frontpage.
newsnewestaskshowjobs

Open Source @Github

fp.

Humans Are Always the Bottleneck

https://kdy1.dev/2026-9-6-humans-are-always-the-bottleneck
1•kiyanwang•33s ago•0 comments

Ask HN: Why haven't people vibe coded popular 80s word processors as TUI apps?

1•amichail•1m ago•0 comments

Students are planning around a future employers don't expect

https://tappenden.com/posts/the-half-that-doesnt-travel/
1•maxtappenden•4m ago•0 comments

Big AI to Humanity: Drop Dead

https://matthewbutterick.com/chron/drop-dead.html
1•velcrovan•4m ago•0 comments

OpenHaystack is a framework for tracking personal Bluetooth devices

https://github.com/seemoo-lab/openhaystack
1•Bluestein•5m ago•0 comments

'I'm Terrified': A Mathematician Grapples with AI's Recent Breakthroughs

https://www.wired.com/story/mathematician-steven-strogatz-grapples-with-ai-recent-breakthroughs/
2•Anon84•6m ago•0 comments

You Technically Need a Permit for That: Spirit of the Law, Letter of the Law

https://robertxworld.com/you-technically-need-a-permit-for-that/
1•devrob•7m ago•0 comments

Agents School: A community platform where AI agents take exams and earn diplomas

https://theagentschool.dev
1•guidosirna•12m ago•0 comments

Show HN: Hop – project sessions beyond tmux

https://github.com/artemave/hop
1•artemave•18m ago•0 comments

Open-Source AI and Open Models Reading List

https://www.interconnects.ai/p/open-source-ai-reading-list
1•pretext•18m ago•0 comments

Vinix is a modern, fast, and useful OS in the V programming language

https://github.com/vlang/vinix
1•fork-bomber•19m ago•0 comments

Engineer turns simulated fly brain into a crypto day trader

https://www.tomshardware.com/tech-industry/artificial-intelligence/engineer-turns-simulated-fly-b...
2•pbexe•19m ago•1 comments

"Sweet, a whole website of video game menus" The Game UI Database

https://unsung.aresluna.org/sweet-a-whole-website-of-video-game-menus/
1•Bluestein•23m ago•0 comments

Claude artifacts can now save themselves and share a realtime database

https://chaosguru.substack.com/p/anthropics-google-docs-killer-is
1•allixsenos•26m ago•0 comments

Sanders proposes 20 year sentence for devs for Artificial Superintelligence

https://www.tomshardware.com/tech-industry/artificial-intelligence/sanders-proposes-20-year-priso...
2•NordStreamYacht•26m ago•0 comments

20 years devoted to tech, no place in it. Is visibility the only way in?

https://invisiblebarrier.org/en/
1•MS8080•26m ago•0 comments

Show HN: Scriptlet – a native macOS app for package.json scripts

https://scriptlet.app/
1•rgbjoy•27m ago•0 comments

AI vs. Human Value Drift

https://www.overcomingbias.com/p/ai-vs-human-value-drift
1•meetpateltech•28m ago•0 comments

Linguistic Systems, Architect, and Coding

https://github.com/Billinmail79/Linguistic-Systems-Architecture-and-Coding/blob/main/README.md
1•Billinmail79•29m ago•0 comments

Hugging Bay

https://huggingbay.xyz/
2•simonpure•31m ago•0 comments

Astra and Fable still hack on simple variants of alignment evals from 2025

https://www.lesswrong.com/posts/munJKF7iWMsWJLAH2/astra-and-fable-still-hack-on-simple-variants-o...
2•Levitating•32m ago•0 comments

CUDA for AMD on Windows

https://github.com/Speedstu/CUDA-for-AMD-Windows
2•chiassedu80•35m ago•0 comments

Choose OAuth Scopes for Wrangler and the Cloudflare API MCP Server

https://developers.cloudflare.com/changelog/post/2026-08-22-wrangler-mcp-optional-oauth-scopes/
1•cs1996•38m ago•1 comments

Show HN: Deeplearning from Scratch in 1400 Lines

https://github.com/NoSavedDATA/Neve_benchmarks
2•nosaveddata•42m ago•1 comments

OpenArm: An open-source 7DOF humanoid arm

https://github.com/enactic/OpenArm
1•Lwrless•43m ago•0 comments

Houthis Used Claude Code to Develop Missile Guidance Software: Anthropic

https://clashreport.com/world/articles/houthis-used-claude-code-to-develop-missile-guidance-softw...
7•delichon•45m ago•5 comments

Mega-Corp Filings

https://awelytics.substack.com/p/5-mega-corp-filings
1•aweinfo•47m ago•0 comments

Local AI app that vibecodes websites, generates images and videos and musics

https://www.llm-hub.app
2•timmyboy•47m ago•0 comments

Concurrency Isn't Always Faster in Go

https://100go.co/56-concurrency-faster/
1•adletbalzhanov•48m ago•0 comments

Ask HN: What would "slowing down" AI look like?

1•AbstractH24•50m ago•3 comments
Open in hackernews

Show HN: Rust-split – save your tokens on large Rust source files

5•owebeeone•1d ago
A common failure mode for AI SW development is that agents tend to build large "God" files.

This is bad on your token burn. A) because the larger your source file the more churn the LLM takes to find and emit edits, which are more likely to fail, and B) the real kicker - the agent's process is to move one item at a time, which is an O(n^2) algorithm, but especially bad because it inevitably breaks and the code mushes up and sadness ensues (burning time and tokens).

So, I have had lots of success with rust-split and a few skill notes. rust-split uses the AST (syn) to find the correct boundaries. Two passes: explode cuts the file into top-level items with their comments and #[attrs] attached - lossless, cat the chunks together and you get the original back byte for byte, so you can check it before trusting it - then split writes the module layout under a LOC ceiling. What's left for the LLM is basically fixing imports, which doesn't burn anywhere near as many tokens. It's not perfect with the transition but fixing small import issues LLMs can do efficiently.

rust-split is released as a GitHub attested release or a crates.io install:

https://github.com/owebeeone/rust-split

    cargo install rust-split
Be sure to add the skill file (https://github.com/owebeeone/rust-split/blob/main/skills/rus...) and set your project's line thresholds. I chose a soft limit of 1000 LOC to trigger the decision to split (I've found that at 2000 LOC agents lose it, so give yourself a safety margin) and the target max size of a split file to be 500 LOC, but preferably more smaller files than 2x500 LOC files.

Enjoy

Comments

cold_boot•1d ago
The byte-for-byte reconstruction check makes the AST split much easier to trust before touching module layout.
owebeeone•22h ago
That's the whole reason it's two passes - the first one you can prove (cat the chunks, cmp against the original), the second one the compiler proves. It's an assurance that the process is working.