Browser compiled into a 200M-parameter model's weights: DNS, HTTP and HTML parsing, no LLM API
cephaloform · x · 2026-11-11
A developer demoed a wild experiment: compiling an entire browser into a model's weights — a tiny 200M-parameter model handling DNS queries and answer parsing, IP selection, HTTP requests and HTML parsing, tested over real HTTPS with exact replay. No training or LLM API involved; the runner is Python/NumPy, with socket transport and TLS still external. The author's point: why use harnesses that make tool calls when you can put the tool call in the weights?
More from coding & agent
- Claude-built game has zero assets: every texture and model is code, everything is shaders — mark_k · 2026-11-11
- Memgraph open-sources Context Graph: shared cross-session memory for Claude Code and Codex — JeremyCMorgan · 2026-11-11
- Fable 5.2-generated PR touches 3,000+ files, adds 441,926 lines in one change — cramforce · 2026-11-11
- Birthday gift: an AI agent emergently deployed the exact observability stack I wanted — sloppenheimer · 2026-11-11
- No Claude account? This DSH plugin lets you try Claude-style Whiteboards anyway — dotey · 2026-11-11
- Running 6 concurrent ds41-flash agents: persistent enough to need pi-durable coordination — dbreunig · 2026-11-11