frontpage.
newsnewestaskshowjobs

Made with ♥ by @iamnishanth

Open Source @Github

fp.

Open in hackernews

Show HN: Semble – Code search for agents that uses 98% fewer tokens than grep

https://github.com/MinishLab/semble
4•Bibabomas•1h ago
Hey HN! We (Stephan and Thomas) recently open-sourced Semble. We kept running into the same problem while using Claude Code on large codebases: when the agent can't find something directly, it falls back to grep, reading full files or launching subagents. This uses a lot of tokens, and often still misses the relevant code. There are existing tools for this, but they were either too slow to index on demand, needed API keys, or had poor retrieval quality.

Semble is our solution for this. It combines static Model2Vec embeddings (using our latest static model: potion-code-16M) with BM25, fused via RRF and reranked with code-aware signals. Everything runs on CPU since there's no transformers involved. On our benchmark of ~1250 query/document pairs across 63 repos and 19 languages, it uses 98% fewer tokens than grep+read and reaches 99% of the retrieval quality of a 137M-parameter code-trained transformer, while being ~200x faster.

Main features:

- Token-efficient: 98% fewer tokens than grep+read

- Fast: ~250ms to index a typical repo on our benchmark, ~1.5ms per query on CPU (very large repos may take longer)

- Accurate: 0.854 NDCG@10, 99% of the best transformer setup we tested

- MCP server: drop-in for Claude Code, Cursor, Codex, OpenCode

- Zero config: no API keys, no GPU, no external services

Install in Claude Code with: claude mcp add semble -s user -- uvx --from "semble[mcp]" semble

Or check our README for other installation instructions, benchmarks, and methodology:

Semble: https://github.com/MinishLab/semble

Benchmarks: https://github.com/MinishLab/semble/tree/main/benchmarks

Model: https://huggingface.co/minishlab/potion-code-16M

Let us know if you have any feedback or questions!

Where Have All the Cycles Gone? [video]

https://www.youtube.com/watch?v=B-aDBB34o6Y
1•tosh•1m ago•0 comments

Show HN: Typeset sitelen pona and copy a PNG (for toki pona speakers)

https://sitelen.vercel.app/
1•vitaly-pavlenko•3m ago•0 comments

Scaffold a complete 1990s Geocities-themed static website in seconds

https://gitea.com/sugardaddyapp/geocities-boilerplate
2•hankerapp•6m ago•0 comments

Age Assurance on the Internet: Identity, Privacy, and the Limits of Verification

https://sphericalcowconsulting.com/2026/04/14/age-assurance/
1•mooreds•7m ago•0 comments

Bun's problem may be developing in the open

https://00f.net/2026/05/17/developping-in-the-open/
1•avinassh•7m ago•0 comments

Polylux: Open-source replacement for Asus Armoury Crate that saves 470MB RAM

https://github.com/vladulus/polylux
1•polylux•8m ago•0 comments

U.K. Economy Accelerates to Outpace U.S. as War Headwinds Loom

https://www.wsj.com/economy/u-k-economy-accelerated-at-start-of-year-7c70878b
2•mmarian•12m ago•1 comments

Breakdown of X algorithm published on GitHub

https://twitter.com/javilopen/status/2055681064085938429
1•guiambros•13m ago•0 comments

Show HN: Agnt – Free open-source CLI to run any public or MIT-licensed AI agent

1•meajsinghk22•16m ago•0 comments

Show HN: Due Diligence Agents – 13 AI agents for M&A contract analysis

https://github.com/zoharbabin/due-diligence-agents
1•zoharbabin•16m ago•0 comments

Realtime Stocks Prices that you can put Fullscreen

https://stocks.stagas.deno.net/
1•stagas•16m ago•0 comments

Open Itinerary

http://bryanhu.com/open-itin/
1•thatxliner•18m ago•1 comments

A Family Secret No More

https://www.nytimes.com/2026/05/15/us/degrange-family-history-race.html
1•js2•18m ago•1 comments

Checking Multiplication Overflow

https://lemire.me/blog/2026/05/06/checking-multiplication-overflow/
1•tosh•19m ago•0 comments

CarTalk-AMP: Search and Play Car Talk Episodes

https://cartalk-amp.exe.xyz
1•indigodaddy•19m ago•0 comments

Back in My Day

https://old.reddit.com/r/opencodeCLI/comments/1tftit9/back_in_my_day/
3•theanonymousone•20m ago•0 comments

Show HN: Kami Subs – Live AI subtitles overlay for any browser video

https://github.com/MohammdKopa/kami-subs
2•mohamedkeba•21m ago•1 comments

Neptune: Direct3D Virtualization for QEMU

https://blog.getutm.app/2026/introducing-neptune-direct3d-virtualization-for-qemu/
1•linolevan•21m ago•0 comments

Almost there: Aptera has built five solar EVs on their validation assembly line

https://aptera.us/first-five-vehicles/
1•TeaVMFan•21m ago•2 comments

Why this 26-year-old software engineer doesn't want a promotion

https://www.reuters.com/markets/on-the-money/why-this-26-year-old-software-engineer-doesnt-want-p...
2•tartoran•24m ago•0 comments

Can Bloom Energy build them?

https://theloadgrowth.substack.com/p/can-bloom-energy-build-them
2•agordhandas•27m ago•0 comments

WHO declares Ebola outbreak in Congo, Uganda emergency of international concern

https://www.reuters.com/business/healthcare-pharmaceuticals/who-declares-ebola-outbreak-congo-uga...
2•tartoran•27m ago•0 comments

F.03 Livestream – Watch a team of humanoid robots running a full 8-HR shift [video]

https://www.youtube.com/watch?v=luU57hMhkak
1•Teever•28m ago•0 comments

A bad blog post led to GamerGate and the rise of the alt-right

https://www.dontbeasucker.blog/p/you-are-what-you-beat
1•betterthanever•28m ago•0 comments

Curated list of 736 verified remote tech companies hiring (May 2026)

https://dj5333111-ux.github.io/traction-data/
3•tractiondata•28m ago•1 comments

Dontsurveil.me

https://opencivics-labs.github.io/dontsurveil.me/c22.html
1•laurex•29m ago•1 comments

The clean-up cost of AI-generated code is what the velocity narrative leaves out

https://webflow.com/blog/cleanup-cost-ai-generated-code
2•Brajeshwar•30m ago•0 comments

AI can fix the fragmented online public transport space

https://simianwords.bearblog.dev/ai-can-fix-the-fragmented-online-public-transport-space/
1•simianwords•34m ago•0 comments

He Federal Data Field Guide

https://www.federaldatafieldguide.us/
1•firexcy•37m ago•0 comments

Pokemon NPCs Powered by Local LLMs

https://www.owenmc.dev/posts/local-inference-npcs
1•owenmccadden•38m ago•0 comments