Pi extension for people who use a lot of Agent Skills. It keeps Pi's normal skill autocomplete, but it stops the complete skill catalog from taking space in the LLM system p…

    badgidsTypeScript★9

  • C peep A hand-written PE header parser in C++17 that reads exe files byte by byte, reaching 53 stars in two days. Good reading for anyone wanting to understand the PE format, but mature libraries like pefile and LIEF already exist; it's neither a library nor a unique analysis angle, more like practice notes. #Reverse Engineering#Windows#App Security

    PE file parser in plain C++17 - reading exe headers byte by byte

    doxsirC++★53

  • C jev-shell-history Uses Jev to semantically rank zsh history for Fish-style autosuggestions, improving terminal completion hit rates; but the feature is narrow, depends on an external inference service, and 66 stars is modest growth, so it cannot commercialize independently, and its biggest threat is the built-in history search in the zsh and fzf ecosystems. #Terminal#CLI#LLMs#jev

    Fish-style zsh history autosuggestions ranked by Jev (TypeSafe)

    mrnuggetTypeScript★66

  • C right-panel A Rust screen-edge sidebar packing emoji, clipboard history, color picking, notes, and quick launch. Cross-platform support is a plus, but the features are too scattered and shallow, and launchers and clipboard tools are already mature. #Productivity#Windows#macOS#Linux

    A liquid side panel for your screen edge: emoji, clipboard history, colors, notes, apps and more. Tiny Rust app for Windows, macOS & Linux.

    raminturneHTML★11

  • C four-horsemen-pocs It studies what happens when multiple products fork the same driver and each trims away security boundaries, an interesting topic. But the repo has only a one-line description, with no affected products, exploit conditions, or reproduction steps; security researchers must first guess what it's saying. #App Security#Exploits#Reverse Engineering

    what happens when multiple products fork a driver and strip it's security?

    0xdeadbeefnetworkC★15

  • C oai-adversarial-plugin A small toolkit for debugging agent anomalies, offering timezone normalization, model consistency monitoring, and turn-state observation for proxy plugins. The problem is it serves a very narrow self-built proxy scenario, and the terms are unexplained, so outside readers can't judge whether it applies. #LLMs#APIs#Debugging

    CLIProxyAPI plugin: timezone normalization, model consistency monitor, X-Codex-Turn-State observation & override, probe track.

    FlashyyLGo★20

  • C arabic-ui A rule set for Arabic microcopy and RTL engineering aimed at agents and frontend projects, filling a gap in minority-language details, but the deliverable is essentially a rules document whose effectiveness depends on how the host uses it and is hard to verify. #UI Design#Agent Skills#Web Frontend

    Natural, high-fidelity Arabic UI microcopy and RTL engineering rules for AI agents and modern webapps.

    JKc66TypeScript★11

  • C jegrep A 'semantic grep' that searches code from natural-language descriptions—the starting point is concrete and the CLI form is right, but it relies on Jev rather than a self-hostable index, and neither practical usability nor latency is addressed. #CLI#LLMs#Debugging

    Semantic grep: find code by describing what you're looking for, powered by Jev.

    can1357Rust★13

  • C HINT A research project on long-horizon robot operation whose title coins "Human-Intent Inception." Worth following if papers and experiment videos come later, but it's currently neither stated as a paper repo nor given install steps, so 17 stars means watch for now. #Robotics#ML Research#AI

    HINT: Human-Intent Inception for Long-Horizon Robot Manipulation

    robot-hintPython★17

  • C xuan A Linux image editor ported from Compositor and rewritten in Rust. The description is honest and scope clear, meaningful for Linux users wanting a native lightweight editor; but it's a port rather than something new, and 9 stars in a day means it's not daily-driver ready. #Images#Editors#Linux

    An image editor for Linux, inspired by Compositor and ported from it.

    silverlingRust★9

  • C naver-blog-automation It drafts, reviews, and optionally publishes Naver blog posts for agents, but the "frozen core + taste packs" phrasing is vague and doesn't explain how it differs from just asking a model to write. Only somewhat useful to Korean bloggers on Naver, and 8 stars shows no real usage. #AI Agents#Agent Skills#Content Writing

    Draft, review, and optionally save posts to your own Naver blog. Frozen core + taste packs so any agent can fit it to its user.

    0x8905Python★8

  • C youtube-transcript-minimal A headless tool that pulls YouTube subtitles into an agent with one click, needing no account or third-party server, which is cleaner than similar tools. But subtitle downloading is already crowded, and the project offers no new method, just fewer distractions. #Video#Downloaders#AI Agents

    A minimalist YouTube transcript downloader made for the agent era — one click gets any video’s captions into Codex or Muse. No popups, ads, accounts, or third-party servers.

    BradGrouxJavaScript★9

  • C WaterCuda It uses CUDA to generate procedural islands and spectral ocean surfaces, then renders them in the browser via WebGPU—an interesting technical route, but more a personal graphics demo. There are no install instructions or performance data, and the same author shipped two similar repos within three days. #GPU#3D#Web Frontend

    Infinite procedural islands and a spectral ocean, authored in CUDA and rendered in the browser with WebGPU.

    SamG-CoderJavaScript★8

  • C FreeCast A 'dual-stream free' multi-platform live-stream combo player with a niche positioning, but the description is too short—no install target or supported platforms. Similar aggregator players are plentiful, and no differentiation is visible here. #Streaming#Video#Productivity

    Two-way free, multi-platform live-stream combination player

    Sur4lisLanguage unknown★25

  • C image-prompt-reverse It reverse-engineers images into bilingual Chinese-English prompts with style tags and structured JSON. Many img2prompt tools already exist, and here the model source and accuracy are unstated; the AGPL plus separate commercial licensing also makes its positioning awkward. #Image Generation#Prompting#Images#AI

    Reverse-engineer an image into a generation prompt (Image to Prompt): bilingual Chinese-English prompt reconstruction + style tags + structured JSON prompt, zero-dependency validation script | Forensi…

    TechnicalflightPython★11

  • C Accorduon A SwiftUI experiment that plays a foldable iPhone's hinge like an accordion bellows—creative and well-judged, with restrained code. The problem is it depends on the not-yet-existing iPhone Duo and iOS 27.1 APIs, so it's only a concept demo. #iOS#Animation#UI Design#iphone

    An accordion for the foldable iPhone Duo. The hinge is the bellows: fold and unfold to play. SwiftUI, onHingeChange, synthesized reeds.

    artemnovichkovSwift★14

  • C pi-agent-for-claude Plugs an external agent into Claude Code's subagent interface with streaming echo, a glue layer in the toolchain. Useful to power users, but it depends on internal interfaces on both sides that may break anytime, and the description covers integration but not failure boundaries. #AI Agents#LLMs#Plugins

    Run pi as a native Claude Code subagent: pi agents in Claude Code's own agent UI, streamed live, on any model pi can reach.

    FazalAAliTypeScript★8

  • C UserAgent-list Every five minutes, Actions scrapes a UA library covering 100+ browsers and 100+ systems—diligent automation, but essentially a data collection. Its value depends on source quality, which the README fails to explain. #Automation#Testing#CI

    Automated User-Agent database for 100+ browsers on 100+ operating systems. Updated every 5 minutes via GitHub Actions. Weighted by real-world popularity, grouped by OS and browser, with a cumulative A…

    ThisTakouPython★34

  • C a-share-chart-lab A pure front-end A-share technical analysis dashboard with candlesticks, chip distribution and nine-turn indicators, connecting directly to public market data APIs. Attractive to retail investors, but data source stability and indicator correctness are unaddressed, making it look more like a personal practice project. #Data Analysis#Web Frontend#Fintech

    Pure frontend A-share technical analysis dashboard: candlestick charts / chip distribution / MACD divergence / nine-turn / sector capital flow, directly connected to public market data APIs, no backen…

    conglingkaishi3674HTML★8

  • C AuditHook An HTTP event gateway with SSE real-time telemetry, payload auditing, and local forwarding, with a Vue frontend. Three things bolted together but all described vaguely; 37 stars in half a day doesn't match the description's information density, so wait for a real deployment case. #Monitoring#Observability#APIs#DevOps

    High-throughput HTTP event ingestion gateway with live SSE telemetry, payload auditing, and localhost dispatch capabilities.

    needtobreatheVue★37

  • C sabi A middle layer that schedules models, compute, and providers per turn for coding agents, usable as an OpenAI-compatible proxy. Cost control is a real pain point, but the description is dense with no config examples, and LiteLLM and OpenRouter already cover most scenarios. #AI Agents#LLMs#APIs#Performance

    Adaptive inference scheduling for AI agents — per-round model, effort and provider routing for coding harnesses: a Command Code mod or a local OpenAI-compatible proxy.

    vizuhTypeScript★8

  • C choosekit Wraps llama.cpp and OpenRouter models as typed choices with probability distributions, not a bad idea, but a batch of same-template TypeSafe/Jev repos appeared the same day, looking more like coordinated volume. The README doesn't clearly explain the API or usage, so it's worth only a glance for now. #LLMs#APIs#SDKs

    Turn llama.cpp and OpenRouter models into typed choices with probability distributions

    NotXf1leJavaScript★11

  • C stratum A CUDA-written procedural city that generates and caches analytic geometry on the GPU, paired with layered ray visibility—a good-looking rendering experiment. Reusability is low, and it heavily overlaps the same author's WaterCuda. #GPU#3D#Web Frontend

    CUDA-authored procedural city with GPU generated, cached analytic geometry and hierarchical ray visibility.

    SamG-CoderWGSL★11

  • C jgrep It replaces grep patterns with a sentence, using Jev to judge line by line whether the semantics match, claiming 200ms per line at a thousandth of a cent. Command-line semantic search is genuinely wanted, but that latency and cost mean it only suits small files, and it's locked to a single model service. #CLI#LLMs#AI Agents

    grep, but the pattern is a description. Filters lines by meaning with TypeSafe's Jev decision model: ~200 ms and a thousandth of a cent per line.

    keltokhyPython★12

  • C Academic-Homepage-Template-by-Runqi A config-driven academic homepage template in plain HTML/CSS/JS plus YAML, with no build step. Good enough for researchers wanting a quick personal site, but such templates are everywhere and the only highlight is skipping package managers. #Web Frontend#Learning#Documents

    A polished, configuration-driven academic homepage for GitHub Pages. It uses plain HTML, CSS, JavaScript, YAML, and Markdown—no package manager or build step required.

    runqi-allen-wangCSS★15

  • C qbs Has AI find books around a question, read full chapters, and produce a runnable Skill, and linking "reading books" with "building skills" is somewhat interesting. But the flow is lengthy with no sample output, so readers can't judge whether what it reads is trustworthy. #LLMs#Learning#Agent Skills

    QBS: start from a question, let AI find books, read the relevant chapters in full, and create and trial-run a Skill. You can begin even if you're unfamiliar with a field.

    LearnPromptPython★16

  • C gp-ladder A Rust desktop workbench for Tonghuashun's consecutive-gain rankings, a concrete need for A-share short-term traders. But it depends on Tonghuashun for data, does one thing, and lacks usage docs, and similar small tools are plentiful. #Fintech#Data Analysis#Windows

    Consecutive-Rise Ladder: a Tonghuashun consecutive-rise ranking desktop workbench

    uasierRust★9

  • C fast-browser-use It claims to do browser operations with a local small model, speeding things up via "single-token reflex" and claiming zero hallucination; such absolute wording is itself a minus. No task success rates or comparison data are given, so readers can't tell how it beats mature browser agents. #AI Agents#LLMs#Automation

    A fast browser-use skill powered by local LLMs via single-token reflexes. Fast, local-first, zero hallucinations.

    APUS-AI-LabPython★8

  • C Minimax-H3-Latent-Continuation It claims MiniMax-H3 multi-shot long video will 'stay 100% consistent and completely eliminate degradation'—classic overstatement. No comparison samples or parameters are given, so actual results can't be judged. #Video#AI#Multimodal

    This suite enables **long, seamless multi-shot video continuations** with **100% visual context and environment consistency**—completely eliminating generation-to-generation quality degradation!

    SatoDivePython★14

  • C typesafe-sdk-go A Go SDK for the TypeSafe AI API, one of the TypeSafe/Jev series that appeared the same day. The interface description is fairly clear, but the service itself lacks independent evidence, so the SDK's value can't be established. #SDKs#LLMs#APIs

    Go SDK for the TypeSafe AI API — typed questions in, probability distributions out.

    TangergGo★8

  • C wallpaper-engine-hotswap A small GUI for hot-swapping wallpapers in Wallpaper Engine on Windows—a real but very narrow need, and it doesn't say whether it requires the paid app. It saves a few dozen seconds of clicking, but it's still just a small tool. #Windows#Productivity#Images#wallpaper-engine

    Nice GUI to hotswap wallpapers on windows using wallpaper engine

    VibeCodingBeLikeRust★8

  • C xtags Labels each post in your X timeline with what it wants you to do, with judgments from Jev. Motive detection is a more original idea than most peers, but it relies entirely on a brand-new decision model, and neither accuracy nor the browser-side implementation is explained. #Social#LLMs#Browser Extensions

    On the X timeline, label what each post wants you to do. The judgment comes from Jev, a model that only returns probabilities and generates no text.

    manifoldorJavaScript★8

  • C jev-seo A Rust SEO/GEO CLI plus MCP service pulling data from DuckDuckGo, self-described as a free Semrush replacement. The marketing is over the top (₹0, replacing, System One), with no explanation of how metrics are computed or how far it falls short of real SEO tools; for now it's just a demo. #Marketing#MCP#CLI#Open-Source Alternatives#semrush

    100% free ₹0 agent-first SEO & GEO CLI suite and MCP server in Rust replacing Semrush and OpenSEO via DuckDuckGo and TypeSafe Jev System One

    AkashPriyadarshiiRust★10

  • C jev-capability-atlas It tries to map where Jev fails using real API call records, close in form to an honest failure-case collection, but the evaluated subject itself lacks independent credibility, so the evidence rests on shaky ground. #LLMs#Benchmarks#ML Research

    Independent, evidence-based map of when TypeSafe's Jev actually holds up vs. breaks down — real API-call receipts, not a leaderboard. A primarily Chinese bilingual repo.

    ZaiousPython★18

  • C jev-mac-voice Full-duplex voice control on macOS using OpenAI Realtime plus native accessibility, with Jev as the backend. The idea isn't bad, but the dependency lacks independent verification, and similar voice-control solutions are already plentiful. #Speech#macOS#AI

    English full-duplex voice control for macOS with OpenAI Realtime, native Accessibility, and Jev.

    brudarkoJavaScript★16

  • C jev-voice On Mac it turns a spoken sentence into a system action using local whisper.cpp plus one Jev call, and the voice entry point is a real benefit. But it hard-depends on TypeSafe's paid API, and neither the executable commands nor failure handling are explained, making it a more pragmatic but still early entry in the same-prefix wave. #Speech#macOS#Automation#AI Agents

    Talk to your Mac. Local whisper.cpp + one Jev (TypeSafe) call per command + macOS automation.

    kevinbadiPython★20

  • C css-3d-lab A demo collection explaining CSS/SCSS 3D effects, with copyable snippets and a pure-CSS versus JS comparison, neatly written. But it's essentially teaching material, not an install-and-use tool, and those who need it have likely learned it elsewhere. #Web Frontend#Learning#3D

    Learn 3D effects in CSS/SCSS - live demos, explanations and copy-paste snippets. Pure CSS vs CSS + JS.

    EvirtualTypeScript★8

  • C iPhone-Duo-by-Examples A SwiftUI sample set demonstrating the iPhone Duo foldable's new APIs, clearly written but built on unreleased system interfaces. It'll be worth referencing once real devices ship; for now it's more an extension of the author's fun projects. #iOS#Learning#UI Design

    SwiftUI examples for the iPhone Duo (foldable iPhone) APIs in iOS 27.1: hinge, reserved regions, ArrangementView, vertical toolbar

    artemnovichkovSwift★17

  • C pi-duck-mode A Pi extension claiming 14x token savings and 798x faster TPS—numbers so exaggerated they become unbelievable, with no measurement conditions given. A typical marketing-style README. #LLMs#Prompting#Productivity

    pi extension to spend 14x less tokens and achieve 798x more TPS

    btseytlinTypeScript★13

  • C invalidate It attaches a "lease" to each fact in AI memory, expiring when new evidence arrives; the concept is interesting but the description stays at the metaphor level with no interface examples. The whole cluster of repos appeared the same day, which undercuts credibility. #AI#LLMs#RAG

    The invalidation layer for AI memory. Every fact gets a lease; new evidence ends it. Built on TypeSafe Jev.

    chopratejasPython★11

  • C jev-use A Mac computer-use tool that reads the screen via the accessibility tree instead of vision models. Reading accessibility rather than screenshots is worth discussing, but it doesn't explain installation or supported apps, and several similar harnesses already exist. #macOS#AI Agents#Automation#LLMs

    Say it, and your Mac does it. A computer-use harness on Jev that reads the screen through Accessibility. Fast, no vision model

    savka777Swift★25

  • C Kcoder Promises model integration, a tool system, sandbox permissions, long-term memory, skills, MCP, subagents, a TUI, and a desktop app all at once, a classic do-everything project. With 8 stars against such broad ambition and source published as snapshots in Releases, it's hard to believe it's usable. #AI Agents#LLMs#CLI

    KCoder — agentic AI coding platform by KunlunMeta: model providers, tool system, permissions & sandbox, long-term memory, skills, MCP, sub-agents, terminal TUI and KCoder Studio desktop app. Source sn…

    KunlunMeta-devRust★8

DSkip123
  • D awesome-jev-tools A tool list for the Jev ecosystem whose main value is proving that a dozen typed decision projects emerged within two days, a good momentum signal; but the list itself has no technical moat, its inclusion standards and update speed decide its fate, its competitor is logicrw's similar list, commercialization is nearly zero, and it only suits being a community entry point. #LLMs#AI Agents

    A curated list of tools built for Jev — TypeSafe AI's System One model for typed decisions.

    v-modalLanguage unknown★185

  • D sub2api-state-kit Builds account-level state maintenance, fixed proxy re-verification and anomaly guarding around Sub2API, essentially raising the survival rate of shared or resold API accounts, a gray area under terms of service; 92 stars come from gray-market demand, scaling will trigger platform risk controls and bans, and it has no sustainable business model. #LLMs#APIs#Networking

    Sub2API account-level STATE extension: Pro/Team selection, fixed proxy re-verification, and dynamic guarding against 312 anomalies (experimental version)

    wangyunjeffGo★92

  • D yowes Generates teacher IDs, licenses and official letters from 13 countries via MCP, aimed squarely at forging documents, and is technically just templates plus image compositing; it carries high legal risk, will be banned by platforms, has no sustainable business model, and is a gray-market project to avoid no matter how popular. #Scams#MCP#Documents

    Generate realistic teacher documents (ID cards, licenses, letters) for 13 countries via MCP.

    hirotomasatoPython★61

  • D awesome-jev-projects A radiated list of Jev projects with automatic GitHub sync, showing the ecosystem is spontaneously forming a directory layer and serving as momentum evidence; but list projects have no moat and no revenue, and inclusion bias can mislead selection, making it nearly worthless as an investment target. #AI#Learning#jev

    Awesome Jev: source-backed open-source ecosystem radar, plain-language project discovery, and automatic GitHub sync

    logicrwJavaScript★112

  • D how-to-live-better A zero-dependency single-page version of the 528-entry 'High-Value Life Guide,' readable, searchable and offline on mobile, it is content repackaging plus layout work, with 62 stars from social sharing; it has no technical or business moat, competes directly with paid knowledge products and ebooks, and is not a developer tool. #Knowledge Base#Documents#Learning

    Online single-page reading edition of all 528 entries from the book "High Cost-Performance Life Guide": readable on mobile, searchable, zero-dependency, supports offline

    cdyforeverHTML★62

  • D jev-rules Selects applicable rules per prompt and feeds them to Claude, a common rule-retrieval approach. But it gives no selection strategy, mis-selection cost, or evaluation, and shares its origin with several other jev repos in this batch, looking like different slices of one package. #LLMs#Prompting#AI Agents

    Jev picks which of your rules apply to each prompt, so Claude only sees the ones that matter.

    EliaAlbertiJavaScript★34

  • D streamlit-ollama-chat A local chat interface built with Streamlit and Ollama, with model selection and JSON history; the description is accurate but it's an introductory project everyone has written, more valuable as a tutorial than a tool, with nothing worth sharing. #LLMs#AI#Self-Hosted

    Chatbot local com Streamlit e Ollama, seleção de modelos, streaming de respostas e histórico persistente em JSON.

    EricoBorgonovePython★13

  • D AirCard-Windows A Windows branch of AirCard that fixes the iOS 26 handshake and includes an Indonesian tutorial, meaningful only to Indonesian users who want to use Android phones as AirDrop. The change is essentially one patch that could go upstream, so a separate fork has little long-term maintenance value. #Windows#iOS#Networking

    Windows AirCard fork: Grappa handshake fix for iOS 26+, batched flashes. Tutorial Bahasa Indonesia.

    dhava-gautamaRust★10

  • D alarm_v3 An ESP32 door alarm with matrix keypad, PIR, OLED and ntfy push, with a clear bill of materials. A typical personal DIY project; the v3 naming suggests it's for the author's own front door, so others can only reuse the wiring ideas. #IoT#Embedded#Hardware

    ESP32 alarm system with keypad control, PIR motion detection, OLED display and ntfy notifications.

    erenbkdsC++★8

  • D post-production-skill A prompt skill pack for AI video post-production that outputs descriptive terms for VFX, transitions and 3D UI. Ready-made material for short-film makers, but it's still a prompt collection, and the model version numbers it cites don't yet exist, so the claims need discounting. #Video Editing#Video#Prompting#Agent Skills

    A Seedance 2.5 Skill for AI video post-production VFX creation, able to generate cinematic VFX, creative transitions, 3D UI, dynamic camera and visual compositing prompts.

    huangbai-AILanguage unknown★52

  • D Campus-Flow An auto-login tool for the Hebei University of Water Resources and Electric Engineering campus network, written in C#. Handy for students there using mobile routers, but extremely narrow, essentially a common auto-dial script, and useless outside that school. #Automation#Networking

    Automatic campus network connection, directly usable for China Mobile routers at Hebei University of Water Resources and Electric Engineering

    zuijiu888C#★15

  • D ai-voice-agent A multilingual voice outbound agent for study-abroad loan eligibility pre-screening, with a long feature list, but it's clearly an outsourcing or demo project lacking reproducible deployment instructions and general applicability. #Speech#AI Agents#LLMs

    A multilingual AI voice agent that automates education loan qualification for study-abroad students. Features intelligent call handling, real-time sentiment analysis, lead qualification, and seamless …

    cstar0521Python★8

  • D agent-dispatcher Capability-aware task dispatch for Claude Code with 27 expert roles, skill combinations, and MCP routing. The role count itself signals padding, and with no evaluation showing dispatch beats just asking directly, it easily degrades into a prompt collection. #AI Agents#MCP#Agent Skills#Productivity

    Capability-aware Claude Code dispatcher with 27 specialist roles, composable skills, MCP/tool routing, verification workflows, and direct role commands

    nahid-sparktalesPython★14

  • D jev_local It replicates some "typed decisions" model with a local LLM, but gives no evaluation results, hardware requirements, or differences from the original implementation, so whether the replication holds can't be verified. An isolated replication project with too little information to justify a click. #LLMs#AI#Benchmarks

    Replicating Jev with a local LLM

    Argos1111Python★14

  • D ai-girlfriend-v2 A local AI companion for Windows emphasizing long-term memory, local voice cloning, and an optional digital human. The direction is heavily homogenized, and the README doesn't clarify models or privacy boundaries—more a concept mashup. #AI#Speech#LLMs

    Windows AI companion with DSH long-term memory, local voice cloning and optional WSL2 digital human.

    Work-FisherPython★8

  • D swift-typesafe An unofficial Swift SDK for TypeSafe, appearing in the same wave of TypeSafe repos that day. API coverage and integration are unclear, and its unofficial status means maintenance is uncertain. #SDKs#iOS#LLMs

    Unofficial Swift SDK for TypeSafe

    ainameSwift★8

  • D Discord-Zapret A Zapret config pack tuned for Discord—essentially a ready-made parameter set for DPI bypass, with very little originality. Useful to users in specific regions, but it offers no explanation of how it works and no update commitment. #Networking#Network Security#Windows#discord

    An optimized and stable version for Discord. Minor issues fixed.

    alaaqutfaBatchfile★39

  • D cryptoterm A crypto and meme coin market terminal in Go, with Binance WebSocket plus Solana/Base on-chain tracking, alerts and positions. The features are all off-the-shelf assembly, and the lightning emoji and "Ultra-fast" tone cost it points upfront; similar TUI tools already abound. #Crypto#Terminal#CLI

    ⚡ Ultra-fast, real-time crypto & meme coin terminal. Live Binance WebSocket tickers, DEX meme tracking (Solana, Base), sparkline trends, price alerts & portfolio tracker.

    fchyogaGo★8

  • D ai-engineering-interview-questions-company-wise A company-organized AI engineer interview question bank; 45 stars show demand. But it's a pure content repo whose value depends on the author's collection rather than code, and similar lists abound, so it's fine for practice but not worth studying. #Learning#LLMs#AI#Knowledge Base

    Your Cheat Sheet For AI Engineering Interviews at Top AI Companies - Questions and Answers.

    pallavi-shekharMarkdown★45

  • D mint-desk An NFT monitoring and auto-mint tool for Windows aimed at users watching whitelists. It's an automated front-running script with no compliance or risk warnings whatsoever. #Crypto#Automation#Blockchain

    Mint Desk V1.0 — Windows local NFT monitoring and auto-mint, supporting multi-project waiting and coordinated sending from the same wallet.

    cryptoresetlifeJavaScript★8

  • D K7R3 A practical handbook covering reconnaissance, validation and reporting in bug bounty workflows, a structured set of notes for beginners. But it's a pure content repository with nothing new methodologically, and free material of this kind is everywhere. #App Security#Exploits#Learning

    A practical bug bounty field guide covering recon, vulnerability research, attack scenarios, exploitation, validation, and reporting.

    fawadqureshi007Language unknown★12

  • D swe-agents Claims to be a portable engineering reference for Codex/Claude Code/Copilot, but it's really a collection of materials and templates. It can save newcomers some fumbling, but the content is thin with no executable output—documentation, not a tool. #AI Agents#LLMs#Learning

    Portable Codex, Claude Code, and GitHub Copilot agentic engineering reference

    startmeupaiJavaScript★10

  • D von Claims to be an open-source System One decision model, sub-15ms and non-autoregressive, pitched as a local replacement for TypeSafe Jev. The metrics are specific, but 41 stars in a day with no paper or reproducible experiments is thin evidence for a 'new paradigm model'. #LLMs#ML Research#AI

    The open-source System One decision model. Sub-15ms, non-autoregressive, local drop-in alternative to TypeSafe Jev.

    wfzyxPython★41

  • D jev-paint A small JS tool for drawing with Jev, described in one sentence with no interface or output details. It's one of the crowded Jev-series repos, with too little information to judge its value. #AI#Image Generation#LLMs

    Use Jev to make art!

    achimalaJavaScript★36

  • D dashcope-sdk-python It claims to be a Python SDK for Alibaba Cloud's DashScope Model Studio, but an official implementation of the same name already exists, and the org name and repo shape suggest a mirror or name squat. Forty-seven stars in two days isn't enough to prove real adoption. #SDKs#LLMs#APIs#dashscope

    python sdk for dashscope [model studio](https://www.alibabacloud.com/en/product/modelstudio?_p_lc=1)

    agentscope-ai-javaPython★47

  • D copilot Yet another AI coding agent CLI, described in one self-introductory sentence with no model choices, tool calling, or differences from existing CLI agents. The space is extremely crowded, and it offers no reason to pick it. #AI Agents#LLMs#CLI

    Apex Copilot CLI — AI coding agent

    apexaccsTypeScript★16

  • D jev-sift A portable agent plugin and MCP tool for batch text classification, filtering before reading. The token-saving idea is reasonable but stays at the slogan level, with no data on classification quality or actual savings, and it too carries this week's jev prefix. #MCP#AI Agents#Plugins

    Classify first. Read selectively. A portable agent plugin and MCP tool for batch text classification.

    kbhuwJavaScript★45

  • D sigillo It turns interactions between people and AI companions into "sealed receipts" fed back to the model, a concept more like an essay than a tool: no inputs/outputs or integration are explained, and readers still don't know what to do with it. #AI#LLMs#Prompting

    A wax-sealed feedback receipt between a human and their AI partner — rate it, seal it, let the agent learn without turning love into a checklist. CC BY 4.0

    29-CuJavaScript★14

  • D jev-cli A command-line tool for some typed decision model that, beyond calling an interface, doesn't say what it does or outputs. It appears alongside several other jev repos in this batch, looking more like systematic volume around an unfamiliar term. #LLMs#CLI#AI

    Command-line tool for TypeSafe's Jev AI model

    Nasrallah-ALTypeScript★11

  • D cursor-cpa-plugin A Cursor enhancement plugin for CLIProxyAPI that adds streaming of reasoning content and system constraint injection. Its audience is heavy users already on the forwarding proxy; the plugin itself is thin, openly admits AI-assisted development, and shows no maintenance commitment. #Plugins#Editors#LLMs

    Enhanced Cursor plugin for CLIProxyAPI with reasoning content streaming, system constraint injection, and official brand logo (AI-assisted development)

    kilolonionGo★9

  • D ts-browser-agent A browser agent built on langchain-typesafe with a one-line description that doesn't say which sites it supports, what actions it can take, or its success rate. Projects like browser-use are already mature, so there's no reason to pick this one. #AI Agents#Browser Extensions#LLMs

    Fast browser agent on langchain-typesafe

    ndreznPython★21

  • D niimbot Named 'Niimbot B1 unlocked binary,' it appears to be unlock firmware or a binary for a label printer, with no explanation of how it works and no risk warnings. It's a small tool for bypassing vendor restrictions, and the code can't be audited. #Crack#Hardware#Embedded

    Niimbot B1 unlocked binary

    ThreeDaPrintPython★23

  • D LLM-Training-Handbook Titled as an LLM training handbook, it's actually a collection of notes. It might save beginners a few searches, but with no code and no reproducible workflow, its value rests entirely on the author's expertise, and the repo offers no credentials. #LLMs#Learning#ML Research

    h9-tecLanguage unknown★15

  • D jev-recruiter A LinkedIn recruiting agent based on Jev that browses candidates, saves links, and gives evidence against job requirements. Recruiting has real demand, but scraping compliance and what Jev actually is are both unaddressed, so treat it as a concept demo. #AI Agents#LLMs#Jobs#Automation

    A Jev powered LinkedIn recruiting agent. Watch it browse relevant profiles, save links, and review evidence against your hiring brief.

    skeptrunedevPython★30

  • D Physics-Lab-Report-Maker An auto-generation Skill for physics lab reports at Harbin Institute of Technology, Shenzhen, targeting one school and one course. Convenient for classmates there, but it is essentially templating a single assignment, so it is useless for other people or courses. #Agent Skills#Documents#Learning#Automation

    A foolproof Skill for producing HITSZ physics lab reports.

    tech-catsPython★11

  • D growing-space-vr A VR experience that guides drawing for stress relief, described only with a one-line vision and no engine, platform, or run instructions. Such work needs a demo video to be worth discussing; currently 9 stars. #3D#Health#Animation

    GrowingSpace — A Guided VR Art Journey for Stress Relief

    vart-labLanguage unknown★9

  • D jev-chat A tool-calling chatbot built on Jev that explicitly avoids LLMs, an interesting premise. But the repo never explains how Jev decides which tool to call, and its 42 stars look name-driven, so usability cannot be judged. #LLMs#AI#APIs

    A tool calling chat bot built with Jev and no LLM.

    w3cjTypeScript★42

  • D openJev-verdict-2.0 Claims a 151M non-autoregressive model beats a similar system on a self-made benchmark, but the reported accuracy and ECE only hold for its own dataset, with no reproducible scripts or baseline details. The "verdict 2.0" in the name makes it look like a sequel to the same hype. #LLMs#Benchmarks#AI

    Calibrated 151M Non-Autoregressive Decision Engine beating TypeSafe Jev & Laya on LocalLLaMA/typed-decisions (77.10% acc, 0.0636 Brier, 0.0144 ECE)

    Heman10x-NGUPython★30

  • D stratum-hyperdetail Another city-rendering experiment in the same series, with even the description left blank, so content can only be inferred from the repo name and the author's other projects. No docs and no comparisons make it not worth the time as a standalone project. #GPU#3D#Web Frontend

    SamG-CoderWGSL★7

  • D sub2api-codex-ticket A rewritten Sub2API that adds per-model Codex ticket control—essentially account-pool scheduling, with clear compliance risk. The README only mentions the maintainer and doesn't cover deployment or limitations. #APIs#LLMs#Automation

    Sub2API modified edition with persistent per-model Codex ticket controls, maintained by 2679373771.

    TuavaGo★9

  • D open-alternative-jev It bills itself as an open-source alternative to TypeSafe Jev, producing typed decisions in a single forward pass. If Jev really exists this positioning has value, but the repo was born during the same-prefix spam wave, and the benchmarks are self-written, so nothing can be verified for now. #LLMs#AI Agents#Benchmarks

    Open-source alternative to TypeSafe's Jev: a System One style model layer that gives typed, calibrated decisions from any open-weights LLM in one forward pass (HF + vLLM), with honest benchmarks

    ikermoelPython★11

  • D qijing-ai-guide AI prompts, templates and teaching examples for Chinese fan fiction writing, aimed at writers. A pure content repository whose value depends on template quality, yet it offers no sample outputs or scope, no different from writing prompts scattered everywhere. #Content Writing#Prompting#Learning

    Wonderland AI Chinese fanfiction writing guides, creation templates and teaching examples | Chinese fanfiction writing guides and templates

    jesse-engineerLanguage unknown★27

  • D DrClojure-Neo A new repository claiming to be a Clojure IDE, with 11 stars in a day but only a one-line description: no screenshots, and no word on how it differs from CIDER or Calva. An IDE is a long-cycle undertaking, and a claim of this size is hard to believe is already usable. #Editors#Plugins#UI Design

    Clojure IDE

    kimtgClojure★11

  • D cfo Wraps ChatGPT/Claude as "your CFO," essentially prompts plus role-play. Financial decisions have no credibility on this, and it solves no concrete engineering problem—a typical big-word shell. #Prompting#Fintech#LLMs#AI

    Turn ChatGPT or Claude into your own CFO (Chief Financial Officer).

    rohasnagpalPython★14

  • D localjev It has 401 stars in a day yet not a single line of description, and shares its name root with the batch of jev-prefixed repos that appeared this week. Repos that don't explain their purpose get no recommendation; the star count here looks more like an anomaly signal than evidence. #LLMs#AI Agents#typesafe

    githubnextTypeScript★401

  • D hermes-jev-skills A set of skills for Hermes (and Claude Code, Codex) covering model routing, memory, compression, skill selection and browser operation. It blankets the entire agent pipeline at once, which is exactly why nothing is done thoroughly, and there's no evidence of effectiveness. #AI Agents#Agent Skills#LLMs

    Jev-powered model routing, memory, compaction, skill selection, computer and browser use for Hermes agents (also Claude Code and Codex)

    kerpopulePython★30

  • D sub2api-state-reuse A state-reuse plugin for Sub2API, plus Clash IP collection and admin monitoring. The whole thing points to grey-area sharing of subscription quotas, and the docs read like internal notes, so its purpose and compliance make it unfit for a recommendation slot. #Networking#APIs#Automation

    Sub2API STATE 292/332 reuse plugin, Clash/IP collector and administrator live monitor, with deployment and customization guides

    little-greenbeanGo★27

  • D graphite A design Skill that claims to build feature web pages, but the one-liner never says which agent it is for and gives no sample output. Design Skills suffer most from vagueness, and this one does not even clarify its output format. #Agent Skills#UI Design#Web Frontend

    A design skill for creating monographic webpages!

    thtbeeTypeScript★16

  • D awesome-jev Yet another awesome list, collecting examples around a barely recognizable term. List repos don't produce tools themselves, and its appearance alongside several jev repos in the same batch looks more like coordinated volume-filling than an independent project. #Learning#Prompting#AI

    Curated Jev resources and runnable examples for typed AI decisions.

    AppitStudioPython★41

  • D VBA_Automation_Engine A VBA automation engine with no description, not saying whether it targets Excel, Access, or another host, and no sample macros. It might help those still maintaining Office macros, but the repo itself offers no information to judge by. #Automation#Spreadsheets#Windows

    GrzegorzDokowiczVBA★14

  • D do-not-safety-me A two-sentence description about refusing before history and retrying from a clean context, not saying which model it targets or how it differs from a normal context reset; with the repo name it looks like a small safety-bypass experiment, with unclear value and risk. #LLMs#Prompting#Jailbreak

    Reject before history. Retry from a clean context.

    GwendolenmaveLanguage unknown★17

  • D Nai64ExtraPatches An undescribed Kotlin patch repo whose name only suggests a connection to some 64-related project, with nothing on what it patches or how to apply it. It's completely unusable to anyone unfamiliar with the author's other repos. #Games#Android#Plugins

    Nai64Kotlin★16

  • D Dawn-installer An installer and launcher for Dawn, but what Dawn is and what you can do after installing is never explained, leaving cross-platform support as the only useful information. A wrapper script whose purpose can't be judged is worse than just using the upstream release package. #Windows#Linux#DevOps

    An installer and launcher for Dawn. Cross-platform Windows and Linux support

    echo-mattRust★12

  • D EX520v-tek-tus-root-araci A Turkish "one-click root" tool for a specific device model with only a one-line description. It states neither the exploit chain nor applicable firmware versions, making it high-risk for ordinary users and nearly information-free for researchers. #Android#Exploits#Hardware

    EX520v için tek tuş root aracı

    emryclPython★11

  • D streamlit-ollama-chat-templates A template offshoot of the previous repo with no description, so it's unclear which templates it offers or how to use them. A consecutive practice output from the same author, it adds nothing for readers on the list. #LLMs#AI#Learning

    EricoBorgonovePython★11

  • D typesafe-mcp An MCP service written in Go, but the README doesn't explain what problem it solves or how to connect it to a client, leaving only the name to guess from. It appears amid a cluster of TypeSafe repos and lacks a verifiable use case. #MCP#LLMs#AI

    PyModelGo★21

  • D ntlmscout A Python repo with only a name and no description, guessed from naming to relate to NTLM reconnaissance. Its function, usage, and legality boundaries can't be confirmed, so it's not advisable to recommend. #Network Security#App Security#Exploits

    boydhacksPython★18

  • D jevcache It claims to provide decision caching for TypeSafe Jev-style models in a single 2MB file, avoiding recomputation for repeated decisions. The problem framing isn't fantasy, but it depends on a model referenced by dozens of repos this week, and neither the implementation nor usage can be verified. #LLMs#AI Agents#Performance

    A decision cache for TypeSafe Jev-class models — memoize decisions so repeats are free, deterministic, and shareable. One 2 MB binary.

    hyperspaceaiLanguage unknown★15

  • D jev-macos-loop Native macOS GUI automation on Apple silicon using Jev plus OmniParser, requiring your own OpenRouter and other tokens. The idea isn't new, many computer-use projects already exist, and it outsources reproducibility to a third-party paid gateway. #macOS#AI Agents#Automation

    Open-source macOS AI computer use and native GUI automation on Apple silicon. Jev + OmniParser CoreML + Apple Vision OCR. Bring your own OpenRouter, Vercel AI Gateway, or TypesafeAI token.

    jcpsimmonsJavaScript★13

  • D jev-reflex-autonomy-lab A multi-drone autonomy testbed using Jev for reflexive decisions and optional second-order policies. The name is grand, but there's no simulation environment description or flight logs, so drone people get nothing reproducible. #Robotics#AI Agents#LLMs

    Multi-drone autonomy lab demonstrating TypeSafe Jev reflex decisions with optional System 2 strategy guidance.

    khordooTypeScript★8

  • D transcript-lens A Turkish YouTube subtitle analysis tool that uses Jev for summaries and exports subtitles. Serving Turkish users is a niche, but it's just transcription plus summarization, overlapping many existing tools and tied to a model of unknown origin. #Video#LLMs#AI#Learning

    YouTube transkriptlerini anlamına göre keşfedin. Türkçe arayüz, Jev analizi, altyazı dışa aktarma ve Vercel kurulum rehberi.

    sensahinTypeScript★10

  • D vllm-jev-decison Adds "classification-only limited-mode scoring" to vLLM, and even misspells decision in the repo name. Classifying without generation isn't a bad idea, but it's all jargon with no usage or results—leftovers from this same-name wave. #LLMs#ML Research#AI#APIs

    Classification-only typed decisions for vLLM: finite-schema candidate scoring, probabilities, and abstention. No generative fallback.

    siliconkernelPython★8

  • D codex-state-kit An empty repo with no description, only clues like Rust and "state kit." It doesn't say what state it manages or how to integrate it, leaving nothing to evaluate, and similar small state-management tools are already everywhere. #AI Agents#CLI#Debugging

    DouDOU-startRust★13

  • D QwQramdisk It lets 64-bit iOS 7/8 devices mount a writable ramdisk and handles SEP, falling under old-version jailbreak research. Of some reference value to retro iOS enthusiasts, but it's dated, undocumented and serves only a tiny circle. #Jailbreak#iOS#Exploits#Reverse Engineering

    A Ramdisk can offer iOS7-8 64Bit /mnt2(user space) R/W , unlock the alive SEP

    j1ansPython★9

  • D bonsai-apple-silicon Packages and publishes llama-cli binaries for Apple platforms with no note on source changes. Users really want the upstream llama.cpp release page; repost repos like this only muddy version tracking. #macOS#LLMs#GPU

    llama-cli release for apple

    l0d0v1cLanguage unknown★17

  • D jev-lint jev-lint has no description, so you can only guess from the name that it's some kind of lint. It appeared alongside a batch of same-prefix repos and explains neither rule sources nor workflow integration, so it can't be evaluated. #Code Quality#LLMs#Plugins

    mizchiTypeScript★9

  • D x-scanner Labels each post in your X timeline with a Jev verdict and shows a cost counter, but nobody explains what Jev is and the extension itself does very little. A small toy riding a suddenly popular model name—not worth your time. #LLMs#AI#Browser Extensions#Social

    Chrome extension that labels every post you scroll past on X with typed Jev judgments and a live cost counter

    oso95TypeScript★12

  • D jev-social Uses Jev to scrape Instagram/TikTok/LinkedIn content for research, claiming to bring back evidence and citations. The feature list is long but never shows what output looks like, and the compliance risk of such scraping far outweighs its value. #AI Agents#LLMs#Social#Automation

    Jev-powered Instagram, TikTok, and LinkedIn research: typed routing, real browser evidence, streamed post cards, video capture, and cited socai reports.

    socai-ioJavaScript★8

  • D patdown Uses 'Jev' as a fuzzy linter to health-check code, an amusing metaphor, but it never explains the rules or how false positives are handled. It rides the wave of same-named projects, and mature solutions already exist for code quality. #Code Quality#LLMs#AI

    typesafe's jev as a "fuzzy linter". give your code an ocular patdown.

    tyler-dot-earthTypeScript★8

  • D HyperDock The repository has no description, just a Kotlin project name and 10 stars within a few days; there's no telling what it does or who it's for, and no runnable entry point is documented. No reason for readers to spend time guessing. #Android#Productivity#Hardware

    1812zKotlin★10

  • D oxygen-ui-solana An empty-description Solana UI component library with no indication of who it's for or how it differs from existing libraries. Eleven stars in half a day provides little signal either. #Blockchain#Web Frontend#Crypto

    Subhan-codeTypeScript★11

  • D TypeSafe A Swift TypeSafe SDK that is blank apart from the title: no interface shape, no proof it runs, no integration guide. It lacks even a minimal usage example, so it can't be evaluated. #SDKs#LLMs#AI

    TypeSafe SDK in Swift

    krzyzanowskimSwift★9

  • D live-jev A browser-based 2D self-driving car simulation using Jev for decisions. Fine as a teaching demo, but with no description or screenshots its value rests entirely on the code, and 10 stars suggest few really care. #AI#Games#Web Frontend

    2D autonomous car simulation in the browser, driven by TypeSafe's Jev decision model

    vinilanaJavaScript★10

  • D CivicPulse-AI An empty-description repo whose name only suggests political data analysis. With no data sources, analysis goals, or run instructions, it's a typical coursework-style project from which readers can get nothing usable. #AI#Data Analysis#Social

    Anubhavsingh311Jupyter Notebook★8

  • D Aryx_F22E_StrikeRaptor An undescribed C# project that from its name looks like a 3D or aircraft asset or small game, but with no text, screenshots, or run instructions, readers can't tell what it is or get it running in ten minutes. #Game Dev#3D#Animation

    Aryx3DC#★8

  • D cross-asset-market-network An undescribed notebook repo whose name points to cross-asset market network analysis, but with no data source, method, or charts, it can't be reproduced or evaluated and is an unfinished research draft. #Fintech#Data Analysis#Crypto

    GoodyNickJupyter Notebook★10

  • D RoboICL Claims embodied context learning for "GPT-6 Astra," a model that doesn't exist, so the description rests on a fictional premise with no method or experiment details. Star counts on such repos aren't evidence, and the content isn't worth reading. #ML Research#Robotics#AI

    Embodied In-Context Learning for GPT-6 Astra

    Mosi-AILanguage unknown★12

  • D skills Three explanation and summarization agent skills (eli5, concise version, summary) that got 39 stars in half a day. They're essentially a few prompts that anyone could write, with no evaluation or integration notes, worth bookmarking more than using. #Prompting#Agent Skills#Learning

    Skills for explaining and summarising AI output: eli5-succinct, simple-summary, summary

    isas1Language unknown★39

  • D awesome-jev A resource list around TypeSafe Jev, appearing right in the window when repos with that prefix are flooding in. The list itself is low-density and covers an unverified ecosystem, so it offers readers no independent value. #LLMs#Learning#Knowledge Base

    A curated list of Jev use cases, projects, SDKs, and resources. Jev is TypeSafe AI's System One model for fast, typed decisions in software — Choice, Score, and Noul with calibrated probabilities.

    kraayenjonLanguage unknown★31

  • D jev-mail-classifier Uses the Jev model to label, move, and remind about emails based on config. The config-driven approach is clear, but it depends on a new model of unknown origin and doesn't say how it beats plain rules—one of the thinner entries in this same-name wave. #LLMs#AI#Email#Automation

    Classify your inbox with Jev (TypeSafe's System One model) — tag, move, flag, and notify, all config-driven.

    parth-kpPython★8

  • D jot Claims to be "the first general System One agent based on Jev," but the one-liner is all stacked nouns with no specifics on what it does or how to start. A whole batch of Jev-named projects appeared these two days, and this is among the emptiest. #AI Agents#LLMs#AI

    The first general-purpose System One agent for Jev

    runta-devTypeScript★12

  • D agentflow-platform 'Enterprise-grade multi-agent self-evolving workflow orchestration platform MVP' packs four buzzwords into one sentence without saying what it orchestrates or how to run it. A typical vision repo, so skip. #AI Agents#LLMs#Automation

    Enterprise multi-agent self-evolving workflow orchestration platform MVP

    trickytricklolPython★8

  • D jev-gateway Over a dozen repos bearing the 'Jev / TypeSafe System One' name appeared within two days, and this one has an empty description yet already 53 stars. Whatever it intends to do, without a single sentence of explanation it is not worth a look. #LLMs#AI#APIs

    vinilanaTypeScript★53

  • D tombstone-reaper-skill The title and description are all emoji and slogans about cleaning dead code and reducing AI context pollution. The direction is somewhat interesting, but there is not one concrete rule or usage example, and the marketing tone is heavy. #Agent Skills#Code Quality#AI

    🪦 Let dead code and tombstone skills truly rest in peace, show off your subtraction achievements, and end context pollution in the AI era.

    webkuborShell★7

  • D airlift-rw There's no README, and the repo name's Activator and rw hint at some kind of activation or read-write unlock, but no text explains what it does or what target it applies to; it can't be assessed and isn't worth the risk. #Activation Bypass#macOS#Plugins

    AldazActivatorObjective-C★9

  • D devfest-natal-2026 Companion material for a DevFest talk—lecture-style content, not a reusable tool. Valuable to the live audience, but not to readers of this list. #Learning#AI#App Security

    AI DevSecOps: Segurança em LLMs e Agentes - material da palestra no DevFest Natal 2026

    ahirtonlopesJupyter Notebook★17

  • D jev-directory An HTML directory page with no description, tucked among the batch of jev-prefixed repos this week. It neither says what it collects nor offers runnable code, giving readers no basis for judgment. #LLMs#AI Agents#Knowledge Base

    everyai-comHTML★10

  • D fable-jev A TypeScript layer giving coding agents "hundred-millisecond reflexes" via Jev and get-fable. The description is all jargon and lightning emoji yet can't name one concrete use case, a typical hanger-on to the same-name-prefix ecosystem. #LLMs#AI Agents#Automation

    ⚡ Sub-100ms cognitive reflexes for autonomous coding agents. Powered by TypeSafe AI's Jev & get-fable.

    imMamdouhaboammarTypeScript★9

  • D Brute A brute-force tool for 64-bit iOS device passcodes with a one-line description, stating neither applicable versions nor principles. Tools like this are high-risk with unclear provenance, and the repo itself provides nothing to inspire trust. #Crack#iOS#Exploits#Reverse Engineering

    A tool can brute force the 64Bit Device 's Password

    j1ansC★7

  • D Game-Catalogue A binary release of a closed-source Windows game catalog browser with no source and no description. It can't be audited, and whether its resources are compliant is unclear, so it makes no sense as an open-source list candidate. #Games#Windows#Downloaders

    Closed-source Windows game catalogue browser. Binary releases only.

    kerrdec97Language unknown★16

  • D jev-hub Aggregates long posts and demo videos about Jev on X, keeping original links and authors. It's a useful index for people following the model, but it's essentially content aggregation, and amid the same-prefix spam wave,热度 doesn't equal quality. #Social#LLMs#Knowledge Base

    JEV HUB · Aggregation of long-form articles and demo videos about TypeSafe AI's "System One model" Jev on X (original links and authors preserved) | Produced by Who Is the Expert

    mizzleloverCSS★15

  • D csss The description is self-coined jargon like 'Codex Sleep State Sugar, Surge module', revealing neither the problem it solves nor any verifiable features. Skip. #LLMs#AI#Automation

    Codex Sleep State Sugar: Surge module for validated Codex turn-state capture, reuse, renewal, and status panel

    tzf1003JavaScript★9

  • D awesome-jev-typesafe A 'curated list for Jev', appearing within two days alongside a Chinese version, a plugin version, a gateway version, and more. The list creates nothing new and its sources are a mystery, so its reference value is doubtful. #LLMs#AI#Knowledge Base

    Typed decisions with TypeSafe's Jev, the first System One model

    valentynkitJavaScript★26

  • D awesome-jev-zh A Chinese resource list for Jev with daily auto-collection. List repos create nothing new, and what is collected is mostly same-named projects that appeared in the past few days, so its reference value is doubtful. #LLMs#AI#Knowledge Base

    A curated Chinese list for Jev / TypeSafe System One: official materials, SDK, hit apps, Agent tools, open-source reproductions and independent reviews, with a Chinese getting-started guide, automatic…

    yzflyHTML★13

  • D OpenGFW An empty-description repo sharing the name of a well-known existing project, with 141 stars in 0.6 days and no code explanation or clarification of its relation to the original author, most likely name-squatting or star-farming. No reason to click before the source is clear. #Network Security#Networking#Privacy

    HyNetworksGo★141

  • D dsh-plugin-dev-notes A plugin development notes repository with no description or structure, it's a personal record rather than a tool. It offers no actionable value to readers of the list. #Plugins#Learning#Documents

    cholf5Language unknown★8

  • D android-file-recovery An Android file recovery repository with no explanation, its purpose only guessable from the name. Recovery tools involve root, partition formats and overwrite risks from mistaken deletion; if the author won't explain the principles and scope, you shouldn't hand over your phone data. #Android#Backup#Reverse Engineering

    eslamfaisalKotlin★8

  • D TicTacToe A tic-tac-toe game written in CSS with 1v1 support and nothing else. A typical practice project with no AI opponent and no networking, so few besides the author would need it. #Games#Web Frontend#Learning

    simple tictactoe game....just 1v1 for now

    muthokaricky-altCSS★20

  • D shoemoneyx Only says "crypto trading desk, backtest farm, 110 exchanges" and "star if you want it open-sourced"—a typical empty repo fishing for stars. With no code or explanation, it can't be treated as a project. #Crypto#Fintech#Marketing

    SMX · the ShoeMoneyX crypto trading desk. Backtest farm, walk-forward optimizer, built-in AI agent, 110 exchanges. Star it if you want it open-sourced.

    shoemoneyLanguage unknown★7

  • D jevocks 'Use Jev to see daily stock status' gives no details at all: no data source, metric definitions, or interface. It is one of the weakest of this wave of same-named projects. #Fintech#LLMs#AI

    Everyday Stocks Status with Jev

    unicodeveloperTypeScript★8

  • D oxlint-plugin-jev Empty description; the name suggests a Jev rules plugin for oxlint. With no explanation of the rules or usage, and 12 stars in a day, it cannot be evaluated, so skip. #Code Quality#LLMs#Plugins

    wobsorianoTypeScript★12

  • D jev-desktop Plugs TypeSafe Jev's action selection into Codex Computer Use, a one-liner full of proper nouns that never says when this works better. An offshoot of the same-named projects, with too little information to judge. #AI Agents#LLMs#Automation

    TypeSafe Jev action selection inside Codex Computer Use

    yikangy873-gifJavaScript★13

  • D Microsoft-365-Cracked A Microsoft 365 piracy distribution under the banner of unlocking all premium features, with typical marketing copy that doesn't and can't explain its source or safety. Such repos only expose machines to risk and offer developers nothing. #Crack#Piracy#Windows

    The ultimate productivity suite. Microsoft 365 — Full version with all premium features unlocked. Experience advanced Office apps, cloud storage, AI tools and professional collaboration without limita…

    ClutchLarvaSpindle28Language unknown★14

  • D obisidan-flowtask-sp An Obsidian task plugin with no explanation, and the repository name is even misspelled. No screenshots or usage, and no word on how it differs from Tasks or Dataview, so it's impossible to judge whether it's worth installing. #Productivity#Plugins#Knowledge Base

    giba0TypeScript★8

  • D Multichain-Drainer Self-described as a multi-chain wallet stealer and phishing kit, with the body stuffed with SEO keywords. Such things shouldn't appear on the list, the code can't be verified, and it should be skipped outright. #Scams#Crypto#Blockchain

    A specialized software application or automated script engineered to surreptitiously or forcibly extract digital currency, tokens, non-fungible tokens (NFTs),threat intelligence reports, and the Web3 …

    OrderKomodoFosterLanguage unknown★11

  • D jevbrain A JS repo with no description whose name collides with the Jev series, leaving function and usage unclear. It can't be evaluated, so skip it. #AI#LLMs#Prompting

    SynxneuosJavaScript★9

  • D TareminCloth No description and no usage, so it can only be guessed from the name to be cloth simulation related. As a standalone project it offers nothing readable, so skip it for now. #3D#Animation#GPU

    TareminPython★9

  • D ahb_panel A panel-type repo with no description—only a name and language, making its purpose and tech stack impossible to judge. Readers can't open a project with no explanation. #DevOps#Automation#Self-Hosted

    ahb-panelPython★9

  • D agentic-crm An agent CRM repository with no description, a grand but empty name, and a README offering no usage instructions. Projects combining "agent" with an industry noun are currently the least scarce thing around. #AI Agents#LLMs#Collaboration

    cstar0521Python★8

  • D jevscan-evm Not a single line of description, so it's unclear whether it scans EVM contracts or malicious code, yet it has 18 stars in a day. The name carries jev, sharing roots with the batch of same-prefix repos this week; a repo that won't say what it does isn't worth ten minutes. #Blockchain#App Security#LLMs

    devtooliganPython★18

  • D findme The description is just "jev + fs = findme," which says nothing. The name hints at model-based file search, but with no usage or examples there is nothing to evaluate. #LLMs#CLI#AI Agents

    jev + fs = findme

    marc2332Rust★8

  • D macbrow Empty description, just 32 stars and a one-day-old repo. The name suggests a small macOS browser-related tool, but there is no information to judge it by, so skip. #macOS#Browser Extensions#Productivity

    timpratimPython★32

  • D Figma-Cracked A mirror of a cracked Figma, full of templated 'unlock all premium features' copy and no implementation. It's infringing content and clearly mass-distributed—exclude it. #Crack#Piracy#UI Design

    The ultimate collaborative design software. Figma — Full version with all premium features unlocked. Experience advanced UI/UX design, real-time collaboration, prototyping and professional design tool…

    brinkrhythmcrochetLanguage unknown★14

  • D jmtt-rm No description, no language tag, and the repo name jmtt-rm doesn't reveal its purpose. With no README saying what it does or who it's for, it falls into the unevaluable category—skip it. #Automation#CLI#APIs

    newliver666Language unknown★9

  • D third-hand A Swift repo with an empty description that got 39 stars in 0.4 days. There's no telling what it does and no verifiable usage signs, so the stars look pushed—skip it. #iOS#macOS#Productivity

    shhivvSwift★39

  • D Ready-or-Not-Ultimate-Assistant A cheat panel for Ready or Not listing 31 options like invincibility, infinite ammo, and enemies not firing. It is pure cheating, technically just memory editing, with nothing to learn and a ban risk. #Game Cheats#Games#Automation

    Ultimate assistant for Ready or Not. Features: Immortality (Player/Team/Civilians), Infinite Ammo, No Bleeding, Dismemberment Protection, Infinite Chemlight, Infinite Pepper Spray, Enemies Don't Shoot…

    tixeguPython★9

  • D Glassnode-Cracked A cracked mirror pitched as 'unlocking all premium features,' with nothing but marketing filler and no implementation details. It is both infringing and unauditable, so skip it. #Crack#Piracy#Crypto

    The ultimate on-chain analytics platform. Glassnode — Full version with all premium features unlocked. Experience advanced blockchain metrics, market insights, charts and professional crypto intellige…

    RunnerUncoverLanguage unknown★14

  • D CoinStats-Cracked A mirror of a cracked CoinStats, full of 'unlock all premium features' ad copy and no technical content. It is infringing and untrustworthy—low score. #Crack#Piracy#Crypto

    The ultimate crypto portfolio tracker. CoinStats — Full version with all premium features unlocked. Experience advanced portfolio management, real-time data, DeFi tracking and professional crypto tool…

    TarantulaChangeLanguage unknown★14

  • D aniimo-external-mod A cheat tool for Aniimo listing 15 toggles like instant capture and infinite resources. Useful for players wanting shortcuts, but it's just memory editing with no interesting tech, and game updates or bans will kill it anytime. #Game Cheats#Games#Automation

    Master assistant for Aniimo. Features: Instant Tame, Unlimited Capture Orbs, Max Bond, Infinite Resources, Super Damage, One-Hit Tame, No Aggro, Max Level, and more. 15+ functions.

    oscark-ops1901a2Python★8

  • D UPilot-docker Empty description; the repo name only hints at Docker packaging for UPilot. Not a word on what it does or who it is for, so skip. #DevOps#Automation#Self-Hosted

    xushierLanguage unknown★8

  • D Jev-cu It gathered 276 stars in about a day, yet the repo has no description, leaving its language and purpose to guesswork—a typical star-farmed placeholder. With no way to tell what it does, it isn't worth opening. #Scams#LLMs#AI

    Sac-YJavaScript★276

  • D Fahrenheit The description is only 'tick tock, poc out', revealing no target software, vulnerability type, or impact scope. Repos that do not say what they do should not be recommended, even if it really is an exploit. #Exploits#App Security

    tick tock, poc out.

    thebelxPython★12

  • D raw13g.github.io A personal homepage repo with an empty description, showing nothing reusable and no clue what the author intends. 16 stars in 0.1 days looks more like script-driven boosting—skip it. #Web Frontend#Documents#Content Writing

    raw13gJavaScript★16

Show HN 1

S

Show HN: CUA-S1 – A System One Model for Computer Use

Show HN

Trains a 700K-parameter small model for the many local decisions in computer-use (checking boxes, skipping, filling values), pulling System1-specific judgment out of the general LLM loop at 7ms versus 260ms; backed by the 24K-star cua, open-source drivers and benchmarks, and HN's 56 points show agent vendors are genuinely watching; the risk is that jev-style services become the standard first.

Hello HN! We're Dillon and Francesco from Cua. We were wondering how many computer use tasks actually need a full general purpose LLM (e.g. gpt-6-astra, claude-opus-5 etc.) to think through all their …

#AI Agents#LLMs#ML Research#Automation

▲56

Hot New Repos

  1. NandhaKishorM/laya★1246
  2. arvindear/wp2shell-PoC★562 CVE-2026-63030 & CVE-2026-60137 RCE chain proof-of-concept
  3. Continuum-AI-Corp/OrcaBonsai-27B-Uncensored★462 Runtime behavioral ablation for compressed LLMs. First target: Ternary Bonsai 2 27B — no weight modification or re-quantization. by OrcaRouter team
  4. bespokelabsai/nimble★421 Local typed decisions, contrastive data curation, and model evaluation.
  5. githubnext/localjev★401
  6. gylive/ccodex-sleep-state★332 An attempt to improve Codex's degraded intelligence, rate limiting, and connection experience. One-click local launch, web-based config repair, supports Astra / Sol / Terra, subscriptions, and proxies; no guarantee of obtaining a specified state or improving model quality.
  7. incoai/splash★293 A local inference engine for Apple silicon, built around the model.
  8. featherless-ai/simple-jev★282 Turn any open model into a classifier/jev endpoint
  9. Sac-Y/Jev-cu★276
  10. kyotofin/tax-doc-classifier★261 Tax document page classifier built on Jev decisions. 100% strict accuracy across 261 IRS forms, ~$0.001 per page.

Star gain 1d

  1. cloudflare/security-audit-skill+3935★15978
  2. browser-use/jev-ultrafast+2881★8345
  3. affaan-m/ECC+1134★261321
  4. NandhaKishorM/laya+1024★1249
  5. tamaratran/fast-jev-compaction+1023★4213
  6. eternity4719/HowToLiveBetter+901★5932
  7. TianyuCodings/NanoJev+496★866
  8. TencentCloud/Octop+452★4207
  9. arvindear/wp2shell-PoC+399★562
  10. jaredpalmer/kev+382★511