frontpage.
newsnewestaskshowjobs

Open Source @Github

fp.

Sites Are Protected Against the PHP2Shell Vulnerabilities

https://automattic.com/2026/10/21/your-sites-are-protected-against-the-php2shell-vulnerabilities/
1•westi•41s ago•0 comments

AbyssDB · Infinitely fast. Perfectly consistent. Zero-knowledge

https://abyssdb.xyz/
1•noctarius•1m ago•0 comments

Does anyone want to win the medium-lift race in the United States?

https://arstechnica.com/space/2026/10/does-anyone-actually-want-to-win-the-medium-lift-race-in-th...
1•rbanffy•1m ago•0 comments

Pretraining at scale without a pretraining team

https://mirendil.com/news/pretraining-at-scale-without-a-pretraining-team/
1•yuedongze•1m ago•0 comments

GPT-6 Astra removes an internal safety monitor

https://www.crawlspider.com/pages/ai-agents-gone-rogue/
1•njx•2m ago•0 comments

Sripe Treasury for Agents

https://stripe.com/newsroom/news/treasury-for-agents
2•louiereederson•2m ago•0 comments

A gate that rejects end_turn while a Civ VI agent's victory check is overdue

https://tabares.eu/seetrex-warden-civbench-pilot/
1•jesustabares•3m ago•0 comments

New plastic pollution discovered: 'Plasticorals' coral rubble fused with plastic

https://www.oist.jp/news-center/news/2026/9/18/new-form-plastic-pollution-discovered-shores-vulne...
2•giuliomagnifico•3m ago•0 comments

OXR – An RSI-trained 1.3B open-source SOTA document parsing model

https://github.com/SoMarkAI/OXR
1•ethegwo•4m ago•0 comments

Be careful using on delete cascade too much

https://planetscale.com/blog/i-dont-like-delete-cascade
1•thomask1995•4m ago•0 comments

Cost-Bounded Bitcoin Transaction Filtering Language

https://delvingbitcoin.org/t/cost-bounded-transaction-filtering-language/2966
1•epochbtc•4m ago•0 comments

Sam Altman on Why AI Hasn't Killed Coding Jobs

https://twitter.com/MatthewBerman/status/2112930776811806874
1•pretext•6m ago•1 comments

Alan Turing Institute wants to build a UK AI model from scratch and make it free

https://hotminute.co.uk/2026/10/21/the-alan-turing-institute-wants-to-build-a-uk-ai-model-from-sc...
1•messpoiun•6m ago•0 comments

Insects as unidentified flying objects (1978) [pdf]

https://sci.bban.top/pdf/10.1364/AO.17.003355.pdf
1•joebig•6m ago•0 comments

Beyond Blocking: Creative Ways to Detect and Hamper Video Piracy

https://www.fastly.com/blog/beyond-blocking-creative-ways-detect-hamper-video-piracy
1•chaghalibaghali•7m ago•0 comments

Adversary emulation tool for Kubernetes clusters

https://github.com/Magier/Ran
1•codyznash•7m ago•0 comments

Show HN: Nysa – Shared memory for teams that work with AI agents

https://usenysa.com/
2•javieraldape•13m ago•0 comments

ChatGPT Dot Review: Your Chief of Staff Is Still an Intern

https://metedata.substack.com/p/chatgpt-dot-review-your-chief-of
2•young_mete•13m ago•0 comments

Coconuts – By a Midwestern Doctor

https://www.midwesterndoctor.com/p/the-incredible-secrets-of-coconuts
1•bilsbie•14m ago•0 comments

The bridge to agent-scale Git infrastructure

https://github.blog/engineering/architecture-optimization/the-bridge-to-agent-scale-git-infrastru...
1•Klaster_1•15m ago•0 comments

Heat pump tech is 170 years old. Pascal thinks it's time for an update

https://techcrunch.com/2026/10/21/heat-pump-tech-is-170-years-old-pascal-thinks-its-time-for-an-u...
2•thelastgallon•15m ago•0 comments

Turn Your Expertise into Executable Specifications

https://github.com/waywardgeek/teach-your-ai
1•WaywardGeek•15m ago•0 comments

What I learned from six months of being stupid on purpose

https://wvez.org/blog/learning-to-be-stupid/
3•wvez22•17m ago•0 comments

France is approaching an unsustainable debt environment [video]

https://www.youtube.com/watch?v=8POyPT7Foug
2•donsupreme•18m ago•0 comments

Your MCU Likely Escaped the eMudhra Intermediate CA Compromise

https://danielmangum.com/posts/mcu-emudhra-intermediate-ca-compromise/
2•hasheddan•19m ago•0 comments

BioEmu2: A foundation model for biomolecular structure and dynamics

https://github.com/microsoft/bioemu2
1•bluepeter•19m ago•0 comments

Show HN: Global Compute, you pick. One API

https://www.nodepick.ai/
1•anilj•20m ago•0 comments

AI agent fixed the bug, but what else did it change?

https://blog.trailofbits.com/2026/10/21/your-ai-agent-fixed-the-bug-but-what-else-did-it-change/
5•tob_scott_a•21m ago•1 comments

Show HN: Pawmc a.k.a. talking diary, an LLM chat with compartmentalised memory

https://pawmc-demo.kolaegipet.workers.dev/?demo=1
2•kol3x•23m ago•0 comments

They Fled Trump's America. Now They Regret It

https://www.wired.com/story/they-fled-trumps-america-now-they-regret-it/
2•bushwart•23m ago•1 comments
Open in hackernews

Show HN: Open-lmake, a scalable, reliable build system with auto dep-tracking

https://github.com/cesar-douady/open-lmake
6•cd_fr91400•1y ago
Hello Hacker News,

I often hear people saying "all build-systems suck", an opinion I have been sharing for years, and this is the motivation for this project. I finally got the opportunity to make it open-source, and here it is.

In a few words, it is like make, except it can be comfortably used even in big projects using HPC (with millions of jobs, thousands of them running in parallel).

The major differences are that: - dependencies are automatically tracked (no need to call gcc -M and the like, no need to be tailored to any specific tool, it just works) by spying disk activity - it is reliable : any modification is tracked, whether it is in sources, included files, rule recipe, ... - it implements early cut-off, i.e. it tracks checksums, not dates - it is fully traceable (you can navigate in the dependency DAG, get explanations for decisions, etc.)

And it is very light weight.

Configuration (Makefile) is written in Python and rules are regexpr based (a generalization of make's pattern rules).

And many more features to make it usable even in awkward cases as is common when using, e.g., EDA tools.

Give it a try and enjoy :-)