Topics /
The most popular blog posts about ai & machine learning on Hacker News. 1,520 posts.
Literate programming is the idea that code should be intermingled with prose such that an uninformed reader could read a code base as a narrative, and come away with an understanding of how it works and what it does. Although I have long been intrigued by this idea, and have found uses for it in a couple1 of different…
Dual GPU setup: run Qwen 3.6 27B at a Q8 quantization at 80+ tokens/sec with 39GB total VRAM
OpenAI addresses Apple’s baseless lawsuit, corrects claims about its employees, and shares messages documenting what happened.
A substantial improvement in intelligence and behavior over Composer 2, particularly on long-horizon agentic tasks.
I gave Claude Code access to my pen plotter. Not directly. I was the interface between the two machines. Claude Code produced SVG files that I plotted with my pen plotter. With my smartphone I captured photos that I pasted into the Claude Code session, asking Claude what it thought about the pictures. In total, Claude…
We've been experimenting with running coding agents autonomously for weeks at a time.
Andrej Karpathy tweeted a mini-essay about buying a Mac Mini ("The apple store person told me they are selling like hotcakes and everyone is confused") to tinker with Claws: I'm …
if you want to do world-class research, temperament matters more than talent
LFM2.5-24B-A2B: an open-weight 24B reasoning model with 2B active parameters, built for fast, private on-device agents. Beats Qwen3.6-35B-A3B on LiveClaw.
The toolkit for spec-driven development. Write feature specs, not prompts. Ship better software with AI agents that understand your requirements.
I test Unsloth GGUFs of Qwen3.8 27B (Q4_K_M, UD-Q2_K_XL, UD-IQ1_S) with llama.cpp on GPQA Diamond, IFBench, Terminal-Bench 2.1. Q4_K_M matches BF16 and fits an RTX 4090.
GPT-2 is a direct scale-up of GPT-1 , with more parameters and trained on more data. However, it was deemed too dangerous to release by OpenAI:
Anthropic secures multiple gigawatts of next-generation TPU capacity starting in 2027, its largest compute commitment to date.
Welcome to the digital home of Daniel Ball. Explore my portfolio to see the projects I’ve worked on and learn more about my professional journey.
Strands Decider 2B is a small, open source, decision model optimized for fast experimentation, local development, and innovation.
I’ve been in the AI space since ChatGPT first dropped. I’ve toyed around with a lot of Language Models, built random side projects, built a couple from scratch and I’ve spent hours looking at the math behind it all.
This post is a high-level explainer for my Master’s thesis, which involves designing hardware architectures for ultrafast inference and online learning using the Kolmogorov-Arnold Network (KAN) architecture. I’ll assume familiarity with standard machine learning concepts, as well as some understanding of hardware and…
An unreleased Claude model improved the lower bound for the fraction of zeroes of the Riemann zeta function that satisfy the hypothesis, raising it from 41.6% to 67.2%.
1.6万亿总参大模型，训练全程由国产芯片完成
Calibrate your enthusiasm
09 Apr 2026 The Vercel Plugin on Claude Code wants to read all your prompts!
Deep learning has been built around backprop, the only credit assignment algorithm capable of training modern neural nets, including transformer-based language models. Backprop requires differentiability and produces first-order gradients, and deep learning’s architectures, optimizers, and hardware have co-evolved…
Ever since Golden Gate Claude I’ve been fascinated with “steering”: the idea that you can guide LLM outputs by directly manipulating the activations of the…
From the Zed Blog: Run multiple agents at once, in the same window.
Discover the infinite possibilities of the Prompt API.
Most of what an AI coding agent does for me isn't thinking. It's I/O.
We find that roughly half of test-passing SWE-bench Verified PRs written by recent AI agents would not be merged into main by repo maintainers. A naive interpretation of benchmark scores may lead one to overestimate how useful agents are without more elicitation or human feedback.
We compared old and new agent swarms building SQLite from scratch and found that better coordination delivers similar quality at a fraction of the cost.
