tierlist.tech
中EN

10.02 Today's Picks

382 new projects found today, 189 rated.

Created 2026-09-30 – 2026-10-02Fetched 600Filtered 50Rated 34

A · Worth watching

A

gpuslider

niklaspTypeScript

A shader-drawn slider using WebGPU with a WebGL2 fallback, 4.7KB core with everything else as plugins. The scope is cleanly cut and the promises match the code size; worth a look if you want non-standard controls, but don't expect it to work out of the box.

A slider drawn by shaders: WebGPU, or WebGL 2 where there is none. A 4.7 KB core, everything else a plugin.

#GPU#Web Frontend#UI Design

★28

A

skkmonkey

taniTypeScript

Turns the SKK Japanese input method into a pure front-end userscript that runs directly in Firefox and Chrome without a server, solving the awkwardness of installing Japanese input on Chromebooks and Linux browsers. The scope is small and the privacy notes restrained. The limitation is that it only serves longtime SKK users, far narrower than mozc.

A local SKK Japanese IME for Firefox and Chrome, written in TypeScript and built with Node.js. Runs as a userscript; no server is required and typed text is never sent over the network.

#Browser Extensions#Privacy#Productivity

★9

A

aime

zoolappSwift

A RIME-based Chinese input method for macOS, local-first with AI as an optional enhancement, with clear positioning and install path. It is a new option for Chinese users tired of Squirrel's fiddly configuration who want to try AI suggestions; but competing head-on with Squirrel and WeChat Input, its maturity is still unknown.

Aime Input Method · An open-source Chinese input method for macOS, built on RIME. Local first, optional AI.

#macOS#AI#Privacy

★21

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

BHas a spark43
  • B playport Wirelessly casts CarPlay to any screen on the LAN as a head unit, reusing old tablets or browser screens and saving the hardware cost of replacing the head unit; 120 stars in 0.6 days is strong momentum, spreading fast in modding and garage geek circles. Monetization can come from hardware accessories plus software licensing. The risk is the legal boundary of protocol reverse engineering and long-term stability maintenance. #IoT#Self-Hosted#carplay

    Wireless CarPlay in your browser — a server-side receiver that turns any screen on your network into a head unit.

    youcciKotlin★120

  • B listing-to-3d A Codex skill that reconstructs property listings into walkable flattened 3D scenes, hitting real-estate marketing's need for low-cost virtual tours; 47 stars in 1.2 days shows agent-skill distribution momentum, and it could be packaged as B2B SaaS priced per unit. The risk is head-on competition with mature solutions like Matterport and general 3D generation models. #AI Agents#Agent Skills#3D#Image Generation

    Codex skill for reconstructing real-estate listings as plan-based, walkable 3D scenes.

    aek15Python★47

  • B find-my-nuts Uses the Apple Find My network as a transport layer so Cashu ecash on an ESP32 can send and receive tokens offline, a clever angle that honestly labels itself a proof of concept. It is inspiring for people studying off-grid payments and side channels; the limitation is that neither the wallet nor the receiving end is close to daily use. #Crypto#Embedded#Hardware#Privacy#cashu

    Proof-of-concept ESP32 Cashu wallet with Find My transport and a host receiver

    zeugmasterC★21

  • B Comma A sessionless, always-running personal agent positioned against AI companion products like Muse, Dots, Instinct and Town; 141 stars and 1.6 days of momentum are decent, and open-source plus self-hosting differentiates it for privacy-sensitive users, with cloud sync and premium subscriptions as monetization. The risk is that companion agents generally retain poorly and it collides head-on with big-vendor personal assistants. #AI Agents#Open-Source Alternatives#Productivity

    The sessionless, relentless personal agent. Open source alternative to Muse, Dots, Instinct and Town.

    AFK-surfElixir★141

  • B JEV_sees Provides real-time visual decision-making with RGB, video, and RGB-D for the JEV robot platform, a perception-layer component for embodied intelligence; 92 stars in 1.9 days show the robot-vision community is following. It could monetize via robot-vendor licensing or a hardware SDK, but the risk is JEV's niche ecosystem and competition with mature perception stacks like NVIDIA Isaac. #ML Research#AI#Images

    Eyes are All JEV Needs - real time visual devisions from RGB, video and RGB-D cameras.

    CharlesFeng0314Python★92

  • B threejs-procedural-animals Procedurally generates animals in three.js using SDFs, auto-rigging skinning, adding fur or scales, and IK gaits without any models, textures, or animation clips. Technically substantial and worth a look for web 3D and game prototyping, though still far from production quality. #3D#Animation#Game Dev#three.js

    Fully procedural animals for three.js: sculpted from SDFs, auto-rigged and skinned, with fur, feathers and scales, and IK gaits and actions. No models, no textures, no animation clips.

    majidmanzarpourJavaScript★14

  • B Hyperliquid-Leaderboard-Analytics Tracks PnL, ROI, and rank changes on Hyperliquid leaderboards in the terminal, distinguishing consistently profitable wallets from one-off lucky ones, using public read-only APIs; 60 stars in 1.9 days shows real trader demand. It could become a paid alpha-subscription tool, but the risk is unofficial APIs changing at any time and signal decay as strategies homogenize. #Crypto#CLI#Data Analysis

    Hyperliquid analytics right in your terminal. Track top traders by PnL, ROI, and account value, monitor ranking changes over time, and identify consistently profitable wallets while separating them fr…

    BarbarakarlssonPython★60

  • B skillscout Scans which skills a coding agent actually loads, which agents can use them, and how often, filling the blind spot of opaque agent configuration; 43 stars in 1.4 days, and the author has their own distribution channel. It could become the diagnostic standard for the skill and MCP ecosystem, but the risk is a still-small ecosystem and platforms potentially building this in natively. #AI Agents#Agent Skills#Observability#macOS

    Find the skills your coding agents load, see which agents can use each one, and count how often you use them

    flaviocopesSwift★43

  • B herringbone A pure-Ruby Parquet reader and writer that lets Ruby services read columnar files without installing Arrow native libraries. Narrowly scoped with an honestly self-deprecating description; the limits are a small ecosystem and unverified performance, but Ruby engineers writing ETL should take a look. #Databases#Data Analysis#Performance

    A (slightly unhinged) pure-Ruby Parquet file reader/writer

    julikRuby★14

  • B xentop-ng A Xen resource monitoring TUI modeled on btop, from the author of Xen Orchestra, self-described as a demonstrator. Xen admins have long lacked a good monitoring interface, so it is worth a look; the feature set is still narrow. #Monitoring#Observability#Linux#btop

    A modern, btop-inspired resource monitor for the Xen hypervisor (demonstrator)

    olivierlambertRust★13

  • B glm53-flash-offload A concrete plan for fitting a large MoE model on a single 24GB 3090: expert caching, zero-copy offload, CPU layers and an OpenAI-compatible interface are all spelled out. It's directly useful to people wanting to run large models locally, though it only targets GLM-5.3-Flash + EXL3 and needs edits for other models. #LLMs#GPU#Performance#AI

    GLM-5.3-Flash EXL3 on one 24 GB RTX 3090 + DDR4: elastic GPU expert cache, zero-copy experts, AVX2 CPU tier, OpenAI API

    0xSeroPython★20

  • B dvcp-aac Adds AAC audio codec and import support to DaVinci Resolve, by the maintainer of Proton-GE, addressing a real problem with restrained scope. A direct pain point for people editing in Resolve on Linux; the limitation is that it only solves one codec format. #Audio#Video Editing#Linux

    Davinci Resolve AAC audio encoder/decoder with import support.

    GloriousEggrollC++★22

  • B float-android Builds an "AI virtual phone" inside Android: a realistic home screen, social feed, location and calls, with characters that keep schedules and long-term memory, all data stored locally. The concept is complete and rare, worth ten minutes for anyone interested in AI companions; but it only connects to your own API and the actual experience is unverified. #AI Agents#Android#LLMs#Image Generation

    An AI virtual phone inside Android: realistic desktop, chat, Moments, phone checking, map location, voice/video calls, diary and calendar. AI characters have schedules, routines, locations, and long-t…

    shiaho777TypeScript★14

  • B mikrotik-home-netflow-plus A self-hosted NetFlow/IPFIX traffic panel for MikroTik RouterOS 7, a single Go binary with a live web UI, runnable today by anyone with RouterOS at home. The scope is clear and the claims are modest. ntopng and Elastiflow exist, but it wins by focusing only on MikroTik and deploying lighter; depth is not yet visible. #Monitoring#Observability#Self-Hosted#DevOps

    Self-hosted NetFlow/IPFIX traffic monitor for MikroTik RouterOS 7 with a live web dashboard.

    thedyermanGo★8

  • B fpvdash A full-screen telemetry panel for EdgeTX color-screen radios, usable by Betaflight pilots on ELRS/Crossfire by copying it to the SD card, with a find-my-model QR code and voice announcements. The need is real and the boundaries restrained. The limitation is that it covers only this small group, with limited gain over the built-in telemetry page. #Hardware#Embedded#Monitoring

    Full-screen telemetry dashboard for EdgeTX colour radios: Betaflight quads on ExpressLRS or Crossfire, with a find-my-quad QR code, logbook and voice callouts

    vcazanLua★8

  • B witness-vault Generates self-decrypting HTML files with WebCrypto and AES-256-GCM that open offline, suited to passing small amounts of sensitive text. The approach is not new but it is cleanly done with a credible description; the key derivation and tamper-resistance details are worth a look. #Secrets#Privacy#Documents

    Zero-install, air-gapped, self-decrypting HTML vaults using AES-256-GCM and native WebCrypto.

    HamidoohPython★25

  • B ai-agent-gateway A gateway for AI agents' MCP tool calls and LLM traffic, unifying auth, policy and observability. It fills a real gap in enterprise agent deployment, implemented in Go with a clear focus; the risk is that similar gateways are appearing rapidly, so the policy engine's granularity will decide. #MCP#AI Agents#App Security#Monitoring

    AI Agent Gateway - an open-source gateway that secures, governs and observes AI agents' MCP tool calls and LLM traffic

    TuskiraGo★28

  • B satisfactory-logistics-map A self-hosted map for Satisfactory dedicated servers: live positions, production lines, power, logistics history and capacity planning, up in five minutes with Docker. A real need for small teams playing the game, but mature map tools already exist, so actual completeness matters. #Games#Self-Hosted#Data Analysis

    Web map for Satisfactory dedicated servers: live map, production, power, logistics, history and a production planner — self-hosted with Docker in 5 minutes

    Fade97Python★16

  • B factorylog An agent activity timeline for macOS that shows by day and week what Claude Code and similar tools did across your projects and where time went. From Flavio Copes, so build quality is likely solid; watch whether it collects via git or process hooks, and the privacy boundaries of session data. #macOS#AI Agents#Productivity#Observability

    See what your coding agents did for you: a timeline of each day and week on your Mac, and where your time went, project by project

    flaviocopesSwift★24

  • B perspica Claims to review code by what a change does rather than line by line. The direction targets review fatigue, but the one-line description does not say whether it uses AST or an LLM or how it plugs into PR workflows, and 27 stars reveal nothing about actual results. #Code Quality#CLI#Debugging

    Review code changes by what they do, not line by line.

    sshah03Rust★27

  • B WiimKnob Turns the Waveshare ESP32-S3 knob screen into physical remote firmware for WiiM streaming amplifiers, with the board model and feature boundaries spelled out clearly, flashable in ten minutes by anyone with the board. Its value depends entirely on whether you use WiiM; generality is near zero. #Embedded#Hardware#IoT#Audio

    WiiM remote control firmware for the Waveshare ESP32-S3-Knob-Touch-LCD-1.8

    xon94010C★8

  • B tex-revision-cursor-extension A small Cursor / VS Code extension for accepting, rejecting and stepping through TeX revision marks, filling the gap where AI edits to LaTeX cannot be reviewed item by item. Scope is restrained and usage clear; the downside is it only serves LaTeX users and depends on a specific revision package. #Editors#Plugins#Documents#latex

    Accept, reject, and navigate TeX revision markup in Cursor and VS Code.

    Amazing-BirdsJavaScript★10

  • B Roblox-BlenderBridge Preview and push meshes, textures and skeletons from Blender straight to Roblox. It's a clear convenience for Roblox creators and the audience is large, but similar plugins already exist, so success depends on stability and keeping up with Roblox API changes. #3D#Game Dev#Plugins#blender

    Preview or upload mesh, texture, and armatures to roblox from blender easily

    TylerAtStarboardLanguage unknown★18

  • B swe-sweep An evaluation repo from Meta's research team asking how many bugs an LM can find and fix in large codebases. Valuable for anyone evaluating code agents, but the key question is whether the task set and eval scripts are released; if it's only a paper, it's just a page of results with figures. #Benchmarks#LLMs#Code Quality

    How many bugs can LMs find & fix in large codebases?

    facebookresearchPython★15

  • B fleet Runs Claude Code and the pi agent in parallel inside Docker sandboxes on a Mac, automatically producing tested PRs. The idea is clear, but similar orchestrators like Conductor and claude-squad already exist, it has only 8 stars after one day, and usability is still unclear; watch for now. #AI Agents#LLMs#macOS#DevOps

    Runs Claude Code and pi agents in parallel on your Mac, one Docker sandbox per issue. A host agent hands out the work, watches every sandbox and brings back tested, reviewed pull requests.

    jaqubowskyTypeScript★8

  • B open-finance-agent A financial research framework that bakes "check trusted sources, compute numbers for real, cite conclusions" into the agent loop. Its design is more honest than most wrapper agents, but with 7 stars and no sample output, reliability in financial settings is unproven. #AI Agents#LLMs#Fintech#RAG

    An open-source financial research harness that enforces analyst standards in the agent loop itself: trusted sources first, no mental math with every figure computed in a Python sandbox, and every numb…

    nemofqTypeScript★7

  • B gleaner A local-first browser extension that grabs original images, videos with sound and audio from any web page, with zero telemetry. The scope is clear and the privacy stance honest, but similar download extensions are mature, so it lacks a reason to be indispensable. #Browser Extensions#Downloaders#Privacy#Video

    Gleaner — collect full-resolution images, video with sound, and audio from any web page. Zero telemetry, local-first browser extension.

    ssssssanjiuTypeScript★12

  • B linuxqq-wayland-fix Fixes screen sharing, screenshots and clipboard for Linux QQ under Wayland, a real problem with a single narrow scope. A must-have for Chinese Linux desktop users, but it is a patch-style solution that may break as soon as QQ updates. #Linux#Video#wayland

    Fixes screen sharing, screenshots, and clipboard issues in Linux QQ under Wayland

    SHORiN-KiWATAC★14

  • B ntpods A small Windows tool for managing AirPods, rebuilding in Rust a need currently held by paid software. Useful for people forced to use AirPods with a PC; right now there's only a one-line description, and success depends on battery display, noise-mode switching and auto device switching, competing with MagicPods. #Windows#Audio#Hardware

    Open-source AirPods control for windows

    arctumnRust★11

  • B optimaizr A local CLI that scans Claude Code, Codex and self-hosted API calls for wasted tokens, pointing out which parts of the context are burning money. Very concrete and directly useful for people running agents daily; the limitation is that conclusions rest on its own heuristics, so compare with each vendor's usage panel first. #CLI#LLMs#Productivity

    Spend fewer tokens on the same work. A local CLI that finds wasted tokens in Claude Code, Codex and your own API calls.

    blendbunjakuTypeScript★8

  • B outis A Go anti-spam idea: return fake user-unknown bounces to suspicious senders so bulk address-harvesting bots waste their time. Small, novel and narrowly scoped, suited to people running their own mail servers; the limitation is it only works against crude harvesting and you must wire it into your MTA yourself. #Email#Network Security#Self-Hosted

    Fight spam by generating and sending a fake "user unknown" bounce emails

    dtononGo★11

  • B gsc-full-render A bookmarklet that shows the full Google-rendered HTML and links directly in Search Console's URL inspection. A tiny niche that genuinely solves an SEO pain point; the audience is narrow, though, so just bookmark it. #Browser Extensions#Debugging#Automation

    Bookmarklet for Search Console's URL Inspection: the whole page from Google's rendered HTML, with its links

    natzirJavaScript★11

  • B homelab-agent Turns a 70-euro used mini PC into an always-on AI agent server, with Hermes, n8n, Pi-hole, Tailscale and encrypted backups all codified. Useful for self-hosters, but it is essentially another homelab orchestration list with nothing new. #Self-Hosted#DevOps#Automation#Productivity

    Turn a €70 used mini PC into an always-on AI agent server: Hermes, n8n, Pi-hole, monitoring, Tailscale and encrypted backups, all as code.

    sjespersenShell★13

  • B shipstores Wraps App Store Connect, Google Play and EAS as an MCP service so coding agents can ship releases directly. The entry point is specific and saves manual upload back-and-forth, but releasing is high-risk, and permissions and rollback strategy are unclear, so it should not touch production yet. #MCP#AI Agents#iOS#Android

    Ship iOS and Android apps from your AI agent. MCP server for App Store Connect, Google Play Console and Expo EAS.

    FidelisMMPython★8

  • B codex-approval-notifier Pushes Codex approval events to Windows popups and Bark phone notifications, solving the need to watch the screen during long tasks. Small and precise, it works as soon as installed; the limitation is that it only covers Codex and the notification channels are personal. #Windows#AI Agents#Productivity#Automation

    Windows desktop popups and optional Bark mobile notifications for Codex conversations.

    Lyu-ZHPowerShell★16

  • B promptfilm Generates a research-flavored 3D motion short from one sentence, producing a loopable Three.js HTML file and a frame-by-frame MP4. Grounding 'code as visuals' in concrete artifacts is more substantial than prompt-only video, but results depend heavily on the model and the export chain needs independent verification. #Agent Skills#AI Agents#Video#Animation

    One sentence in, a researched 3D motion graphic out — a Claude Code skill that makes looping real-time HTML films (Three.js) and frame-exact MP4s

    SeokwooooJavaScript★17

  • B jellyfin-portal A dark-themed landing page for self-hosted Jellyfin users, with recently added items, status widgets, and a proxy that keeps API keys off the client. Restrained in scope, but portals like Organizr are already mature, so unless you specifically want this skin there's no need to seek it out. #Self-Hosted#Streaming#Web Frontend#APIs

    A dark Jellyfin-themed homepage portal with Recently Added row, status widgets, and a serverless API proxy to keep keys off the client.

    jnracreatesCSS★14

  • B openink Describes interfaces in YAML and generates clickable hand-drawn-style wireframe prototypes. A neat idea, but text-to-UI tools are already plentiful and the hand-drawn style helps little in real reviews, unless you happen to write specs in YAML. #UI Design#Web Frontend#Productivity

    Describe screens in YAML. Get a clickable, hand-drawn wireframe prototype.

    mediuswareltdJavaScript★10

  • B loci Typed long-term memory for agents, with a core library plus an MCP server and Hermes plugin, reasonably structured. But agent memory is crowded with mem0, Zep and the like, and the repo doesn't explain which retrieval failure 'typed' solves, making it hard to justify switching stacks. #AI Agents#LLMs#MCP

    Typed long-term memory for AI agents. One core, an MCP server, and a Hermes plugin.

    cmoraes10Python★24

  • B procedural-pixel-creatures A procedural pixel creature generator for Godot that batch-produces enemy assets from given rules, saving indie developers exactly the drawing time. A restrained, clean little tool; the limitation is that results depend entirely on parameters, and the repo has no sample image library yet. #Game Dev#Images#godot

    Godot Tool to create procedural pixel creatures

    idlerunner00C#★12

  • B whiteboard-video A pipeline turning a topic or SRT into hand-drawn whiteboard videos: AI scripting, Vietnamese voiceover, storyboarding, handwriting animation and MP4 export. Practical for Vietnamese content creators, but whiteboard animation tools are already plentiful and the voice only covers one language. #Video#Animation#Speech#AI

    Biến chủ đề, kịch bản hoặc file SRT thành video vẽ tay bảng trắng hoàn chỉnh: AI agent viết kịch bản, tạo giọng đọc tiếng Việt (Vbee/Edge TTS), phân cảnh, vẽ nét bút liền mạch, gắn phụ đề và xuất MP4.

    Cuongyd196HTML★19

  • B tensward Profiles your own vLLM deployment by running real workloads and giving tuning suggestions. It targets the pain point of local inference tuning, but the project is too new, with sampling methods and advice unverified, so vLLM's built-in benchmark is usable for now. #LLMs#Performance#GPU#Benchmarks

    Profile and diagnose LLM serving on your own GPU. Measures your vLLM setup on your real workload and suggests what to change.

    TenswardPython★9

  • B Model-TG An unofficial firmware extension for the Elektron Model:Cycles that adds a sampler, resampling and beat-repeat. Assembly-level hardware hacking done with clear judgment makes it a skilled project, but its audience is limited to musicians who own this machine. #Audio#Hardware#Embedded

    Unofficial firmware additions for the Elektron Model:Cycles (OS 1.13): a full Sampler machine, resampling, beat-repeat with master FX, and more

    TinyGregAudioAssembly★8

COrdinary61
  • C openpages An AI workspace for both humans and agents, with SuperCompress as a context-compression layer; 45 stars in 1.9 days is middling, but the AI workspace track is extremely crowded and compression lacks independent benchmarks, making general gains for mainstream agents hard to prove, with no clear paid scenario. #AI Agents#Knowledge Base#Productivity

    Open-source AI workspace for humans and agents — SuperCompress as the core context layer.

    arjunkshah12345-hashTypeScript★45

  • C claude-code-pro-course A full Russian-language Claude Code tutorial set with 8 labs, a prompt library, and a starter kit; 140 stars prove real demand in non-English markets. But it is educational content rather than infrastructure, monetizable only via paid courses or lead generation. The risk is a limited language market and official docs plus free content continuously diluting willingness to pay. #Learning#MCP#Prompting

    Claude Code PRO — a complete course and guide in Russian: CLAUDE.md, settings, skills, subagents, hooks, MCP, plugins, CI. 8 labs, a prompt library, and a ready-made starter-kit. Free ChatGPT courses …

    justxorShell★140

  • C AuthFlowMap Turns OAuth, Entra conditional access, Agent ID, and MCP authentication flows into an interactive map, solving the pain of explaining identity configuration; 39 stars in 1.6 days is steady. It can drive security training and consulting leads, but it is essentially documentation-style content, hard to turn into a paid product or be relied on as a dependency. #App Security#MCP#Learning

    Interactive map of OAuth, Conditional Access, Entra Agent ID and MCP authentication flows, step by step

    JinssiHTML★39

  • C dsh-libreoffice-kit A document-processing component inside DeepSeek Harness, with 80 stars largely from the DeepSeek brand; the code is optimized around its own agent runtime, with insufficient generality, docs, and interface stability, raising the bar for external adoption. It lacks an independent commercialization path and looks more like an internal dependency spilling out than a general standard part. #Documents#Plugins#AI

    An internal component used by DeepSeek Harness

    deepseek-aiJavaScript★80

  • C mcp Wraps Gupshup's WhatsApp, SMS and RCS marketing capabilities as an MCP, including template approval, A/B testing and multilingual broadcasts, with human confirmation before sending. It saves integration work for teams already on Gupshup, but it is essentially a vendor SDK wrapper tied to one platform. #MCP#AI Agents#APIs#Marketing

    Run WhatsApp, SMS and RCS from Cursor: Gupshup analytics, template drafting and approval, and A/B, multilingual and scheduled campaigns, with a human review step before anything is sent.

    GupshupSuperAgentPython★12

  • C sevoflurane Tucks Windows Steam into a Wine bottle and rewrites the interface in native Swift, aiming for a true Mac Steam. The idea is appealing, but Wine's Steam pitfalls and Valve's stance are hard obstacles; 12 stars in a day makes it look like a personal experiment. #macOS#Games#UI Design#steam

    Steam for macOS, as a real Mac app: the Windows Steam client runs hidden in a Wine bottle; everything you see and touch is native

    kageroumadoSwift★12

  • C cua-swe A visual software engineering benchmark for computer-use agents, with 105 tasks spanning web, games, mobile, and ops. The direction has value, but with 7 stars nobody has run it yet, and task quality and judging criteria remain unverified; no rush. #AI Agents#Benchmarks#Testing#ML Research

    CUA-SWE: When Computer-Use Agents Meet Visual Software Engineering. 105 tasks across Web, Game, Mobile, and DevOps.

    kingofspace0wzzTypeScript★7

  • C ComfyUI-Olm-SoL-Refiner An experimental ComfyUI node wrapping NVIDIA's SoL-Refiner. It is a timely small patch for people running ComfyUI, but the audience is very narrow and experimental status means the interface can change at any time. #Image Generation#Images#ML Research#comfyui

    Experimental ComfyUI nodes for NVIDIA's SoL-Refiner

    o-l-l-iPython★10

  • C Rallo A macOS desktop pet that puts notes and reminders on screen and waves when a coding agent needs you, fully local and CLI-first. The positioning is clear and fun, but desktop widgets are common, so its value depends on whether you want it always on screen. #macOS#Productivity#CLI#AI Agents

    A macOS desktop pet that keeps your notes and reminders in view and waves when your coding agent needs you. Local only, CLI first.

    EyakubRust★13

  • C LiquidGlassEffects Exposes 24 material variants of macOS's private NSGlassEffectView, including sub-variants the public API does not provide. Attractive to macOS developers fussing over visual details, but it relies on private APIs, and App Store and OS-upgrade risks are not mentioned. #macOS#UI Design#Reverse Engineering

    Access to all 24 Liquid Glass material variants and 10 subvariants of macOS's NSGlassEffectView, including the private _variant / _subvariant ones the public glassEffect API doesn't expose.

    HoYeonPark1221Swift★8

  • C GLM-5.3-Flash-EXL3-2x-DGX-Sparks-TensorFold Quantizes GLM to EXL3 and runs it on two DGX Sparks with TensorFold, solving local deployment on very large-VRAM machines; 128 stars show high geek interest. But it is essentially a personal config script that breaks as models and drivers iterate, lacking a durable product form or commercialization path. #LLMs#GPU#Self-Hosted

    GLM-5.3-Flash EXL3 on 2x DGX Spark with TensorFold

    MiaAI-LabShell★128

  • C awesome-homelab Templates homelab configurations into a GitOps repo with real configs and guides, significantly lowering the cold-start cost of self-hosting; 68 stars in 0.8 days shows list-style momentum. But the value is in curation and ongoing maintenance rather than a product, monetization is limited to sponsorships or paid docs, and it is not a dependable component. #Self-Hosted#DevOps#Learning

    Your homelab, as a Git repository: free tools, real configs, and guides for self-hosting with GitOps.

    Minh141120Language unknown★68

  • C orion An ESP32-S3 desktop voice assistant with double-click flashing and a bundled API key, combining a local device with cloud models. The hardware and setup flow are clearly explained, but open-source voice boxes are already numerous, and it lacks one sentence on how it differs. #Embedded#Speech#Hardware#IoT

    A little voice assistant for your desk that you build and own. Flash an ESP32-S3 board with one double click and bring your own keys.

    MultiX0C★10

  • C computer-use-fast A macOS agent skill that avoids the GUI when a single shell command will do, otherwise completing clicks, typing and screenshots in one cu.py call. 'CLI first, GUI second' is cheaper than pure screenshot-and-click, but it is single-platform, single-ecosystem, with unknown stability. #Agent Skills#AI Agents#macOS#Automation

    Agent skill for macOS GUI tasks in one or two turns: a shell command when one answers, else a single cu.py call that opens the app, clicks by visible text, types, reads the screen and screenshots.

    Nanako0129Python★8

  • C ProsperoEden An experimental port of the Eden console emulator to PS5, with 196 stars and 1.9 days of momentum—the strongest in this batch of GitHub projects—reflecting an active console homebrew community; it has UEFI and low-level engineering value, but almost no commercialization and legal pressure from rights holders like Nintendo, making it a community watch item only. #Game Porting#Reverse Engineering#Games

    Experimental Eden port for PS5

    blackbearreloadedC++★196

  • C ESP32SDR Builds a full-duplex SDR from two ESP32 boards, explaining the idea in one sentence; a low-cost approach for hardware hobbyists. But with no documentation or measured data, whether it actually works depends on RF skills; an experiment to watch from a distance. #Embedded#Hardware#sdr#esp32

    Full duplex sdr from two esp32

    lozaningC★14

  • C hearr Adds a music discovery step for Lidarr users, and one sentence makes the positioning clear. Unfortunately similar recommendation scripts are common and the repo gives too little information to judge recommendation quality or how to integrate it. #Self-Hosted#Audio#Automation

    Music discovery for Lidarr

    nurulmac11Python★7

  • C Metalio-E-INK4-Plus Hardware specs for a self-built 4.7-inch e-ink reader and voice assistant device, with component models listed in great detail. Useful reference for anyone building open-source e-book hardware, but so far only part selection is shown, no firmware or enclosure files, so it reads more like a personal project manifesto. #Hardware#Embedded#IoT

    Metalio E-Ink4-Plus is an open-source e-ink hardware device for reading and AI assistant scenarios. The device features a 4.7-inch 1216×684 parallel e-paper display (ED047TC2 / EPDiy + TPS65185), FT63…

    CloudZaoC★7

  • C motion-design-skill A Claude Code skill that generates 30-second explainer videos in 22 motion styles using pure code, without calling video models. The direction is specific, but the completeness of the 22 styles and export quality cannot be confirmed, and it heavily overlaps with similar code-generated motion skills. #Agent Skills#AI Agents#Video#Animation

    Claude-Code-Skill: 30-Sekunden-Erklärvideos in 22 Motion-Design-Stilen – jedes Bild ist Code, ohne KI-Videomodelle

    IsmypHTML★10

  • C awesome-ai-motion Curates 637 motion recipes for AI video, infographics, and scrolling presentations with an agent skill, an index rather than a tool; 44 stars in 1.6 days shows list-style momentum, and it could sell courses or templates. But the recipes are easily replaced by native video-model capabilities and frontend animation libraries, leaving a thin moat. #AI#Video#Animation

    637 motion techniques for AI video, infographics and scroll decks — rendered clips, recipes, routes and an agent skill (ko/en)

    gongnyangHTML★44

  • C ComfyUI-H3-IMG-Gen-SatoDive A ComfyUI node for MiniMax H3 focused on precise output resolution, multiple reference images and single-pass generation. It saves a few nodes for existing ComfyUI users, but its feature surface is narrow and official or community nodes will likely cover it eventually. #Image Generation#Plugins#AI

    Simple ComfyUI nodes for MiniMax H3: exact output resolution, multi-reference images, single-pass generation, optional resolution-preserving detail refine.

    SatoDivePython★8

  • C house-planner A self-hosted floor plan designer with 2D real dimensions, 3D, satellite basemaps, plumbing/electrical and a priced bill of materials. The ambition is too big for a project 0.6 days old, and React+PHP without a database struggles with complex drawings; for similar work, see Sweet Home 3D and Planner 5D. #Self-Hosted#3D#Web Frontend

    Self-hosted house planner: true-scale 2D floor plan, 3D, satellite map overlay, sewer, electrical, water and underfloor heating with calculations, priced bill of materials. React + PHP, no database. B…

    egmaltTypeScript★15

  • C project-manager-skill A self-contained project-management skill for ChatGPT covering milestones, implementation coordination, and acceptance records; 50 stars show real demand in the agent-skill directory. But the core is just prompts and template organization, easily copied, lacking data and network effects, with monetization limited to skill-marketplace revenue share and a low ceiling. #Agent Skills#Productivity#Documents

    A self-contained project management skill for ChatGPT: milestones, implementation coordination, exact-build QA, and project records.

    finnaLanguage unknown★50

  • C exynos9810-woa Upgrades a Galaxy S9+ from native Android to Windows 11 via an installer, including a full UEFI and driver port; 43 stars reflect an active WoA porting community. The engineering is substantial but the audience is tiny, with no commercial path and a heavy maintenance burden—a watch-level hack rather than a dependable result. #Windows#Embedded#Hardware#Reverse Engineering#samsung-galaxy-s9

    Windows 11 on the Samsung Galaxy S9+ (SM-G965F, Exynos 9810): UEFI, drivers and an installer that goes from stock Android to the Windows desktop.

    ntdevlabsC★43

  • C openmimir Talk to Claude Code, Codex and OpenCode simultaneously on a Mac using full-duplex voice. The concept is novel, but the real benefit of voice-directed coding agents is doubtful, and 7 stars offer no evidence of stability, so it is a novelty. #AI Agents#Speech#macOS#LLMs

    Talk to all your coding agents. One hands-free, full-duplex voice conversation over Claude Code, Codex and OpenCode. Runs on your Mac.

    openmimirTypeScript★7

  • C ad-audit-wordlist A merged wordlist for AD password auditing and Kerberoasting, with sources, dedup tooling and hashcat rules documented so the recipe is reproducible. Handy for internal network assessments, but it is essentially a repackaging of others' wordlists with no new method and little independent value. #App Security#Network Security#Secrets

    The Kerberoasting / AD password-audit wordlist I use on engagements: priority-merged HashMob large list + The-Viper-One's kerberoast_pws, order-preserving dedup via rling, paired with OneRuleToRuleThe…

    ADScanProShell★7

  • C Ultima7_Decomp A byte-level verified decompilation and reconstruction of Ultima 7, rigorously engineered, with long-term value for reverse engineering and game preservation; 48 stars is normal heat for the retro community. But the result is one-off research with no commercialization path and no reusable standard dependency. #Reverse Engineering#Games#Game Porting

    This is a decompiled and byte-verified rebuild of Ultima 7.

    KitrinxC★48

  • C aevori-app A local AI chat and character agent for macOS built on Ollama, with controllable memory and a browser workspace. Completeness looks decent, but the space is crowded, and the description does not clarify the key differences from existing local clients. #AI#LLMs#macOS

    Local AI chat and personal agents for macOS. Ollama-powered conversations, customizable characters, controllable memory, and a browser-based workspace. Ready-to-run downloads for Apple silicon and Int…

    StxqqTypeScript★16

  • C inlay A virtual game keyboard for Android handhelds, mapping physical controllers to input, serving Retroid, Ayn and similar handheld players. The need is real, but similar key-mapping apps are plentiful, and the repo has only one line of description with no indication of supported models. #Android#Games#Hardware

    A gamepad keyboard for Android handhelds

    dakingemanKotlin★7

  • C liuliang-buddy A small Android tool for checking data balances on China Mobile, Unicom, Telecom, and Broadnet, with a home-screen widget; the author honestly notes Telecom's remaining data is estimated from the official site. A real need but with many alternatives, differentiated only by the widget and small size, so little technical reference value. #Android#Productivity#UI Design

    Data Buddy: a cute Android data-usage query app with desktop widgets, first-time choice of China Mobile / Unicom / Telecom / Broadnet (beta, Telecom remaining data estimated from official website valu…

    huachen19867Dart★15

  • C factory-switcher A macOS menu bar tool for switching between multiple Factory(Droid) accounts and checking quotas, with backup and rollback. The features are concrete, but it essentially circumvents single-account quotas and the description is stuffed with emoji, showing questionable judgment; those who need multiple accounts will find it themselves. #macOS#Productivity#Automation

    🔁 Native macOS menu-bar app to switch between your own Factory (Droid) accounts — quota at a glance, backup-first switching with auto-rollback.

    lhferSwift★13

  • C ResearchPilot An academic research agent with citation verification and prior-work analysis, emphasizing that conclusions are grounded. The problem is that such research agents have clustered in the past six months, and 10 stars with no reproducible evaluation makes it hard to tell what it does better. #AI Agents#LLMs#Science#RAG

    Evidence-grounded academic research agent with claim verification, prior-work analysis, and persistent research workspaces.

    renfei421Python★10

  • C minecraft-crossover-bridge Runs real Minecraft inside Monster Hunter: World and Elden Ring via CrossOver on Apple Silicon. More a one-off technical joke with near-zero usability; ordinary players needn't bother. #Games#Game Porting#macOS#minecraft

    Play real Minecraft inside Monster Hunter: World and Elden Ring on Apple Silicon Macs with CrossOver

    justbustinJava★10

  • C clef-webcam Runs Cloudflare's clef-flash classification model locally on a webcam in real time, a typical one-day demo. Beyond showing the model's responsiveness it has little use; those curious about new models can take a quick look. #AI#Images#Web Frontend

    Run Cloudflare's clef-flash decision model locally on your webcam

    lucatacoHTML★10

  • C digisophie A porting experiment bringing the Sophie synthesizer to the Elektron Digitakt. It appeals to hardware music fans, but the audience is tiny and there is almost no documentation, so outsiders cannot judge how complete it is. #Audio#Hardware#Embedded#digitakt

    Digitakt 1 machine based on Sophie-for-schwung

    soejrdAssembly★13

  • C dissect A Claude Code skill that breaks YouTube and TikTok short videos into shot lists and structural templates for reuse. The idea is clever, but it's ultimately prompt wrapping with no explanation of the breakdown basis or output format, so results depend entirely on the model, and such "viral breakdown" tools are already common. #Prompting#AI Agents#Agent Skills#Video

    Reverse-engineer any video (YouTube, TikTok, Reels, X...) into a reusable blueprint for making similar videos. A Claude Code skill.

    0xpratzyyPython★12

  • C omarchy-agent-notch A Dynamic Island for the Omarchy desktop that shows the status of agents like Claude Code and lets you ask them questions directly. The idea is appealing, but it is tied to one desktop distro, and the actual animation and interaction cannot be judged from the description. #Linux#Plugins#AI Agents#Terminal

    Dynamic-island notch for Omarchy showing live Claude Code (and Grok Bot) agents — animated faces, alerts, ask Claude from the notch

    GNSBQML★8

  • C Bench2Drive-RLinfra Infrastructure for reinforcement-learning post-training of end-to-end autonomous driving policies on CARLA and Bench2Drive. Useful reference for autonomous-driving RL research groups, but it depends on heavy simulation environments, is basically unrunnable for individuals, and no experimental conclusions are shown. #ML Research#Robotics#Benchmarks

    Scalable reinforcement learning infrastructure for autonomous driving, supporting privileged planning and end-to-end policy RL post-training in CARLA.

    Random8756Python★9

  • C cinematic-film A Claude Code skill that turns any topic into a short Remotion + three.js animation with AI images, voiceover, and sound effects. Strong as a demo, weak on reusability; a toy in the current agent-skill boom. #AI#Video#Animation#Agent Skills

    Claude Code skill: turn any subject into a cinematic animated film with Remotion + three.js, gpt-image assets, a rigged TTS narrator and synthesized sound.

    minosdevsTypeScript★18

  • C Corpus A self-hosted AI document workspace emphasizing sourced answers and connections to business data. The direction is right, but enterprise knowledge bases with RAG are already crowded, and 8 stars with no trial path is not enough to justify migrating. #RAG#Self-Hosted#Knowledge Base#AI

    Company knowledge. With receipts. A self-hosted AI workspace for documents, connected business data, and answers grounded in evidence.

    nuanu-aiTypeScript★8

  • C PRISM Counterfactual video generation for a CoRL submission, augmenting data for humanoid robot mobile manipulation, a hot topic right now. But if the repo only has a paper and a few result videos, it helps reproducers little; the org name 'amazon-far' is also easily mistaken, so first confirm whether code is released. #ML Research#Robotics#Video

    [CoRL 2026] Counterfactual Video Generation Enables Scalable Humanoid Loco-Manipulation

    amazon-farPython★12

  • C omarchy-mac-starter A desktop configuration that makes Omarchy look like macOS: frosted effects, familiar shortcuts, optional Apple device integration. It's dotfiles beautification rather than a tool, and its 37 stars come more from Omarchy's popularity. #Linux#macOS#UI Design

    A Mac-friendly Omarchy starter: frosted desktop, familiar shortcuts, and optional Apple device integrations.

    johnloringpollardQML★37

  • C Multimodal-Flow Research code from the hustvl lab unifying multimodal sequence modeling into a continuous-flow framework. The direction is interesting, but the repo has only a one-line abstract, no paper, experiment configs, or weights, so it can't be reproduced yet; wait for follow-up. #ML Research#Multimodal#AI

    A Fully Continuous Multimodal Flow Framework For Flexible Multimodal Sequence Modeling.

    hustvlPython★8

  • C halo-mw2-director A prototype for Halo CE and MW2 mashup cutscenes and a scene editing tool on Mac, requiring your own game assets. The idea is fun but it's still a prototype with no docs or finished demo, and beyond a small circle of cross-game modders it's hard to see who else it's for. #Games#Game Dev#macOS

    Native Halo CE × MW2 cinematic prototype and scene authoring tools for Mac. Bring your own game assets.

    0xburnPython★20

  • C clusterbreak A Pine Script v5 structure-break indicator that marks BOS/MSS, sets stop-loss and take-profit, can push webhooks, and includes multi-timeframe and intraday P&L tables. Plenty of features, but such indicators flood the TradingView community, and without backtest evidence it is just another signal script. #Fintech#Data Analysis#tradingview

    Pine Script v5 indicator that detects BOS/MSS structure, marks a stop and TP when a reversal breaks a swing inside the prior zone, and can send the bracket to a webhook. Supports higher-timeframe stru…

    JferheartLanguage unknown★13

  • C PS5LM Runs GGUF models with llama.cpp on a jailbroken PS5, by author cobanov, a fun technical demo. It requires a cracked console, and shared memory can't run large models; watch it for entertainment, but don't count on it as an inference device. #LLMs#Hardware#ps5

    llama.cpp on a jailbroken PS5: run any GGUF model natively on the console, starting with Qwen 3.8

    cobanovHTML★7

  • C st-tavern-architect A dual-agent pipeline adding 3D scene and building generation to SillyTavern, output rendered live in Three.js. The idea is concrete, but the dependency chain is long and the description leans on showmanship; it is mostly useless outside SillyTavern players. #3D#AI Agents#Animation

    3D Scene & Architecture Generator for SillyTavern via Dual-Agent & Three.js (Real-time Spatial Rendering)

    pathetic777HTML★9

  • C ingetrazo-blender A Blender add-on that imports .igz models with groups, materials and cameras. The problem is that the IngeTrazo format ecosystem is tiny and the author does not explain where the format comes from or the conversion limits, so most people cannot use it; generic glTF/FBX import already suffices, and its value depends on whether you happen to have .igz assets. #3D#Plugins#Images

    Import IngeTrazo (.igz) models into Blender 5 with their groups, components, tags, materials and cameras, and reload them when they change.

    vicc3dPython★10

  • C badfisher-ai-log Java application log analysis with LLM-based anomaly diagnosis, following the common pattern of feeding logs to a model for explanation. Too many such services have appeared this year, and the repo doesn't say which model it uses, how it reduces noise, or how it beats ELK plus an LLM plugin, so differentiation is weak. #Observability#LLMs#DevOps

    Java application log analysis and AI anomaly diagnosis service

    badfisheryhyJava★21

  • C scbz A Go SDK and REST gateway for Indonesian tipping platform SociaBuzz, covering hundreds of payment methods across 18 countries. Ready-made glue for teams doing paid content in Indonesia, but the applicability is very narrow and it's a wrapper around a single commercial platform, so it can only follow that API's changes. #SDKs#APIs#Fintech

    Go SDK & REST gateway untuk SociaBuzz donations. 18 negara, 100+ metode pembayaran

    hirotomasatoGo★12

  • C skills Two Agent Skills, one for UI consistency and one for Russian phrasing, claimed to work with Claude, Codex and Cursor. They are essentially prompt packs with unverifiable effects and no English version, so the audience is narrow. #Agent Skills#UI Design#Prompting

    Two Agent Skills that help an AI agent make an interface clear, consistent, easy to use and scalable, and Russian text clear and correct (English version planned), without the signs of AI-generated wo…

    nicktureLanguage unknown★8

  • C seoo-skills A free SEO skill pack for AI agents in SKILL.md format. It falls into the prompt-collection category, so its value depends on what is actually written inside, and the repo offers no credible account of its effects. #Agent Skills#Marketing#Prompting

    Free SEO skills for your AI agent (SKILL.md format). MIT licence.

    seoo-toolsLanguage unknown★9

  • C minecraft-dungeons-2-skin-manager A small tool for changing skins in Minecraft Dungeons II, supporting Java skin conversion and 3D preview. The function is clear but the audience is narrow, the 'ultimate' wording overstates the actual completeness, and similar skin tools already exist. #Games#3D#Images

    The ultimate skin manager for Minecraft Dungeons II. Convert Java skins, preview in 3D, organize your library, and install with one click. Supports Second Skin Layer, 3D Skin Layers, and MRES textures…

    ClaspTransmuterC++★11

  • C next-scada-hmi An attempt at open-source industrial SCADA/HMI: visual configuration, real-time data binding, alarms and trend curves. The direction is scarce, but listing five or six subsystems at once with no protocol drivers, deployment method or demo is a typical big-platform vision; for similar work, look at FUXA first. #IoT#Self-Hosted#Web Frontend

    An open-source industrial SCADA/HMI platform supporting visual configuration, real-time data binding, alarms, trend curves, and custom components.

    agentthinkTypeScript★7

  • C MetalCyan Adds Metal acceleration under macOS for mining boards like the ASRock BC-250, based on a modified NootedRed. The audience is only hackintosh users with unusual hardware, and the description openly says LLMs helped debug it, raising stability doubts; the closest reference is NootedRed itself. #macOS#GPU#Hardware

    Metal acceleration project for the ASRock BC-250 on macOS. Based on ChefKiss' NootedRed project. LLMs are used in this project for debugging.

    amethyst8118C++★10

  • C psgfx A small tool for changing icons and background images of PS5 homebrew and games, using the PS5 Upload payload. The audience is only console tinkerers running homebrew, the function is single-purpose and tied to a specific upload tool version, with no cross-platform value. #Games#Hardware#ps5

    Custom icons and backgrounds for your PS5 homebrew and games. Works through PS5 Upload's payload.

    elripaldaC++★8

  • C Antigravity-Mobile An unofficial Android control client for Antigravity, local-first with a Windows execution station. The description piles up too many terms and the boundaries are vague, and its unofficial status means it could break at any time; not recommended for now. #AI Agents#Android#Self-Hosted

    An unofficial, local-first Flutter/Android control plane for Antigravity, backed by a Windows Fleet Station for local execution, orchestration, and multi-slot management, with LAN/USB connectivity and…

    mohgomaa-artDart★9

  • C nodos-map A visualization page turning UNAM theses into a semantic map. It targets users of a single school and shows no technical novelty, looking more like a course or thesis project. #Data Analysis#Web Frontend#Knowledge Base

    Mapa semántico de las tesis de la UNAM

    ssebastian-diazzHTML★9

  • C remocn-studio Calls itself an 'agent-native video editor' with no other explanation of features, interface or usage. The idea may be big, but to readers it is just an unverifiable slogan. #Video Editing#AI Agents

    Agent-native video editor

    RemocnTypeScript★37

  • C XPSemu An experimental branch porting the original Xbox emulator Xemu to jailbroken PS5s. Technically novel, but it depends on jailbroken consoles and homebrew payloads, making it useless to ordinary developers; the description is one line and doesn't say which layers changed or whether it runs. #Reverse Engineering#Game Porting#ps5

    Original Xbox emulator (Xemu) port for the PS5

    ZiZc3C★14

  • C newsroom Calls itself a "personal world news briefing," but does not say where the news comes from, whether models run locally, or how to install and start it. RSS-plus-LLM-summary tools are already plentiful, and a personal project with no usage instructions is hard to justify trying; the author needs to state clear capability boundaries. #Productivity#Automation#Knowledge Base

    Your personal news briefing of the world events

    yogthosClojure★11

DSkip82
  • D caleb-video-editor A browser video editor self-described as generated in one shot, with 48 stars coming more from curiosity about whether slop can work. Video editing is a very broad feature surface, one-shot code is hard to maintain, and it lacks depth in rendering performance and format compatibility, so it may be abandoned at any time and isn't worth sustained tracking. #Video Editing#Video#Web Frontend#AI

    This is a one-shotted slop that works!

    saeedvaziryTypeScript★48

  • D terrahour A terminal world clock with a day/night map and 24-hour timeline, solving the small daily pain of scheduling across time zones; 39 stars in 0.5 days reflects natural CLI-tool spread. Useful but with many substitutes, no data or ecosystem moat, and no paid scenario. #CLI#Terminal#Productivity

    A world clock for the terminal, with a day and night map and a 24-hour timeline to find a time that works across time zones.

    ACoci86Python★39

  • D x-ambient Uses the dominant colors of hovered images and videos in the X timeline to cast ambient light across the page—a clear visual gimmick with a very narrow use case; 49 stars is natural spread for a browser script. It lacks willingness to pay, integration depth, and any path to becoming a dependency, making it a one-off fun project. #Browser Extensions#Images#UI Design#twitter

    Ambient light from hovered X timeline photos and videos, across the whole page.

    mmngaJavaScript★49

  • D 100-themes Provides 100 themes and 200 wallpapers for Omarchy, pure visual assets, with 74 stars from distribution rather than technical innovation; it is convenient for users, but zero differentiation and copyable by anyone in a day, with no moat, no paid scenario, and no dependency potential. #UI Design#Linux#Images

    100 themes for Omarchy, with 200 backgrounds

    bjarneoHTML★74

  • D Adwair Blends WhiteSur's macOS feel with Adwaita's native look into an icon theme, with hand-polished aesthetic value; 49 stars is normal pacing for the theme scene. Theme projects have zero switching cost and no moat, surviving only on sponsorships, with no commercialization and no dependency potential. #UI Design#Linux#macOS

    A refined WhiteSur icon theme blending its polished look with the clean, native feel of Adwaita.

    kusaidaMeson★49

  • D Riot-Account-Checker-Manager Bulk-verifies and manages Riot accounts with an emphasis on low proxy consumption—essentially a credential-stuffing companion tool; 176 stars in 0.3 days is fast, but the users are account sellers, not developers. Technically it is just concurrency scripts plus a proxy pool, with no moat, clear legal and ban risks, no legitimate commercialization, and not worth tracking. #Games#Automation

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

    xiaomingtongxueqlLanguage unknown★176

  • D AStudio2API Reverse-engineers iFlytek Spark into an OpenAI-compatible gateway, with phone-number code login, account-pool keepalive and check-in rewards. It's a lot of fiddly technical work, but the goal is bulk exploitation of free quota, a gray-area integration that can be banned at any time and isn't worth investing in. #Reverse Engineering#APIs#LLMs

    Reverse-engineers AStudio (iFlytek Xingchen / Astron Studio) into an OpenAI / Anthropic-compatible Go gateway: multi-protocol, phone-number verification-code login, account pool keep-alive, check-in b…

    2951461586Go★7

  • D ps-ppl-bypass Uses a vulnerable driver from Process Explorer to bypass PPL protection, one link in an attack chain for disabling protected processes. Technically substantive, but its purpose is basically fighting security software and it depends on a driver already under scrutiny, so it is not suitable to recommend. #Exploits#Windows#Reverse Engineering

    Process Explorer vulnerable driver PPL Bypass

    R41N3RZUF477C★10

  • D 100-themes-day One hundred light themes for Omarchy, complementing an earlier dark set. Ready-made material for Omarchy users, but essentially a collection of color files that solves no new problem and has no evaluable technical substance. #Linux#UI Design#Terminal

    100 light themes for Omarchy, the day versions of 100-themes

    bjarneoHTML★14

  • D Overcast The description is a single line, "weather is also a necessity," with no indication whether it is a watch face, a notification or a widget, and Android weather apps are already oversupplied. #Android#UI Design#Productivity

    Weather is also Essential

    sameeraswKotlin★16

  • D linux-security-hardening A beginner-oriented Bash audit script that checks basic Linux security items and gives advice. Honestly written but ordinary; mature options like Lynis and CIS-CAT exist, and it doesn't say which checks it covers or how it differs from Lynis, making it hard to justify switching. #Linux#App Security#Automation

    Beginner-friendly Bash script to audit and recommend Linux system security improvements.

    emmanuelygrShell★8

  • D drama-song-ad An operational workflow for taking a narrative ad from script and song breakdown to finished film, claimed to be tool-agnostic. In essence it's a prompt plus process template whose results depend on your image, audio, and video models; there's no code to evaluate. #Content Writing#Video#AI

    Make a complete drama song ad: a long-form ad sung as a drama, from script to song to finished video. Originally designed for the MaxFusion AI MCP, tool-agnostic: works with any image, audio and video…

    holy-templarLanguage unknown★16

  • D kaiten Calls itself the control plane for the software you sell, but offers only a one-line vision with no indication whether it handles licensing, billing, or releases, nor how to integrate. Not enough information to justify opening the repo. #Fintech#APIs#DevOps

    Kaiten — the control plane for software you sell

    kaitencloudGo★10

  • D AISafetyHot-Hub A list repo summarizing AI safety papers and links daily, with md, bib and json data plus agent integration notes. It is a convenient feed for people tracking the field, but it is essentially content aggregation with no tool or methodological contribution, a typical list-style repo. #AI#Science#Knowledge Base

    Daily AI safety picks and paper lists (md / bib / json), with Agent integration instructions · AI Safety HOT

    wuyoscarTeX★7

  • D spinlens An internet-jargon detector inspired by a minigame, but it does not say whether judgments come from rules or a model, or what the output looks like. The concept is interesting but the final form is vague, more a casual experiment, and accuracy cannot be judged. #Productivity#Social#AI

    spinlens is an internet rhetoric detector inspired by "Big Fat Fish Defense"

    AolibibiJavaScript★8

  • D dsh-boot-animation-pro An enhanced version adding fullscreen intro animations to DSH, with playback controls, time-based intros and library management; the audience is basically existing DSH users. The description admits it's a fork of someone else's project, a decorative tweak for a small circle with little general value. #Animation#Terminal#CLI

    DSH intro boot animation enhanced edition: playback control, trigger rules, session lists, time-based intros, clip library management, Chinese-English bilingual. A full-frame intro animation for DSH. …

    Yiheng-guoJavaScript★7

  • D jev-llm-guard Claims to be an OWASP LLM guardrail based on Jev, with a Turkish demo. The problem is Jev itself can't be verified, and the repo doesn't say which OWASP risks it covers or its false-positive rate; guardrail tools without evaluation data are basically unusable. #LLMs#App Security#AI

    Contextual OWASP LLM guardrail powered by Jev, with a Turkish interactive demo

    gulbakiJavaScript★9

  • D bootloops Claims to offer a 'certified' high-precision physics and quantitative computing engine for LLM agents, but there is no tool list, accuracy validation or usage example. The vision far exceeds what is visible, and the 'certified' claim cannot be verified. #AI Agents#Science#LLMs

    BootLoops 1.0: certified computational tools and house engines for exact and high-precision physics and quantitative science, built to be driven by LLM agents. MIT; docs CC BY 4.0.

    BootLoops-aiPython★18

  • D ZeldaWWHDRecomp A static recompilation/port attempt for The Legend of Zelda: The Wind Waker HD, aimed at players wanting the original on PC. But the repo currently has only a title, with no build steps, dependencies or asset notes, so it can't be used today; similar projects are plentiful and 15 stars don't prove progress, so wait for the README to be filled in. #Games#Game Porting#Reverse Engineering#zelda#wiiu

    Zelda Wind Waker HD Recomp

    ZeldaWWHDRecompC++★15

  • D aether-clash A pure HTML/Canvas 2D fighting game with a YouTube tutorial, positioned as a teaching demo rather than an operable product. Those wanting to learn Canvas fighting implementations can look through it, but as tutorial companion code it has almost no reusable value. #Games#Game Dev#Web Frontend

    2D browser fighting game — Mecha vs Demi-Human. 12 chibi fighters, cinematic ultimates, smart CPU, mobile touch controls. Pure HTML/CSS/JS + Canvas, no build step. Tutorial: youtu.be/UN_0bNC2wTU

    bangtutorialJavaScript★8

  • D self-love A Skill that turns movie footage into comic-style animation, with rules and 698 frames of data, and copyright for the material listed separately, which is honest. But the use is mostly personal creation, results depend on an undisclosed model pipeline, and ComfyUI's style-transfer workflows are comparable; not worth dedicated study. #Image Generation#Agent Skills#Animation

    Movie original-pixel sample analysis and photo-to-comic/animation conversion Skill; full rules and 698 frames of data, original parts under MIT, movie materials with separate rights boundaries.

    chaneX69HTML★15

  • D RapistOps Claims to do retention and correlation analysis of sexual violence information; the subject is heavy, but the project name itself is highly offensive and the description offers only grand vision without data sources, privacy boundaries, or usage. Not recommended to open. #Data Analysis#Knowledge Base#Privacy#Documents

    A survivor-driven accountability intelligence system for preserving, connecting, and analyzing fragmented information about sexual violence.

    meadowriversPython★13

  • D dai-asr-i18n The title is only "David AI's ASR internationalization benchmark," with no dataset source, evaluation dimensions or reproduction steps, clearly a byproduct of vendor self-testing. It may be useful to people doing multilingual speech evaluation, but there is not enough information to judge credibility, so skip for now. #Speech#Benchmarks#AI

    David AI's ASR I18N Benchmark

    withdavid-aiPython★15

  • D research-chamber It claims to study open models in a "pain mode" experiment, but the repo has no method description, no data scale and no reproduction steps. Outside readers have no entry point; it reads more like private experiment notes than a tool, so there's little to see for now. #AI#LLMs#ML Research

    Code, data and findings behind researchchamber.fun: experiments on open AI models with a pain pattern switched on

    0xD3btPython★12

  • D WARDOGS-2026-AIMBOT-WALLHACK- Provides aimbot, wallhack, and resource-tampering cheat tools, with 62 stars coming from cheaters rather than the developer community; technically it is just injection plus memory read/write, locked in an arms race with anti-cheat and subject to bans at any time. There are clear legal risks, no legitimate commercialization, and no reusable technical value. #Game Cheats

    These tools give you control over combat, resources, progression, and experimentation in your game

    Masszumplain17Language unknown★62

  • D vev Claims to be a Jev-style decision model running on Qwen3.5 with image input, but neither Jev nor Qwen3.5 is a verifiable public object, and the README doesn't explain weight sources or how to run it. Anyone wanting a self-hosted multimodal decision model gets nothing actionable until reproducible evaluations appear. #LLMs#Multimodal#AI

    Jev-like decision models that can also see images — open weights on Qwen3.5, run on your own GPU

    XiaooolongPython★8

  • D Bot-Zenith-Testnet Indonesian scripts for auto-claiming testnet faucets, bulk-creating wallets and checking balances, serving the few chasing testnet airdrops. The functions are straightforward, but such faucet scripts are everywhere and break whenever testnet rules change, so it isn't worth adopting as a tool. #Crypto#Automation#Blockchain

    Bot otomatis klaim faucet, buat wallet & cek saldo di Zenith EVM Testnet

    YusufAbdulahJavaScript★8

  • D dsh-node-addon-require-builtin A Node native plugin used internally by DeepSeek Harness that lets addons require built-in modules. The description is honest and marks it as an internal component; it has no standalone value for third parties unless you're writing native extensions for DSH. #Build#SDKs#Plugins

    An internal component used by DeepSeek Harness

    deepseek-aiC++★9

  • D tf3-modkit A 40-module modifier for Transport Fever 3 that lets you tweak economy, towns and vehicles with one click. It is a single-player game cheat tool, outside our recommendation scope, and highly likely to break with game updates. #Game Cheats#Games

    40-module trainer/cheat for Transport Fever 3 — economy, towns, industries, vehicles and world tools

    ChameleonBoil6Language unknown★34

  • D dspy-system-one-agent-patterns Applies System 1/2 thinking to a set of DSPy agent patterns, but the description is empty and only 29 stars support it. It doesn't say what problems these patterns solve or how they fit DSPy's existing modules, reading more like personal experiment notes. #LLMs#AI Agents#Prompting

    cmpnd-aiPython★29

  • D jailbreak-archives A structured archive of adversarial prompts, claimed for red-team and defense research. It's still essentially a prompt collection whose value depends entirely on curation and versioning, and such repos can be assembled in three days; defenders are better off reading OWASP LLM Top 10 and public red-team datasets. #Jailbreak#Prompting#LLMs

    A structured, versioned archive of adversarial prompt engineering techniques against Large Language Models — for AI safety research, red-teaming, and defensive hardening.

    e2syRoff★9

  • D relapse-y2jb Ports the Relapse exploit to the Y2JB PS5 jailbreak chain. This falls under console cracking, with an audience limited to users on specific firmware versions; the code's lifespan is measured in weeks and it dies once patched, so it isn't worth collecting as a tool. #Exploits#Jailbreak#Reverse Engineering

    Relapse exploit ported to Y2JB.

    edisnordJavaScript★16

  • D ComfyUI-Fisher-Pose-i18n Just English, Japanese, and Chinese copy localization for the ComfyUI Fisher Pose node, an extremely narrow scope. Somewhat convenient for Chinese users of that node, but it contains no capability of its own and its value depends entirely on the upstream project. #Plugins#Image Generation#UI Design

    Fisher Pose · Camera & Pose for ComfyUI [en/jp/cn]

    hiderminerJavaScript★8

  • D BasicLaravelReact-template An empty scaffold template with no description; 44 stars in 0.6 days looks more like mutual upvoting, and nothing distinguishes it from Laravel's official scaffold plus React. #Web Frontend#APIs#Learning

    markleduna321JavaScript★44

  • D pkg-mutant-shop A PKG store and installer for jailbroken PS4/PS5, with a Windows client. Its core use is installing game packages of unknown origin, making it part of the piracy chain; not recommended. #Piracy#Games

    PKG MUTANT SHOP - a homebrew PKG store and installer for jailbroken PS5 and PS4, with a Windows companion.

    XavyProdC★10

  • D network-recon-toolkit A teaching-oriented scanning tool wrapping Nmap in Python, described as educational. People doing real asset mapping would use Nmap or Naabu directly, leaving this repo with only the value of a learning example. #Network Security#CLI#Automation

    Beginner-friendly network reconnaissance tool using Python and Nmap for educational port scanning.

    emmanuelygrPython★8

  • D password-security-lab An interactive CLI demonstrating hashing, salting and authentication flows, positioned as teaching. The content is correct but goes no further than any introductory article on password storage; fine as a security exercise, but there's no need for it to exist as a tool. #App Security#CLI#Learning

    Interactive Python CLI demonstrating secure password hashing, salting, and authentication workflows.

    emmanuelygrPython★9

  • D Rex-Widgets-Hub Uses Cloudflare Workers and KV for online subscription and collection management for some module called Rex, but the description gives no background on what Rex is, so outside readers cannot understand or use it. #Plugins#APIs#Web Frontend

    An online Rex-specific module/module collection storage and subscription tool deployed with Cloudflare Workers and KV storage

    MakkaPakka518JavaScript★13

  • D genpark-sparkfly-influencer-marketing-creator-intelligence-skill A vendor's Agent Skill wrapper for its SparkFly influencer marketing platform, handling TikTok/Instagram/YouTube creator discovery and campaign attribution. It's essentially a manual for a commercial API, unusable outside that platform, with no evaluable methods or data. #AI Agents#Agent Skills#Marketing

    GenPark AI Agent Skill - SparkFly (Tec-Do 2.0) global creator intelligence, TikTok/Instagram/YouTube influencer discovery & campaign attribution.

    alphaparkincPython★7

  • D ai-companion-telegram-guide An illustrated tutorial for connecting an AI companion to Telegram; it's a content repo rather than a code project. Useful for people who want to build one following along, but outside today's scope of looking at new tools. #LLMs#AI#Learning#telegram

    Bring your AI companion into Telegram: MochiBot × OpenRouter × VPS hands-on tutorial

    liora-valeLanguage unknown★27

  • D break-your-assumptions An interactive experiment demonstrating assumptions software takes for granted. The concept is not bad, but the description stops at slogans and does not explain what the experiment is or how to run it. The same organization opened three similar repos in three days, looking more like unpacked course material than a standalone tool. #Learning#Testing#Code Quality

    Interactive experiments in the assumptions that break software. Predict, break, repair, and keep the test.

    wenslink-osJavaScript★10

  • D build-the-world A local engineering experiment claiming to be "technology for building worlds," with executable tests, but it is all vision rather than a content list, so readers cannot tell what they could learn or use today. Highly similar to two other repos from the same organization, it is tutorial material and not recommended. #Learning#Testing#Build

    Build the technologies that built the world. Free, local-first engineering labs with executable tests and build, break, repair workflows.

    wenslink-osJavaScript★7

  • D software-without-the-world Aims to rehearse offline degradation scenarios such as network loss, expired logins and payment timeouts with a local simulator. The topic is well chosen, but the description only lists scenarios and does not say how the simulator is built or how assertions are verified. Three sibling repos launched the same day, looking more like a syllabus than a runnable tool. #Learning#Testing#Debugging

    Cut the internet, break login, or time out a payment. A free local simulator for offline fallbacks, data ownership, and safe recovery.

    wenslink-osJavaScript★10

  • D omarchy-tsukiyo-theme An Omarchy desktop theme with no description and no screenshots, so its effect can only be guessed from the QML source. Theme repos have limited value anyway, and without a showcase almost nobody will install it. #UI Design#Linux

    Arjun010011QML★10

  • D Administracion-fuentes-publicas No description; the name suggests scripts processing public finance source data, and 67 stars clearly do not match the amount of content. With no data sources, pipeline or outputs visible, it cannot be evaluated, and inflated stars cannot be ruled out. #Fintech#Data Analysis#Documents

    BquantFinancePython★67

  • D Map An offline-capable map PWA written in HTML/CSS/JS, a common learning-project template that does not say where the data comes from or how it differs from existing map services. #Web Frontend#Learning#Productivity

    A web-based map application built with HTML, CSS, and JavaScript. It is set up as a Progressive Web App (PWA), so it can be installed on a phone or desktop and keep working with limited connectivity

    Gl1tchRazeJavaScript★9

  • D perplexity-auto-creator Bulk-registers Perplexity accounts with random identities and temporary email, bypassing Cloudflare; essentially an automation script for free quota. It violates the terms of service, can break at any time, and has no reusable technical value. #Scams#Automation#AI

    Automated, end-to-end Perplexity account creator — random identity, temp inbox, session cookies and access token. Cloudflare-aware browser automation.

    0xgetzJavaScript★7

  • D Bambu_Monitor A Bambu printer monitor written in C, but the README does not say which models are supported, how to build it, or where the data comes from. A project with no description is not a usable tool for anyone. #3D#Hardware#Monitoring

    Alloyd21C★11

  • D online-quiz A PHP online quiz system where admins create questions and students answer them, a typical course assignment with nothing new and no deployment instructions. #Learning#Web Frontend#Databases

    A web-based online quiz application built with PHP. It provides separate areas for administrators and students: admins manage quizzes, while students register, log in, and take them.

    Gl1tchRazePHP★7

  • D Learning-Analytics-Tools A learning analytics toolkit from an NTNU course project, essentially coursework. The description is just a course name, with no indication of what data it analyzes; unless you're doing similar coursework, there's no reusable value. #Learning#Data Analysis#Web Frontend

    Learning Analytics Tools Implementation Applications (NTNU)

    bigmoumouJavaScript★19

  • D Sekiro-Android-research An undescribed Sekiro Android-related study; it is unclear whether this is a port, an emulator or asset extraction, with no explanation or output. #Game Porting#Games#Reverse Engineering

    LukeValenPython★8

  • D alice_fanart A character fan-art asset library and reference-image guidelines; a content collection rather than a tool project. Useful for a specific fan community, but not an open-source new tool worth ten minutes today. #Image Generation#Prompting#Images

    Alice fan-work asset library and base-image guidelines (Alice / Luo Xiaoshan)

    miyang-aiPython★8

  • D easyread 475 stars in 1.5 days with no description at all; the growth is completely disconnected from the content, a classic case of suspicious star inflation. There is no way to tell what it does, and nobody should spend time on it. #Productivity#Documents#Learning

    EdwardxlaiJavaScript★475

  • D bloodborne_pc Named Bloodborne PC Edition, but with no description or credible technical explanation; most such 'X on PC' repos are empty shells or gimmicks. Real PS4 emulation progress is happening in shadPS4, so this repo has no reference value for now. #Game Porting#Games#Reverse Engineering

    deadinside28C++★13

  • D offensive-combat-reload No description at all, just TypeScript code and a name that sounds like combat reload logic, so it is unclear whether this is a game, a demo or assets. Readers cannot judge its purpose and can only skip it. #Games#Game Dev#Animation

    AiramVieiraTypeScript★16

  • D world-observer An empty-description repo from the KAIST CV lab; the name suggests world-model research code, but there's no explanation, paper link or run instructions. Research repos without a reproduction path can't be evaluated, so skip for now. #ML Research#AI#Video

    cvlab-kaistHTML★10

  • D Beyond-sigma-resources A repo named as a resource, with no description and mostly JavaScript files, a typical list or asset collection with no engineering content to evaluate. #Learning#Productivity#Content Writing

    Devendradhote001JavaScript★23

  • D model-house No explanation at all, so it's unclear whether it's model routing, a model repository, or something else. Ten stars in a day provides no information either; skip it. #AI#LLMs#SDKs

    jikimeTypeScript★10

  • D pi-clm No description, and the name doesn't reveal what it does, so its usability can't be judged; skip it. #AI#LLMs#Productivity

    lolipopshockTypeScript★17

  • D filmcraft The repo has no documentation at all; only the name and Rust implementation hint at film and video processing, and it is unclear whether it is a script or a complete tool. Created one day ago with no stated purpose or usage, there is nothing to evaluate; check back once the author writes a README. #Video#Images#Animation

    storytoldRust★11

  • D photocraft Appears alongside the same author's filmcraft, also with zero description and zero usage notes, only a repo name hinting at image processing. It offers zero information for anyone evaluating it, and mature Rust image-processing libraries already exist, so skip for now. #Images#3D#GPU

    storytoldRust★12

  • D jarvisdemo001 The repo has no description and the name suggests a personal demo, so it's impossible to tell what problem it solves or who it's for. Excluded by rule; check back once the author adds documentation. #AI#LLMs#Learning

    jaturapornchaiPython★24

  • D mybotdemo01 Another undescribed bot demo where only the language and star count offer clues; neither usable nor novel. The kind of repo most easily clicked by mistake while browsing trending lists. #AI#LLMs#Learning

    jaturapornchaiJavaScript★11

  • D hexagon-art No description, created just hours ago, with only the name hinting at hexagonal drawing. It explains neither purpose nor usage, so it can't be evaluated. #Images#3D#Animation

    kieraneglinPython★9

  • D familycfo No description; the name suggests a family finance tool, but with no explanation or usage there is no way to tell how it differs from existing budgeting apps. #Fintech#AI#Productivity

    nmazuzTypeScript★7

  • D CatMaster No description, so it is unclear whether this is a tool, a scraper or a practice project; it cannot be evaluated, skip. #AI#LLMs#Automation

    q734738781Python★13

  • D kabarwarga No description at all, only Blade templates and a star count; there is no way to tell what product this is or judge it. #Web Frontend#Social#Learning

    rizalahmadddBlade★10

  • D brasil-gov No description; the name hints at Brazilian government data, but with no explanation of the API, data sources or usage it cannot be evaluated. #Data Analysis#APIs#Learning

    sibeliusTypeScript★12

  • D drawcraft No description, just a Rust repository; there is no way to tell whether it is a drawing library or a drawing tool, so it cannot be evaluated. #Images#3D#Learning

    storytoldRust★7

  • D NewHypFullSprintScaffold The description is only 'a solution...', so it is impossible to tell whether this is a scaffold, a course assignment or a half-finished project. There is no information anyone could use to judge it. #Learning#Build#Productivity

    A solution...

    Derkorn1541Java★7

  • D pair-extraordinaire-badge Only a name and no description; it's unclear whether it's a badge generator or something else. A repo that can't explain its purpose or maintenance status isn't worth clicking; wait for the README to say what it does. #Productivity#Collaboration#Learning

    brunnojobPython★13

  • D thund-lethalcompany-menu A cheat menu for Lethal Company whose all-caps title boasts BEST. Repos like this often ship unknown executables, ruining multiplayer fairness and posing security risks; not recommended. #Game Cheats#Games#Scams

    BEST Is Lethal Company Mod Menu

    ianward6727Language unknown★20

  • D CashPilot A Kotlin project with no description and no README, so it's unclear whether it's bookkeeping, payments or something else. Repos that don't say what they do are never accepted; wait for the author to add positioning and usage before reviewing. #Android#Productivity#Fintech

    abu-pixelKotlin★12

  • D espanachat A JavaScript repo with an empty description; the name suggests Spanish-language chat, but there's no explanation. It's impossible to tell whether it's a chat app, an AI wrapper or practice, so per the rules, skip it. #Social#Web Frontend#Learning

    apolmigJavaScript★8

  • D ahbpanel A Python repo with an empty description; the name suggests a panel but it's unclear what resources or services it manages. A project that doesn't say what it does can't be evaluated and has no usable entry point; wait for the author to add a line of positioning. #Monitoring#DevOps#Self-Hosted

    ardm6105-dotcomPython★8

  • D gate A Python repo with an empty description; the name gate doesn't reveal whether it's a gateway, approval flow, or access control. With no explanation of what it does, it can't be evaluated, so skip it. #Self-Hosted#DevOps#APIs

    hezhanleiokPython★13

  • D terraria-piscat-bot The description literally says cheat menu, speedhack, aimbot; it's a collection of Terraria cheat scripts. No technical substance, plus the risk of executables from unknown sources; skip it. #Game Cheats#Games#Scams

    -terraria-cheat-menu,-terraria-speedhack,-terraria-glitch,-terraria-aimbot

    iruiz2726Language unknown★11

  • D diablo-4-fixi-menu A Diablo 4 modifier claiming infinite gold, max paragon, and all legendary gear. It's a cheat that breaks the game economy and will almost certainly require running programs of unknown origin; not worth any developer's click. #Game Cheats#Games#Scams

    Diablo IV trainer — infinite gold, max paragon, all legendary items, and god mode.

    ivan-daniels85Language unknown★10

  • D fortnite-respock-swapp-tool Titled Best Mod Menu For Fortnite, it hit 38 stars 0.3 days after creation, a classic star-farming plus cheat-distribution pattern. The feature description is vague and it likely bundles malware; blacklist it outright. #Game Cheats#Games#Scams

    Best Mod Menu For Fortnite Yet

    ivy-cruz155Language unknown★38

  • D NATIVE-EMUS-PS5 The name claims a native PS5 emulator but there's no description or documentation; repos like this are usually gimmicks or name-squatting for search traffic. Don't waste time. #Games#Scams#Reverse Engineering

    klortekhqC++★8

  • D DF-EXP-99240040222 The description is just a hyphen and the repo name looks like an experiment number, so its purpose is completely unclear; excluded per the rules. #Learning#Science#Data Analysis

    -

    shungjanwooLanguage unknown★9

  • D ProjetoTeste The author states it's a test project: no features, no description, no content, just a practice repo. No one needs to spend time on it, and it has no basis for evaluation. #Testing#Learning#Productivity

    Isto é um projeto de teste

    apedromLanguage unknown★7

Show HN 5

S

Show HN: Open-source model routing for coding agents at Astra-level performance

Show HN

An open-source model router that plugs into any coding agent, using HMMs plus classifiers to compress a 10^100 search space and precisely calculating cache-switch gains, claiming frontier-level results at about half the cost; 74 points is the top momentum in this batch, with both a repo and a hosted version. Saving tokens drives willingness to pay; the risk is doubts about benchmark credibility and competition from LiteLLM and Portkey.

A few months ago we started building a model router for coding agents because we thought we could outperform any single model with an ensemble approach. Recently we’ve achieved that milestone and I wa…

#LLMs#AI Agents#APIs#Benchmarks

▲74

B

Show HN: Breadcrumb, record everything on your mac + context manager for AI

Show HN

Turns Mac screens, meeting transcripts and AI conversations into searchable memory, then feeds it to coding agents via 30+ MCP tools, filling Claude Code's missing work context; 14 points plus a downloadable beta and full MCP definitions show it's a product actually in use. Monetization can come from E2EE cloud sync and team plans; the risks are privacy sensitivity and being limited to M-series Macs.

Hi HN, I'm Justin. Breadcrumb records everything you do on your Mac (screen + meetings + AI transcripts + what you and your AI decided) and turns it into memory your AI can search. It's local and encr…

#MCP#AI Agents#Privacy#macOS

▲14

B

Show HN: Helo – An email API from former Postmark folks

Show HN

A transactional and marketing email API from the former Postmark team, differentiated by isolating domains, credentials and unsubscribes for multi-tenant sending; email is high-margin, sticky developer infrastructure and the team's track record is itself an endorsement, with 15 points and the founder answering questions personally. The risk is the red ocean of Resend, SES and others plus migration costs, so it needs multi-tenant scenarios to break through.

Hi HN, I'm Ryan, engineer and co-founder at Helo ( https://www.helohq.com) - an email API for sending transactional and marketing email from your application: password resets, receipts, notifications,…

#Email#APIs#SDKs

▲15

B

Show HN: Janus – Go binary that runs GGUF models via Vulkan on AMD/Intel/Nvidia

Show HN

A single Go binary running GGUF via Vulkan, enabling local inference on AMD/Intel integrated GPUs and bypassing CUDA and Python dependencies; 44 points and 22 stars in a day show real demand for non-NVIDIA backends. It could charge for enterprise local deployment or an inference gateway, but must go head-to-head with llama.cpp, Ollama, and LM Studio on quantization compatibility and ecosystem.

#LLMs#GPU#Self-Hosted#APIs

▲44

B

Show HN: Yantra – an LALR(1) parser generator for C++

Show HN

Solves the problem that semantic actions in Bison-like tools execute during reduction without parent-node context, by building the whole AST first and generating multi-pass walkers; 31 points and 24 stars show recognition from the compiler toolchain crowd. Commercialization is weak, with the path being adoption as a standard dependency by DSL and compiler projects; the risk is a single maintainer and pressure from the tree-sitter ecosystem.

Yantra is a C++ parser generator: lexer, parser, and AST walker all generated from one tool. It builds the whole AST first, then walks it. Most LALR parser generators (Yacc, Bison, Lemon) run your sem…

#Build#CLI#Code Quality

▲31

Hot New Repos

  1. rehan-remade/universal-modder★1546 Point Claude at any game. Skills, tools and the fal MCP that let Claude Code mod almost any PC game you own: recon, reverse engineering, fal-generated art/3D/audio, in-game testing, showcase videos.
  2. nanaism/yomiyasu★941 Agent Skill for Refining AI-Generated Japanese into Natural Japanese / Agent Skill for Refining AI-Generated Japanese into Natural Japanese
  3. OpSafari/hypoarena★553 Scientific hypothesis-discovery workbench: grounded hypothesis-evidence graphs, synthetic literature with planted causal chains, generate-debate-evolve loops over pluggable offline adapters, Elo tournaments recovering planted skill order, paraphrase dedup, Bayesian evidence accumulation, reproducible reports. NumPy core, CPU-only torch extra.
  4. Edwardxlai/easyread★475
  5. pengchujin/livecanvas★293 Turn a sentence into a Live Photo with a designed cover and animation. Turn a sentence into a Live Photo with a designed cover and animation.
  6. ZacharyZhang-NY/Ely-GPUI-Components★271 A component library for GPUI, in light and dark. Every component runs live in the browser.
  7. chasmlol/SkyCraft★255 Play Skyrim as a Minecraft player: Minecraft physics, inventory, blocks and combat inside Skyrim's world (SKSE plugin + Fabric mod).
  8. deepseek-ai/DeepEP-Ascend★199 A high-performance communication library for machine learning training and inference on Huawei Ascend NPUs.
  9. blackbearreloaded/ProsperoEden★196 Experimental Eden port for PS5
  10. blendi-remade/dioramas★181 A free, open-source framework for cinematic, interactive 3D websites with AI-generated assets (Nano Banana 2 + Meshy 7.1 on fal). 20 example sites.

Star gain 1d

  1. debpalash/VoiceStudio+2999★49599
  2. Niko1221/Strata+1665★4697
  3. eternity4719/HowToLiveBetter+1481★33486
  4. pingdotgg/t3code+1481★25527
  5. Louis-CFM/coucou+1244★2455
  6. OffGridPete/Fieldwatch+1063★1271
  7. nanaism/yomiyasu+747★941
  8. rehan-remade/universal-modder+745★1547
  9. affaan-m/ECC+660★268561
  10. mvschwarz/openrig+608★2735