New
The latest blog posts to take off on Hacker News. 80,828 posts.
Writing an LLM from scratch, part 22 – training our LLM
Finally, we train an LLM! The final part of Chapter 5 of Build an LLM (from Scratch) runs the model on real text, then loads OpenAI’s GPT-2 weights for comparison.
ImapGoose
ImapGoose is a small program to keep local mailboxes in sync with an IMAP server. The wording “keep […] in sync” implies that it does so continuously, rather than a one-time sync. ImapGoose is designed as a daemon, monitoring both the IMAP server and the local filesystem, and immediately synchronising changes. When…
Acrobat is intrusive, slow and non-customizable
Acrobat is intrusive, slow and non-customizable. Of course there are alternatives, specifically bad ones...
Cheap DIY solar fence design
A year ago I installed a 4 kilowatt solar fence . I'm revisiting it this Sun Day , to share the design, now that I have prooved it out.
A Gemma model helped discover a new potential cancer therapy pathway
We’re launching a new 27 billion parameter foundation model for single-cell analysis built on the Gemma family of open models.
Getting syntax highlighting wrong
Applying human ergonomics and design principles to syntax highlighting
Things I've learned in my 7 years implementing AI
Even though the impacts of LLMs have never been seen before, they feel familiar to earlier assumptions.
Recursive Language Models (RLMs)
We explore language models that recursively call themselves or other LLMs before providing a final answer. The aim is essentially unbounded context, and a way through context rot.
US Passport Power Falls to Historic Low
For the first time since the Henley Passport Index was created 20 years ago, the United States is no longer ranked amongst the world’s Top 10 most powerful passports.
Are hard drives getting better?
After 13 years, the data is clear—the bathtub curve does not apply to hard drives. And, more importantly, hard drives are failing less and lasting longer.
Claude Haiku 4.5
Claude Haiku 4.5, our latest small model, is available today to all users.
Zed is now available on Windows
From the Zed Blog: Zed for Windows is finally here. Download it today.
Recreating the Canon Cat document interface
Exploring the emergence of user conventions in a minimal system.
Exploring PostgreSQL 18's new UUIDv7 support
Exploring what's interesting about UUIDv7 support using a demo crab store.
You are the scariest monster in the woods
We shouldn’t worry about AI destroying society because there’s something scarier out there (it’s us)
A kernel stack use-after-free: Exploiting Nvidia's GPU Linux drivers
This article details two bugs discovered in the NVIDIA Linux Open GPU Kernel Modules and demonstrates how they can be exploited. The bugs can be triggered by an attacker controlling a local unprivileged process. Their security implications were confirmed via a proof of concept that achieves kernel read and write…
Pwning the Nix ecosystem
ptrpaws's blog on reverse engineering, programming, vr finds and miscellaneous stuff.
iPad Pro with M5 chip
Apple today introduced the new iPad Pro with M5, delivering a huge boost in performance and taking the next big leap in AI for iPad.
Apple Vision Pro upgraded with M5 chip
Apple introduced the upgraded Apple Vision Pro, featuring the powerful M5 chip, the Dual Knit Band, and innovative features with visionOS 26.
Apple M5 chip
Apple today announced M5, delivering advances to every aspect of the chip and the next big leap in AI.
I almost got hacked by a 'job interview'
I was 30 seconds away from running malware, Here's how a sophisticated scam operation almost got me, and why every developer needs to read this.
Garbage collection for Rust: The finalizer frontier
Abstract Rust is a non-Garbage Collected (GCed) language, but the lack of GC makes expressing data-structures that require shared ownership awkward, inefficient, or both. In this paper we explore a new design for, and implementation of, GC in Rust, called Alloy . Unlike previous approaches to GC in Rust, Alloy allows…
Reverse engineering iWork
So you don't have to.
Leaving serverless led to performance improvement and a simplified architecture
Every millisecond matters when you're in the critical path of API authentication. After two years of fighting serverless limitations, we rebuilt our entire API stack and slashed the end-to-end latency.
Bots are getting good at mimicking engagement
Click fraud protection in one tool: DataCops flags bot clicks, blocks repeat bots at your Cloudflare edge, keeps them out of the conversions 8 ad platforms learn from, and exports Google refund evidence.
The scariest "user support" email I've received
Hi, it's Takuya. As your app grows in popularity, you occasionally start to attract attacks aimed directly at you—the developer or site owner. Just the other day, I got one that was honestly terrifying, so I'd like to share it. The Email Subject: Cookie consent prevents platform
Europe's Digital Sovereignty Paradox – "Chat Control" Update
Can you build tech independence while breaking encryption?
How to stop Linux threads cleanly
Stopping a Linux thread is a surprisingly annoying affair. In this post I present common pitfalls and some solutions -- although no truly satisfactory method exists.