Look today31
2026-10
S neurband An open sEMG wristband in the spirit of Meta's Neural Band: 8-channel ADS1298, nRF52 BLE, dry electrodes, with KiCad, FreeCAD and Zephyr published. A rare full-stack reference for EMG hardware hackers; you still fabricate and solder it yourself.
Open, minimal sEMG wristband like the Meta Neural Band: 8-ch ADS1298 + nRF52832 BLE, dry gold electrodes on a flex, 3D-printed band. KiCad, FreeCAD, Zephyr, Python, simulations.
S reylo Imports .sql directly into TypeScript with full type inference and zero codegen, closing the type gap between ORMs and hand-written SQL. The author is the creator of Zod with built-in distribution, evidenced by 656 stars in two days. It could become the standard dependency for SQL type checking; the risk is that Kysely and pgtyped already hold the niche.
Import .sql files into TypeScript with full type inference and zero codegen.
S Show HN: Omnara – Open-source alternative to Claude Managed Agents Turns the agent main loop into a durable harness of a Postgres event log plus stateless workers, solving real pains of losing state on restart, no multi-machine takeover, and idle cost burn. Its 2969 stars and founder's long post show a real user base; open-source core plus hosted version is the classic enterprise-standard play, directly replacing Claude Managed Agents; the biggest risk is Anthropic/OpenAI official hosted services crushing it.
Hey all, Kartik from Omnara ( https://omnara.com) here. We're building an open-source API for running durable agents. You create an agent with an API call, and Omnara runs the agent loop, gives the ag…
S 0c-ime A no-selection Zhuyin input method that uses an on-device Transformer to decide characters from context, supporting Fcitx5, IBus, macOS, and Windows. It addresses a real pain point for traditional Chinese typists, with 110 stars in a day and a clear approach; the risk is that model size and per-platform install experience are unverified, and it can be compared against rime-based solutions.
A Zhuyin input method with no manual character selection: uses a local Transformer to pick characters from context. Supports Linux (Fcitx5/IBus), macOS and Windows.
S stalemap Fills coding agents in on dependency changes after their training cutoff by comparing type definitions and changelogs and writing them into AGENTS.md or CLAUDE.md. It targets the real pain of agents writing outdated APIs with a small, specific approach; effectiveness depends on how parseable each ecosystem's changelog is.
Tell your coding agent which dependencies changed after its training cutoff. Diffs types and changelogs, writes a short block into AGENTS.md / CLAUDE.md.
S git-ai Refactors git into version control AI agents can safely operate, solving the real pain of runaway branches and hard rollbacks when agents write code at high frequency. 1113 stars in two days is a very strong validation signal, and the git layer is a natural gateway for all coding workflows. Path: become a default dependency for agent frameworks or offer enterprise hosting. Risk: git's native worktree/staging mechanisms and IDE vendors building their own agent solutions.
Git designed for AI agents.
S open_scion Version control for coding agents, hitting the core pain of agents frequently changing code and needing rollback and auditable operation records—a genuinely new category beyond git. It passed a thousand stars in 0.8 days with backing from Anthropic's official Scion concept, showing validated demand. Monetization could be SaaS or adoption as a standard agent-platform dependency; risk is Anthropic's official version open-sourcing and crushing third-party implementations.
Open-source version of Scion: version control built for coding agents.
S skim Syncs only the Ethereum state a P2P app actually needs, returning local eth_call in milliseconds via an app-specific node rather than a full node. Directly useful for wallets, MEV, and on-chain monitoring teams; the hard parts are state integrity verification and long-term maintenance cost.
Application-specific Ethereum state client: sync (and archive) only the state your app needs, straight from P2P. BAL-driven, verified, local eth_call in ms.
S verdi_bridge Connects AI agents to a running Synopsys Verdi via MCP to read waveforms, signal values, driver traces, and RTL source. For chip verification people who want models to help inspect waveforms, this has been a gap; the limits are needing a Verdi license and, as a brand-new repository, no real session examples yet.
Connect AI agents to a live Synopsys Verdi session through MCP: waveforms, signal values, driver tracing, and RTL source.
S DishwasherBench Turns a physical simulation of a real dishwasher into a browser mini-game for evaluating language-model embodied reasoning; such concrete, reproducible benchmarks are more useful than generic leaderboards. Limitation: the task is extremely narrow, so don't extrapolate from one machine model.
A dishwasher-loading game on a physical simulation of one real machine, built to benchmark language models and playable in a browser. The game inside is called Spotless.
S warmctl It snapshots a pre-warmed vLLM engine and restores it within seconds, aimed directly at the real problem of slow cold starts in inference services. People running their own inference services can try it today. The limitation is that it is only one day old with 17 stars, and the compatibility boundaries of GPU state snapshots have not been verified.
Snapshot a warmed-up vLLM engine and restore it in seconds
S esp32-iphone-agent Uses a $7 ESP32 for AirPlay screen capture plus USB HID touch emulation so any OpenAI-compatible vision model can operate a real phone, with no jailbreak or developer account; 562 stars in two days shows mobile agent hardware solutions are scarce. It could target mobile automation QA and device farm subscriptions against Appium, with risks around stability, platform terms, and big vendors' built-in agent frameworks.
Let an AI model use a real iPhone with a $7 ESP32: AirPlay eyes, USB HID hands, any OpenAI-compatible vision model. No Mac, no developer account, no jailbreak.
S agent-chrome-relay Lets an agent reuse your already-logged-in Chrome directly, bypassing login state, CAPTCHAs, and fingerprint detection, which is exactly the key bottleneck in browser automation; 251 stars in 1.7 days confirms the demand. It could be monetized as agent infrastructure, with Browser Use and Anthropic's Computer Use as the biggest competitors.
Let your AI agents use your Chrome, signed in as you, in background tabs. By Alexey Indeev · blog: read.aindeev.com · X: @AlexeyIndeev
S hermes-official-plugins Nous Research releases official plugins for its Hermes Agent covering high-frequency scenarios like Home Assistant and Spotify, with 184 stars in 1.1 days showing the ecosystem is growing; it could become a standard plugin distribution entry point and be commercialized, but the risk is that Hermes installs lag behind mainstream harnesses like Claude Code.
☤ The official plugins for Hermes Agent, built and maintained by Nous Research: Home Assistant, Spotify, Claude Subscription DirectSDK and more, one install away from the Hermes plugin catalog.
S tiktok-web-signature Solves reverse engineering of TikTok web signing (anti-scraping checks) so data collection can hit the API directly without a headless browser; 706 stars in 1.2 days shows bypassing anti-scraping is a real need. Commercialization could be a signing SDK or anti-detection hosted service, and it could become a standard dependency for scraping tools like TikTokApi. The biggest risks are high maintenance from TikTok's frequent signature changes, plus platform bans and legal disputes.
tiktok web signature reverse engineer generates valid tiktok signatures for API requets
S mailcheap Connects self-hosted newsletters to Amazon SES, pushing cost per thousand emails to about $0.1 and directly targeting Substack and Mailchimp's high cut; 463 stars in 0.5 days is strong price-driven growth. The commercialization path is clear: hosted version, deliverability monitoring, and subscription analytics as paid add-ons; the biggest risks are commoditization against Listmonk and Buttondown and SES setup friction deterring non-technical users.
Self-hosted newsletters on Amazon SES. About $0.10 per 1,000 emails.
S sqlite-multiwriter Enables concurrent multi-threaded, multi-process writes to SQLite without modifying its source, directly addressing SQLite's long-standing single-writer pain point, with real value for embedded and multi-process services. The 5.7x figure is for a specific workload, so benchmark it yourself before adopting.
Concurrent SQLite writes across threads and processes without modifying SQLite. No forks, no schema changes, up to 5.7x faster writes.
S android-adb-control Wraps ADB operations into an observe-act-verify loop as an Agent Skill for Pi, Codex, and Claude Code, more reliable than letting the model blindly tap the screen; the author has a long open-source record, and it reached 69 stars in just over a day. The limitation is that you still need your own device and ADB environment.
An Agent Skill for Android phone control via ADB: observe, act, and verify. For Pi, Codex, Claude Code, and OpenCode.
S luobo-ppt Lets agents produce editable native PPT slides and beautify them, targeting the real problem that AI-generated slides can't be edited, immediately useful for anyone who presents daily. Layout quality still depends on templates, and it faces direct competition from tools like Gamma.
An Agent Skill for generating and beautifying editable native PPTs
S MacMiniMode Keeps a MacBook awake with the lid closed and no external display, auto-disabling on battery — exactly the gap in macOS. One sentence states the use and limits, it's native Swift, and 100+ stars in two days shows real demand; the limit is that it only serves MacBook users.
Keep a plugged-in MacBook awake with the lid closed: clamshell mode without an external display, auto-off on battery. Let a MacBook stay awake with the lid closed and use it like a Mac mini.
S fixkit Long-press any control in the iOS simulator, say what is wrong, and the issue description lands directly in a Claude Code session, saving screenshots and retelling. Wiring debug feedback into an agent is a new angle worth trying for iOS developers; for now it only covers the simulator, not real devices.
Long press any element of your iOS app in the simulator, say what is wrong, and the report lands in your Claude Code session
S frankensonos A Rust LAN Sonos controller implementing SSDP/UPnP-SOAP/GENA, plus a CLI, HTTP API, and MCP server so agents can queue music directly. Broad but concrete in each part; the limits are that it only serves Sonos users and S1/S2 compatibility still needs real-world testing.
Memory-safe Rust controller for your own Sonos speakers on your own LAN — reliable SSDP/UPnP-SOAP/GENA interoperability (S1 + S2), a Spotify-library classical DJ, and a CLI + HTTP API + MCP server so …
S podcast-to-article An Agent Skill that rewrites multi-person podcast transcripts into publishable long-form articles, with a concrete method: shift the angle, verify names, cut a third, and self-check for AI tone. Content people can use it today; the limit is that results depend heavily on transcript quality, and it is still prompt engineering.
An agent skill that turns podcast transcripts into publishable articles: re-axes into an article, verifies names, trims by a third, and self-checks for AI flavor. An agent skill that turns podcast tra…
S mullion A project translating D3D10/11/12 to Metal so Windows games run on Apple Silicon, a space previously covered mainly by CrossOver's internal implementation with few open-source options. Whether it can cover modern games' feature levels is the real question, and it is still very early.
Direct3D 10, 11 and 12 on Metal, for Windows games on Apple silicon.
S leviathan Compresses JSONL/CSV/SQLite records into a ranked full-text index, solving expensive agent retrieval over large datasets; a single binary hit 273 stars in 0.6 days. Monetization could be an MCP tool or embedded retrieval layer, but SQLite FTS and Tantivy already exist, so it must prove scale advantages.
**Deep memory for agents over large datasets.** Leviathan is a single static binary that turns your records (JSONL, JSON, CSV/TSV, SQLite, or anything a database CLI can export) into a ranked full-tex…
S Show HN: Offrun – manage every coding agent from one workspace Brings multiple coding agents like Claude Code, Codex, and Grok into one workbench, solving the ops blind spot of the parallel-agent era: who is running, who is waiting, how much quota is left; 72 points is the strongest signal in this batch, and the pain scales linearly with agent count. The path is team-level agent ops SaaS; the risk is IDEs and big vendors building parallel panels natively.
Run Claude Code, Codex, AGY, and Grok Build side by side. See who is working, who needs you, and what every account has left.
S muse-gadget-sdk Meta-incubated official SDK for Muse peripherals, letting developers write gadgets for the new interaction hardware directly in C, solving the lack of a development entry point for a new platform; about 181 stars in 5 hours show the community knows the hardware is open. The commercial path is clearest: the SDK is the platform entry point and naturally accumulates into a standard dependency for that ecosystem. Uncertainty is the very short description and unknown Muse hardware shipments and API stability.
Open source SDK to build Muse gadgets
S CadenceSlicer A fork of OrcaSlicer letting two nozzles use different orifice diameters and layer heights in one print. This is a long-missing capability for dual-extruder users and more practical than most slicer alternatives; the cost is syncing with upstream, and it only matters for multi-nozzle machines.
OrcaSlicer fork that prints with two different nozzle sizes in one print, each at its own layer height.
S Show HN: Open-source model routing for coding agents at Astra-level performance 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…
S dots Solves automated browsers being blocked by anti-scraping and human verification, giving agents a stable browser of their own; 1,932 stars in a day shows the blocking pain is real and spreads strongly. Monetization can be a hosted agent browser API against Browser Use and Hyperbrowser. The biggest risk is head-on collision with these mature competitors and compliance and legal red lines around bypassing blocks.
Open-source dots for the web: an AI agent with its own browser, one that does not get blocked.
S DeepGEMM-Ascend Fills in efficient matrix multiplication kernels for Huawei Ascend, supplying the operator layer most missing from non-CUDA ecosystems; 386 stars in 1.3 days benefit from the DeepSeek brand and rigid demand for domestic compute substitution. It's most likely to become a standard dependency in the Ascend training stack and bring enterprise support contracts. The risk is fast Ascend toolchain iteration, high maintenance costs, and CUDA's still-deep moat.
DeepGEMM-Ascend: clean and efficient matrix multiplication kernel library for Huawei Ascend NPUs