W40 This Week's Picks

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

Turns any article into an HTML-rendered explainer video in seconds at about $0.04 each, pushing video from dollars-per-minute to cents-per-second and creating a new content cost structure. 118 points plus MCP, Chrome extension and ChatGPT plugin distribution validate demand, and Scrimba is a mature YC-backed company that could sell B2B PR/documentation video subscriptions; the risk is rough visuals and easy copying by Loom, Canva and other large players.

Hi HN, I’m Per, founder of Scrimba (YC S20). We’ve spent the last decade teaching people how to code with an HTML-based video format. We’ve now plugged an LLM into it, so that people can create explai…

#Video#AI#MCP

Raven calls itself a 'harness of harnesses,' building a self-evolving multi-agent ecosystem to support RSI (recursive self-improvement), with 4,807 stars and 53 HN points—one of the strongest early momentum in this batch. If the RSI narrative pans out, it could become the infrastructure layer for agent orchestration. The risk is that the concept is ahead of its time, engineering feasibility is doubtful, and it competes head-on with mature orchestration frameworks like LangGraph.

#AI Agents#AI#ML Research

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.

#AI Agents#Productivity#DevOps

Provides deterministic guardrails for multi-tool agents: a data-oriented policy language plus DualLLM, claiming leakage drops from about 10% to single digits without sacrificing utility. Enterprise agent deployment needs permission and leak-prevention consoles, a clear budget line backed by papers and benchmarks, but Cedar/OPA and big-vendor native policy systems compete, and only 23 HN points suggests real customer adoption is still needed.

Hi Hacker News! Matvey, one of the authors, is here. While building enterprise agents, we ran into a problem: the more tools you connect to the AI, the higher the chance it will run out of control and…

#AI Agents#App Security#Privacy

Embeds executable commands and SQL directly into Markdown notes, eliminating copy-paste between notes and the terminal; 74 points is the top HN heat in this batch, proving the pain is widely validated. Monetization can come from ledge-server team collaboration paid plans. The risk is many competitors like Obsidian and its plugins and various runnable-note tools, plus security concerns from executing arbitrary commands.

Hi HN, Ledge is a Markdown notebook that runs shell commands, code, SQL, etc from inside your own notes. I built Ledge because I spend much of my day copy/pasting commands from my notes into the termi…

#CLI#Terminal#Productivity

Solves the high barrier of initial Burp/Caido setup and APK traffic capture, with zero-config out-of-the-box use as the core selling point; 30 Show HN points show practitioner resonance but no breakout yet. Monetization can be pentest team subscriptions or SaaS. The risk is that Burp and Caido already own mindshare, security tool trust builds slowly, and single-person maintenance can't support enterprise features.

This is Katriel, the architect of the tool APIaxess. I have been API pentesting for a good time now, and starting with API pentesting was a bit of a rough patch. Specially the setting up and having to…

#App Security#APIs

A semantic layer with strict naming and partition-aware routing lets LLMs reject wrong definitions, solving inconsistent metrics in self-service analytics and agent data retrieval; 18 points is low because it's a closed-source beta. It offers 25 free users plus enterprise subscriptions, directly against Looker and Cube. The risk is that closed source weakens spread and dbt Semantic Layer already occupies the ecosystem.

Hello HN, I'm Ajo and I built Strata. I spent 4 years at Netflix solving self service for non-technical business users. I think I cracked it with my unique approach to semantic layer design. The key c…

#Data Analysis#MCP#LLMs

A local-first Markdown editor across Mac, iOS and Web with in-place rendering, draggable tables and WebRTC peer-to-peer sync where the server only sees a random room ID, earning 62 points for 10 months of product polish. Free plus optional sync subscription is a clear path, but the editor space is extremely crowded with Bear, iA and Obsidian, so it must target a niche through typography and private sync.

I've spent the last 10 months building Beauty: a local, powerful greco-roman Markdown editor where the page looks finished while you type. It runs on the Mac, iPhone and in the browser, and the browse…

#Editors#Documents#Productivity

Open-source cross-platform multitrack recording and professional audio editing, filling the open-source ecosystem's long-standing gap for a mature DAW; 131 HN points and 145 stars in 0.7 days strongly validate demand. Monetization can come from support, plugins and team licensing, but it's unlikely to become a standard dependency. The risk is that Ardour and Reaper already own mindshare, open-source audio monetization has always been hard, and the momentum may be one-off.

#Audio#Editors