vibehacker
News
Hugging Face / Microsoft ·

Microsoft releases FrogNano-4B, an open-weight 4B coding agent trained only with RL

Built on Qwen3.5-4B under Apache 2.0 and trained with reinforcement learning on about 1,500 synthetic repo tasks (no stronger-model trajectories), FrogNano scores 61.5% on SWE-bench Verified and 37.6% on SWE-bench Pro in Microsoft's own five-tool Leaf harness at ~131K context. Microsoft calls it a research artifact: Python-heavy, BF16 weights need about 9.3GB, and quantized builds aren't validated yet.

More news

View all

REA, an open-source MCP server and CLI that lets coding agents reverse engineer binaries, tops GitHub trending

Launched Oct 7 by developer morluto and past 10,400 stars, REA hands Claude Code, Cursor, or a terminal the ability to decompile native, Electron/JS, .NET, and web apps, trace how a feature works, and recreate it in your own stack, with v4.1 adding Android APK (JADX), firmware (Binwalk, Unblob), and IDA Pro support. Analysis runs locally, and the project says it doesn't recover original source or clone whole apps…

Startup Fortune

Spotted something we missed? Start a thread.