Tools Bench.

Product launches and open-source repos with enough signal to earn a second look.

Last Brew Time: Sep 9, 2026, 11:16 AM PT

Insight

Builders keep teaching coding agents entirely new professions

01.EXTENDING CLAUDE CODE SKILL LIBRARIES INTO HARDWARE AND PHYSICAL DESIGN

02.REBUILDING CORE BUSINESS SOFTWARE AGENT-NATIVE FROM THE GROUND UP

03.WIRING AGENTS DIRECTLY INTO THE TEAM CHAT TOOLS PEOPLE ALREADY USE

Featured

GitHub283.9K

obra/superpowers

An agentic skills framework & software development methodology that works.

Market Signal

Why It Has Market Pull

Superpowers is an open-source Claude Code skills framework and development methodology that has grown explosively since its October 2025 release, becoming one of the fastest-growing developer tools in recent memory. It has real, mixed community debate (not just hype) and was reportedly adopted into Anthropic's official plugin marketplace, giving it strong evidence of durable builder value among similar tools.

  • 283,946 GitHub stars and 25,412 forks as of 2026-09-09, up from roughly 89,000 stars in March 2026
  • 345 open issues with active external contributors filing real feature/bug reports, not spam
  • Hacker News 'Rave Review' thread at 50 points / 25 comments, plus a separate 'Superpowers 6' release thread
  • Reported inclusion in Anthropic's official Claude Code plugin marketplace as of January 2026
  • MIT-licensed, created and still actively maintained by Jesse Vincent (obra), with commits as recent as 2026-09-08

feedbacks

What People Are Saying

  • "Using Claude Code with Superpowers is so much more productive and the features it builds are so much more correct than with stock Claude Code."HN comment

  • "I personally don't like superpowers very much...Claude makes more mistakes when using superpowers than when not."HN comment

  • "I do _not_ like the separation of spec/plan; IMO the models are good enough to get straight to coding once the spec is written."HN comment

  • "STOP everything you're doing and start using SuperPowers"Substack article

  • "I BUILT A BENCHMARK WHERE CODING AGENTS COME TO DIE"GitHub issue

  • "SDD: Approve + Important triggers mandatory re-review loop"GitHub issue

  • "docs: add MiniMax Code marketplace installation"GitHub issue

GitHub103.8K

TauricResearch/TradingAgents

TradingAgents: Multi-Agents LLM Financial Trading Framework

youtube
TradingAgents

Market Signal

Why It Has Market Pull

TradingAgents is a mature, widely-cited open-source multi-agent LLM framework for financial trading, backed by a peer-reviewed paper and nearly two years of sustained growth. It is the most-starred project in this batch by a wide margin and stands out as a genuine reference architecture for multi-agent debate systems, well worth a closer look even outside finance.

  • 103,839 GitHub stars and 19,960 forks, created December 2024, still actively pushed as of September 2026
  • Backed by a public research paper (arXiv:2412.20138) describing the analyst/researcher/trader agent debate architecture
  • Independent backtest coverage found roughly 7% returns over a 30-day window versus 4.5% for the S&P 500, alongside a 22% drawdown warning
  • The project explicitly disclaims real-money trading and documents that results vary by backbone model and market period

feedbacks

What People Are Saying

  • "The most-starred open-source AI trading framework on GitHub."third-party review

  • "A meaningful alpha, but accompanied by 22% drawdowns that most retail traders would not tolerate."third-party review

  • "Backtest results are not guaranteed to match any published figure."GitHub repo

  • "The TradingAgents team explicitly recommends against using real money."GitHub repo

  • "Models a real trading firm by deploying LLM-powered specialist agents who collaborate through structured debate."third-party review

  • "Independent third-party coverage is still thin."evidence gap

Product Hunt510

Switch

Switch brings AI agents into the collaboration tools your team already uses. Agents join as named participants in your Slack, Teams, Discord, or Telegram channels, sharing the same context and history as your team. Connect an agent once and use it across projects. Each room carries its own context, participants, and rules. Works with Claude Code, OpenAI, Google ADK, LangChain, and more. Open source, self-hostable, running in minutes.

github
Switch demo — agents and humans collaborating in Slack

Market Signal

Why It Has Market Pull

Switch is a genuine enterprise-backed product from SandboxAQ's Flint AI line, open-sourced to let any AI agent join Slack, Teams, Discord, or Telegram as a named participant with shared context. Real press coverage and a visibly active internal engineering team give it stronger-than-average builder momentum among similar tools.

  • 510 Product Hunt upvotes, listed under Open Source, Developer Tools, and AI
  • Backed by SandboxAQ; launch covered by PRNewswire, Unite.AI, and several other tech outlets on 2026-08-26
  • GitHub repo (sandbox-quantum/switch): 592 stars, 47 forks, 32 open issues, commits as recent as 2026-09-09
  • Actively developed by a named internal team (Discord/Teams/Slack connectors, OIDC gateway, session recovery), not a one-off open-source drop
  • Released under Apache 2.0 with Commons Clause, free for any team to self-host

feedbacks

What People Are Saying

  • "bring any AI agent into any team chat"PRNewswire

  • "a shared workspace where people and agents operate with the same context, resources, and history"Unite.AI

  • "SandboxAQ Lets Any AI Agent Join Your Slack Or Teams Channel"Tech coverage

  • "fix(discord): anchor inbound message-map on thread root so directly-created threads keep threading"GitHub issue

  • "feat(teams): show that an agent is working — eyes on the message, card in its thread"GitHub issue

  • "rich session activity in the messaging apps (Slack first)"GitHub issue

  • "Integrate shared SDK sessions and recover card publication"GitHub issue

HF Spaces405 likes

MiniMax H3 Turbo LoRA

Video generation with a synchronized soundtrack MiniMax-H3 — unquantized, split across two Spaces Joint video and soundtrack out of a single denoising pass, at bfloat16 with no quantization anywhere. This Space is the denoising half: the 61.73 GiB transformer and the two autoencoders. The 62.14 GiB Qwen3-VL conditioner runs in qwen3vl-conditioner, which this Space calls over the gradio API for every request. The weights are the public MiniMaxAI/MiniMax-H3 diffusers checkpoint. MiniMax-H3 is 195.9 GiB in bfloat16 and a ZeroGPU Space is evicted at 150 GB of storage. An unquantized single Space is therefore impossible, which is why quantized demos of it run NVFP4 or float8 weights. Cut the Mini...

Market Signal

Why It Has Market Pull

MiniMax H3 Turbo LoRA is an official Hugging Face Space published directly by the MiniMaxAI organization, offering a genuine speed boost (roughly 5x fewer sampling steps) for MiniMax-H3's joint video-and-soundtrack generation model. A visible independent ecosystem -- ComfyUI nodes, a Diffusers port, and a from-scratch distillation reimplementation -- shows this has become a reference acceleration method rather than a one-off demo, making it worth tracking.

  • 405 likes on the official MiniMaxAI-hosted Space
  • Independent ComfyUI integrations from two separate builders (Larryvrh and drbaph)
  • Ported to Diffusers format by InstantX for use outside Gradio pipelines
  • Open-source 4-step distillation reimplementation published on GitHub (ModelTC/Minimax-H3-Turbo)
  • Cuts sampling from roughly 20 steps to as few as 4, about a 5x speedup for joint video+audio generation

feedbacks

What People Are Saying

  • "Thanks guys! Keep up the good work"HF discussion

  • "Tested it, still not better than silveroxide's dareties v0.1-v1.0 4-8step experimental lora"HF discussion

  • "much better at keeping the references, even voice references are better"HF discussion

  • "I'm not gonna lie, kinda disappointed with the result. Ref2v model really has quality issues as many claimed."HF discussion

  • "The Sound is a bit iffy, but what can you do"HF discussion

  • "Thank you, i think lot of people will be very happy"HF discussion

Sources

GitHub

affaan-m/ECC

255.0K

The agent harness performance optimization system. Skills, instincts, memory, security, and research-first development for Claude Code, Codex, Opencode, Cursor and beyond.

cathrynlavery/diagram-design

36.4K

38 editorial diagram types for Claude Code, Codex, and Pi. Self-contained HTML + SVG. No shadows. No Mermaid slop.

ayghri/i-have-adhd

33.9K

A skill to stop your coding agent from burying the answer. ADHD-friendly output.

freestylefly/awesome-gpt-image-2

29.9K

Prompt as Code | GPT-Image2 工业级提示词引擎与模板库，530+ 个案例逆向工程，20+ 套工业级模板，并提炼出Skills，持续更新中

Product Hunt

Widgo

418

Widgo is an AI sales rep on your website. It scans your site, adds any docs you give it, and answers every visitor from that alone. It identifies the company, scores the lead, and books the demo on your calendar. One line of code. Free forever, no card.

Kombai Gallery

347

Build standout UIs by giving your agent design taste. Browse curated web and mobile designs, from full pages to individual components. Preview them live and copy a ready-to-use prompt. The gallery is free forever, with no sign-up required.

OpenMarket

279

What if marketplaces rewarded the best product instead of the best marketing? OpenMarket (in research preview), is a multi-agent marketplace where sellers pitch, competitors challenge their claims, and independent truth agents verify the evidence as they compete to earn a buyer's sale. Here, claims don’t count just because they’re written on a product page. They have to survive scrutiny and be proven.

Relaticle

164

Most CRMs bolt AI on. Relaticle is built agent-first: any MCP client gets 37 first-party tools over OAuth, with workspace custom fields in each agent's schema automatically. The in-app assistant is stricter: every write is a proposal showing the exact change, approved record by record. Self-host everything under AGPL, including local inference with Ollama. Free forever self-hosted; Cloud is flat per-workspace pricing, not per-seat.

GoodLads

151

AI growth manager for Google Ads accounts. GoodLads generates hypotheses to improve your ROAS or CPA, ships each one in one click - never without your approval - and tracks it on a kanban board to a verdict.

Jupitrr Cut

116

Jupitrr Cut is an open-source mobile app for recording polished talking-head videos. Use the built-in teleprompter to read line by line, retake anything, and automatically remove silences. Export a finished video without touching a timeline. Record the video. Skip the edit. Add captions and B-roll later with Jupitrr AI—completely optional. No account wall. GPL-3.0 licensed, with the APK on GitHub and one-tap updates through the App Store and Google Play.

YC Launch

Hacker News

TERMy – A fast terminal assistant that does not use LLMs

224 pts

I love research and development, you may have heard of me because of PJON (Padded Jittering Operative Network). It is a network protocol I started developing in 2010, which was recently implemented in silicon by the ETH Zurich university thanks to the research of Pius Sieber. I am excited to share with you TERMy, a terminal assistant built on top of the NPC-Forge framework. Unlike everything else being built today, TERMy does not use embeddings, machine-learning or LLMs. It runs on the CPU (even... (224 points, 45 comments).

Give your AI agent on-screen guides that show users where to click

7 pts

Hey HN. I'm Christian, one of the founders of Frigade (YC W23). I've noticed that a lot of in-app AI agents struggle to actually understand the products they exist in. For instance, let's say a user asks an agent how to do something in a given SaaS product. In an ideal case, maybe that agent replies saying it has a tool to do the task and just automates that work entirely for the user. That's a great outcome. But often that's not the case. Maybe there is no tool call for that exact task, or mayb... (7 points, 4 comments).

Type.com: Multiplayer Codex/Claude in the cloud for non-tech use cases

14 pts

Hey HN, I'm Komran, the CTO and cofounder of type.com. We're launching today. I've been describing type as "the IDE for non technical people." It allows you to build really cool shit without having to worry about maintaining the perfect local dev environment. Everything is cloud first by default. No fumbling with handoff flows or anything weird like that. Every thread is backed by a persistent sandbox VM so you don't even have to worry about git worktrees if you're doing something git backed. Yo... (14 points, 7 comments).

Maxxwell – The IDE for Optimal Tokenmaxxing

9 pts

Hi HN! I’m Michael, one of the founders of Rindler (YC S26). We built Maxxwell, which helps you manage a dozen coding-agent sessions’ progress, context, and blockers. This lets you spend more time with the actual high-leverage decisions (and gives you more confidence in stepping away from your computer). We were originally building browser agents that log in to sites with a cached representation of the site to make for cheap and reliable repeat runs. There were a ton of hard challenges like mapp... (9 points, 8 comments).

Rubato – Retro-Mac desk device mirrors AI coding state-ESP8266

7 pts

Hi everyone! I have built an AI health device (ESP8266 + 240*240 screen). It can show AI Agents' status in real time with breathing bubble. And remind user dringking water, toilet, streth, etc. support: DeepSeek Harness, opencode, OpenClaw, Claude Code, Cursor. Hardware design, firmware and plugins are all opensource. urls are here: https://github.com/lovaxi/Rubato_Device https://github.com/lovaxi/Rubato_Plugins Type-c power supply, 2.4G wifi. Online sell on Tindie: https://www.tindie.com/produc... (7 points, 2 comments).

HF Spaces

Microduck Sandbox

476 likes

Real trained RL policies for the Microduck robot, running fully in the browser: MuJoCo compiled to WebAssembly steps the physics, onnxruntime-web runs the policy network at 50 Hz. No server, no backend. Two locomotion variants of the same robot are included: legs (walking, the default) and rollers (the wheeled skating variant). Press M (or hold D-pad up ~1 s on a gamepad) to switch; the roller model, meshes and policies are lazy-loaded on the first switch. | Mode | Checkpoint | What it does | |--------|-----------|--------------| | Run (legs) | BESTalphawalking.onnx | Velocity-tracking locomotion (arrows / WASD to steer) | | Sit | BESTalphasitstand.onnx | Sits down on its hull, stands back u...

H3 Acceleration Arena

80 likes

Blind A/B ranking of MiniMax-H3 acceleration variants Human-judged ranking of ~26 MiniMax-H3 acceleration variants over a 200-prompt corpus, from blind pairwise votes on pre-generated clips, with confidence intervals, cost and slice breakdowns. The design and its reasoning are in arena/DESIGN.md; the app's own notes are in arena/README.md. This Space is private and must stay private until deliberately flipped. It streams ~3,700 clips out of the private dataset multimodalart/h3-pre-gen-arena. See Going public below. hfoauth: true above creates the OAuth app and injects OAUTHCLIENTID, OAUTHCLIENTSECRET, OAUTHSCOPES and OPENIDPROVIDERURL. arena/space_auth.py implements the flow by hand (this is...

Fruit Fly Simulation

41 likes

Simulate a fruit fly in your browser using WebGPU kernels A standalone fruit fly connectome demo: paint neurons, stimulate the network, and watch an articulated Three.js fly respond. Vanilla JavaScript, Vite, Three.js, and @huggingface/kernels. Simulated neural activity drives crafted walking, turning, and flight animations. The movements are illustrative, not validated predictions of fly behavior. Open the address printed by Vite, then click Download & start. All neural weights, fly meshes, fonts, and kernel templates are included in public/; setup loads them into the browser and caches verified weight chunks. No API key or external model service is required. Deploy the generated dist/ dire...

Wan2.2 14B Fast Preview

1.9K likes

generate a video from an image with a text prompt Wan2.2 14B Fast Preview is a Hugging Face Space tagged with gradio, mcp-server, region:us. It has 1914 likes on Hugging Face.

ProtectBirds

1.0K likes

Protect Birds ProtectBirds is a Hugging Face Space tagged with docker, region:us. It has 1039 likes on Hugging Face.

FLUX.2 Klein multi-LoRA

537 likes

Use multiple FLUX.2-Klein LoRAs Note: This space is experimental and may log image-uploads during certain periods for performance monitoring. Always comply with HF and model Terms of Service. Any stored images are automatically deleted after 7 days. FLUX.2 Klein multi-LoRA is a Hugging Face Space tagged with gradio, mcp-server, region:us. It has 537 likes on Hugging Face.