Open Weights isn't Open Training
How many monkey-patches does it take to post-train a trillion parameter model?
The latest blog posts to take off on Hacker News. 31,668 posts.
How many monkey-patches does it take to post-train a trillion parameter model?
After several intense and incredible years building Bluesky from the ground up, I've decided to step back as CEO and transition to a new role as Bluesky's Chief Innovation Officer.
Repairing a dead power supply and replacing the NVRAM in a vintage UNIX workstation.
FFmpeg is truly a multi-tool for media processing. As an industry-standard tool it supports a wide variety of audio and video codecs and container formats. It can also orchestrate complex chains of…
Social media was once hailed as the great democratizer of information. It promised a world where anyone could share their voice, connect with like-minded communities, and discover authentic human perspectives. But in 2025, the landscape looks very different. Social platforms are drowning in a flood of Generative AI…
How ParadeDB uses principles from search engines to optimize Postgres' Top K performance.
We’re pleased to announce the release of LibreOffice 26.2, the newest version of the free and open source office suite trusted by millions of users around the world. This release makes it easier than ever for users to create, edit and share documents on their own terms. Designed for individuals and organizations…
The European Commission has accepted our request, and starting from today – Friday March 6 – has added the Open Document Format ODS version of the spreadsheet to be used to provide the feedback. We are grateful to the people working at DG CONNECT, the Commission’s Directorate-General for Communications Networks…
The Software Freedom Conservancy provides a non-profit home and services to Free, Libre and Open Source Software (FLOSS) projects.
We analyzed 4 million IRS 990 filings to map where $3 trillion in U.S. nonprofit revenue actually flows. The data is public — nobody had compiled it.
Pushing affordable marine conservation forward with autonomous marine robots powered by Raspberry Pi Compute Module 4.
Sarvam open-sources 30B and 105B reasoning models trained in India: MoE LLMs with leading Indian-language benchmarks; weights on AI Kosh and Hugging Face.
One of the simplest tests you can run on a database:
Today we are excited to announce the Release Candidate (RC) of TypeScript 6.0! To get started using the RC, you can get it through npm with the following
Slack will be the Waterloo of closed data.
Defend yourself against tracking and surveillance. Circumvent censorship. | Good Bad ISPs
Experience the engineering behind today’s leading health wearables including Oura Ring, Dexcom G7, and Jabra Hearing Aids. Industrial CT scans reveal how these devices merge medical precision with consumer design.
Employees who are impressed by vague corporate-speak like “synergistic leadership,” or “growth-hacking paradigms” may struggle with practical decision-making, a new Cornell study into “corporate BS” reveals.
Claude Opus 4.6 found 22 Firefox vulnerabilities in two weeks, 14 rated high-severity, with proofs-of-concept and candidate patches for Mozilla.
Liberty has costs, but it's worth it.
A statement from Dario Amodei
We introduce "observed exposure," combining LLM capability with real Claude usage data, and share early evidence on employment effects.
Google has worked with the open source communities to build OpenTitan, the first open source silicon Root of Trust (RoT).
Introducing GPT-5.4, OpenAI’s most most capable and efficient frontier model for professional work, with state-of-the-art coding, computer use, tool search, and 1M-token context.
The online advertising industry has built a massive surveillance machine, and the government can co-opt it to spy on us.
A prompt injection in a GitHub issue triggered a chain reaction that ended with 4,000 developers getting OpenClaw installed without consent. The attack composes well-understood vulnerabilities into something new: one AI tool bootstrapping another.
Google Safe Browsing missed 84% of confirmed phishing sites in our dataset. Here's what we found and what it means.