2341
Silent speech with ultrasound
We trained a model to predict speech from ultrasound video of the tongue, no speaking required. It reads open text, generalizes across people, and reaches a 15.6% word error rate on open-vocabulary speech.
New
The latest blog posts to take off on Hacker News. 80,798 posts.
2342
An agent in 100 lines of Lisp
Around 2000 I took an AI course at the University of Guelph. I don’t think I learned too much. We didn’t talk about neural networks, as far as I can remember. My end of term project was, I think, a pathfinding algorithm wearing an AI costume. There were certainly no discussions of transformers. No CUDA. No PyTorch…
2343
Every new car sold in the European Union must include a driver monitoring camera
The EU's new mandatory ADDW rules could make roads safer by altering distracted drivers, but its privacy safeguards and data handling rules remain a gray area.
2344
We charge $10k a week to delete AI-generated code
Your AI-built product works, but past 100,000 lines every change breaks two things. Three senior engineers make your codebase maintainable again. One week, fixed price, guaranteed.
2345
AI Meets Cryptography 1: What AI Found in Cloudflare's Circl
We pointed our AI audit pipeline at Cloudflare's CIRCL experimental cryptography library and confirmed seven real bugs, from a critical float64 precision loss in threshold RSA to a complete access-control break in attribute-based encryption. All seven are now fixed upstream. This is the first post in a series on bugs…
2346
Astro 7.0
Astro 7.0 brings faster builds with Vite 8, a new Rust compiler, Advanced Routing, background dev server support, and structured logging.
2347
Local, CPU-Friendly, High-Quality TTS (Text-to-Speech) with Kokoro
Just a few years ago, realistic local speech generation seemed unimaginable. Today, its quality is exceptional and, crucially, it delivers these results without compromising privacy.
2348
Notes on Software Quality
“The absence of problems” is the best definition I can come up with for quality.
2349
2350
Better Auth is joining Vercel
Better Auth is joining Vercel. We're joining to accelerate our work on open-source auth and securing agent workflows.
2351
Why we built yet another Postgres connection pooler
PgDog is a connection pooler, load balancer, and sharding proxy for PostgreSQL. Scale Postgres horizontally without rewriting your application.
2352
Microsoft fire idTech team at Id software
Microsoft have fired most if not all of the team at id Software responsible for developing the idTech game engine during Xbox layoffs
2353
Automating AI Away
A. Karpathy once said that OpenAI researchers are effectively "automating themselves away" by improving their AI. Right now I develop Beagle SCM with Anthropic's Fable and it is of course a brilliant model able to spot nits in a mountain of code, file tickets, make fixes. Still, yesterday it managed to commit the…
2354
Chat Control 1.0 and 2.0 Explained
What is the difference between Chat Control 1.0 and Chat Control 2.0? A clear overview of both EU laws, their timelines, and where things stand today.
2355
An iroh powered smart fan
Putting iroh on a bare ESP32 to build a temperature-driven smart fan you can check on from anywhere.
2356
98% isn't much
98% sounds like a lot. If someone wins the lottery 98% of the times they play, they are clearly blessed. Getting a top mark (e.g.: 10/10) on exams 98% of the time will likely lead to an honour diploma.
2357
Europe's company websites are mostly served by US vendors
Vendor attribution study of 19,450 European company websites. Cloudflare is the largest internet-facing infrastructure vendor in every one of seven EU markets.
2358
Top researchers leave USA for the Netherlands (in Dutch)
Het Tulp Fonds dat in 2025 door OCW en NWO in het leven werd geroepen, biedt internationale toponderzoekers de kans om hun onderzoek in Nederland voort te zetten: enerzijds omdat wereldwijd de academische vrijheid onder druk staat, anderzijds omdat hun kennis en expertise een enorme impuls geeft aan de innovatiekracht…
2359
Inference Optimization for MiMo v2.5: Pushing Hybrid SWA Efficiency to the Limit
An end-to-end engineering practice for the MiMo-V2.5 series inference system, covering KVCache management, tiered caching, SWA-aware prefix cache trees, scheduling, Prefill/Decode pipelines, and multimodal optimizations.
2360
How to sequence your own DNA at home
I have now sequenced my own genome 5 times with an Oxford Nanopore Technologies MinION. This means collecting them from a swab, prepping them for sequencing, running them through a sequencer, then doing analysis over them.
2361
NSA and IETF: Fairness
Secret NSA documents showed that NSA pushed DES in the 1970s to "drive out competitors" while knowing that DES was "weak enough" to break; meanwhile NSA publicly claimed that it would use DES . NSA used export-law exceptions in the 1990s to entrench RC4 and RSA-512, causing security problems for decades . NSA in the…
2362
2363
Alternate clock designs and time systems
A day has 24 hours in it. That's 1,440 minutes or 86,400 seconds. If you wonder why Americans don't understand the need for using the metric system then ask yourself why you use this inconsistent time measurement system.
2364
Learning to code is still worthwhile
Literally no one is uttering what we thought were the immortal lines “learn to code,” right? I don't think that sentence has been spoken for many months now in Silicon Valley
2365
GLM 5.2 and the coming AI margin collapse
GLM 5.2 is the first open weights model I'd call a genuine competitor to Opus and GPT for agentic work - at ~15-20% of the price. Part one of why AI inference margins are about to collapse.
2366
Kernel anti-cheat is an overreach
Modern PC games make you install a ring-0 anti-cheat driver you can't audit.
2367
Girls just wanna have fast MPMC queues with bounded waiting
Disclaimer: An earlier version of this post claimed the structure is wait-free, this is incorrect. Being wait-free requires that failure or suspension of any thread can’t cause failure or suspension of another thread. This queue in fact does not fulfill that requirement. The main section which discusses the wait…
2368
Price per 1M tokens is meaningless
It stops being all about the vibes when the API bill hits you. Many companies are now discovering that AI can indeed be pricey. One habit that might be driving up your AI bill is comparing models by $X per 1M tokens ....
2369
Pruning RAG context down to what the answer actually needs
Pruning agent context down to what the answer actually needs, while keeping 96% of recall
2370
Linux on the Atari Jaguar
The challenges around porting Linux to an old videogame console by Atari