LIVE · ranked by stars/day

Catch open-source AI
while it's still accelerating.

Stars are a lagging vanity metric. We rank by velocity — stars/day, right now — kill ~95% of the noise with AI, and hand-curate what's left. A 5-minute read, every day. Free.

4.3M repos tracked94 editions shippedlinks verified live
Accelerating nowLIVE
sorted by stars/day · re-ranking live · sample data

The signal

Velocity, not vanity.

GitHub Trending refreshes faster than your coffee and star counts lag reality by months. We measure the only thing that matters early — how fast a project is gaining ground.

0
AI repos tracked
across the firehose
0
repo growth,
year over year
0
of noise killed
by AI, for cents
0
to read the
whole edition

Why this exists

Picking the wrong repo is expensive.

Adopt an abandoned or hyped dependency and you pay in rewrites, broken builds, and supply-chain risk — weeks of lost work.

01

The firehose never stops

Dozens of AI repos ship daily — 4.3M and counting. No human keeps up, and Trending is gone by lunch.

02

Wrong calls cost weeks

One dead dependency means rewrites and risk. Knowing what's maintained and accelerating saves real time.

03

Star lists lie

Stars are lagging and gameable. A 50k-star repo can be frozen; a 200-star one can be this week's breakout.

The product

See exactly what you get.

Every edition is a tight, opinionated shortlist — the call already made for you. No dashboards, no digging.

  • ✓A verdict, not a list. Each pick is tagged USE, WATCH or SKIP — with the reasoning.
  • ✓Velocity up front. Stars/day and trend, so you see momentum at a glance.
  • ✓Verified at publish. Live links, no archived junk, no star-farmed hype.
accelbrief.com / 2026-06-06 · voice & speech AI
EDITION · VOICE & SPEECH AI
What's accelerating in open voice this week
axon/whisper-streamUSE▲ 842 /day

Real-time streaming ASR that finally runs on a single GPU. Active maintainers, clean API — safe to build on.

verdant/voicebox-ttsWATCH▲ 511 /day

Expressive TTS with voice cloning. Quality is striking but the license is unsettled — promising, not production.

relay/speechkit-legacySKIP▼ stale

Still riding 40k old stars but last real commit was 11 months ago. The velocity spike is a fork, not the original.

What you get

Five minutes that save you hours.

Hours back, every day

One read replaces scanning GitHub, HN, Reddit and arXiv. We do the digging; you get the shortlist.

Mistakes avoided

We flag what's maintained and accelerating — and what to skip — before you commit a dependency.

Caught early

Ranked by stars/day, not totals — so you see breakouts while they're still an edge, not after everyone moved.

Curated, not aggregated

Every pick is deep-read and noise-filtered by hand. No dead links, no archived junk, no hype padding.

The call we make for you

What to use — and what to avoid.

✓ Worth your time

✓ Actively maintained, accelerating infrastructure
✓ Real substance — engines, frameworks, tools you ship on
✓ Verified live: not archived, links checked at publish
✓ Backed by genuine velocity, not vanity stars

✗ Skip these

✗ Abandoned forks & dead repos riding old stars
✗ Star-farmed or gamed popularity
✗ Hype skill-packs & tutorials with no substance
✗ Day-one spikes mistaken for real traction

One wrong dependency costs weeks. One good call pays for a year of reading us — and we're free.

The archive

Every edition, free.

94 shipped and counting. Each a different slice of the open-source AI stack.

LATESTZCode leads at 3615 stars/day as agent tooling dominates→09·20DeepSeek Harness leads as everything becomes a plugin→09·19fast-jev-compaction leads at 1759 stars/day→09·18handraw-style leads at 186 stars/day as CapCut and world-model repos surge→09·17reef leads as agent self-improvement meets agentic CRM and browser use→09·16Octop leads open-source AI with 475 stars/day→09·15Cloudflare security-audit-skill leads today's AI repo surge→09·14Codex ecosystem dominates: codex-chatgpt-web leads at 237 stars/day→09·13FreeToken hits 121 stars/day as agents get docs and courts→09·12GenOffice leads as Kimi K3 runs on one CPU→09·11Colibri runs frontier MoE models in pure C at 110 stars/day→09·10WeChat intelligence hub leads with 406 stars/day→09·09Experiential leads: BYOK gateway tops open-source AI surge→09·08Scroll-craft leads: AI agents build scroll-driven sites→09·07Commerce agents top the chart at 453 stars/day→09·06Magnitude tops the list: local AI inference servers surge→09·05Ponytail tops: AI that codes less, plus DeepSeek plugins→09·04WorkBuddyGuide tops the list as AI coding agents go practical→09·03Enterprise world models lead open-source AI picks→09·02Codex with ChatGPT leads: AI pairs planning with action→09·01Mono-color skill tops +188 stars/day; agents edit video→08·31Devspace leads: MCP harnesses, ADHD skills, Chinese writing tools→08·30OpenExecutive tops with 296 stars/day as AI exec team→08·29FreeToken tops the chart: datacenter-scale AI on your desktop→08·28FuXi tops the day: terminal AI agents, SRT whiteboards, Kimi K3 on CPU→08·27Shuohao-skills tops AI short-drama agent pack→08·26Sprix Sage Router tops AI agent routing picks→08·25Apache Maka leads: local-first AI agent workspaces→08·24Graft tops the list: AI coding agents get a speed boost→08·23DSH-routing-suite tops the day; agent harnesses flood in→08·22DeepSeek Harness tops the day: everything is a plugin→08·03ADHD skill tops open-source AI, plus Codex jailbreak pack→08·02penecho’s shared canvas tops a day of agent skills and mirrors→08·01QM tops open-source AI acceleration with multiplayer agent harness→07·31Gemma 4 runs in 2GB RAM on MacBooks, Claude watches video→07·30Ponytail leads: AI that codes like a lazy senior dev→07·29Claude-of-Duty leads with 627 stars/day as FPS built from single prompt→07·28Firstmate leads: talk to one agent, ship with a crew→07·27Anti-AI slop leads: Hallmark, video-shotcraft, A-stock data→07·26Colibri runs 744B MoE on 25GB RAM — plus iFixAi, SkillOpt→07·24CodexPlusPlus leads as no-ai-slop tool surges→07·23Harness Engineering leads as img2threejs and Odysseus accelerate→07·22t8y2/dbx leads: lightweight DB client, PriceAI, Codex proxy, TabFM→07·21AOS Community Edition leads open-source AI tools surge→07·20Codex CLI jailbreak leads ADHD-friendly AI tools→07·19OpenWiki leads as AI docs tool hits 262 stars/day→07·18Codex-Dream-Skin leads at 2223 stars/day→07·17Claude watches video, Tencent sandboxes AI agents→07·16xai-org/grok-build leads with 6370 stars/day→07·15Ponytail tops AI repos with lazy senior dev logic→07·14Postgres rewritten in Rust hits 100% test pass rate→07·13AI Berkshire leads as value investing meets multi-agent AI→07·12Synthetic Sciences leads AI science tools, DashiAI slides, LLM Wiki→07·11Skills surge: nature-skills leads, Colibri runs 744B on 25GB→07·10PixelRAG leads as pixel-native search replaces web parsing→07·09Caveman Claude slashes 65% tokens, free LLM proxy stacks 16 providers→07·08Self-hosted AI workspace leads open-source acceleration→07·07Autonomous red teaming leads as T3MP3ST tops open-source AI→07·06palmier-pro leads: AI video editor, agent roadmaps, design tools→07·05OpenWiki leads AI docs surge, Loopy agents, HTML video tools→07·04Archify leads with PPT builder, Hermes desktop, and OpenClaude→07·03Omnigent leads as AI agent orchestration surges→07·02TencentCloud CubeSandbox leads AI sandbox surge→07·01Headless browsers, PPT bots, and Lottie: AI repos accelerating→06·30Ponytail tops AI repos with 2441 stars/day for lazy senior dev code→06·29Vercel Eve leads as AI agent frameworks and local inference tools surge→06·28AI Berkshire leads: value investing framework tops 1203 stars/day→06·27Nature Skills tops open-source AI repos with 334 stars/day→06·26Book-to-Skill leads as PDFs become Claude Code skills→06·25Distill how anyone thinks: nuwa-skill and 5 more repos accelerating→06·24OpenMontage +3604★/day — and 5 more accelerating→06·23AI Engineering from Scratch leads with 366 stars/day→06·16Programmatic video, AI filesystems, and Claude hacks lead→06·15Ponytail tops AI repos with 3073 stars/day→06·14NVIDIA SkillSpector and shadcn/improve lead AI security and code audit surge→06·13OfficeCLI leads as AI agents get their own Office suite→06·12OpenSenseNova leads AI office tools, HyperFrames renders HTML video→06·11Nature Skills leads as AI repos turn books into coding skills→06·10Open-source AI — what's accelerating today→06·09Open-source AI — what's accelerating today→06·08Open-source AI — what's accelerating today→06·06Voice & speech AI — what's accelerating→06·06Vector search & agent memory — what's accelerating→06·06RAG & retrieval — what's accelerating→06·06MCP & the tool-use layer — what's accelerating→06·06Inference & serving — the open-weights stack — what's accelerating→06·06Image & video generation — what's accelerating→06·06LLM evals & observability — what's accelerating→06·06AI coding agents & copilots — what's accelerating→06·06Browser & web agents — what's accelerating→06·06AI security & agent safety — what's accelerating→06·05Local & on-device AI — what's accelerating→06·05AI agents & frameworks — what's accelerating→06·05Open-source AI — what's accelerating this week→