Homescale, PlanetScale at home.

Just yesterday morning, my brand-new MINI Countryman auto-installed software update 03/2026.54. How exciting! The release notes promised groundbreaking chang...

Here’s a conjecture: If you put any significant amount of text on the internet under different names, those identities can be linked using only the text itself.

A worked out example of different job queueing semantics and how they affect workloads.

Found yourself on a data project and have no idea what they all are talking about? Feel excluded from all the fun discussions in the office kitchen? If only there were a humongous guide going over all the concepts and buzzwords...

A German research consortium has released Soofi S 30B-A3B, an open language model trained entirely on Deutsche Telekom's cloud infrastructure in Munich. The model uses an efficient hybrid architecture that activates only a fraction of its 31.6 billion parameters per token, keeping throughput steady even at very long contexts. With a training dataset deliberately weighted toward German, Soofi S tops all fully open competitors on both German and English benchmarks.

Yale researchers have helped to describe a new species of Colobus monkey living high in the forest canopies of the Democratic Republic of Congo.

I’ve been refraining from saying ‘vibe networking’ or ‘vibkrotik’ or some other abomination, but it is true - I have been using LLMs to setup a few networks these last few months, and things have generally gone over swimmingly.

Parents can maximize the impact of their support by ensuring their gifts alleviate financial stress or help their kids move closer to their long-term goals.

A text-mode disk usage visualization utility. Contribute to patonw/leaves development by creating an account on GitHub.

One More Letter is a daily word game where every round adds one letter and rearranges them into a new answer.

Async/await makes concurrency easy to write, but incredibly complex to operate. Read on for why the Tokio/Rayon boundary is failing in production, and how explicit state machines fix it.

Mojibake is a fast, self-contained Unicode 17 library for C11 and C++17.

Some apps are sharing users' health data. New research uncovers which apps lock down your privacy, and which don't.

The Scunthorpe steelworks has been officially nationalised under new government powers passed this week.

Can large language models perform deep technical comprehension of computer architecture papers -- not summarization, but structured critique that names the core mechanism, surfaces buried assumptions, and connects a contribution beyond its own scope? We study Gauntlet, an open-source pipeline that analyzes a paper through five independent expert-persona reviewers and an adversarial synthesis stage. On 20 ISCA 2025 and HPCA 2026 papers, ten researchers each wrote their own analyses and then judged, for papers other than their own, the human analysis against Gauntlet's. Across the 20 comparisons evaluators preferred Gauntlet in 15 (human in 4, one tie); its advantage is significant on per-analyst totals (paired Wilcoxon, p < 0.01) and largest on Critical Rigor, vanishing only on Calibration. Where humans win, it is on trust and usefulness rather than depth: a confident wrong claim, a mechanism described but not taught, or unprioritized breadth. A 98-paper automated ablation shows the gain comes from the multi-agent structure -- the pipeline beats the same model run as a single rich-persona agent on 96% of papers -- and specifically from its synthesis pass. We release all analyses, scores, and the rubric as a community resource.

My latest paper, A Test of the Coase Conjecture Using Prices of Electronic Books, with the excellent Tim Groseclose, has just been published. The Coase Conjecture is another one of Coase’s little ideas — the original paper is six pages — that has spawned hundreds of follow-up papers and thousands of citations. The idea is […]

I’m tired of this phrase and this simple way of thinking about tools. This blog post is a wandering train of thought on the topic of what tools are and why it matters to be even slightly more mature in how we think about them.

Imagine the world as a street. Everyone lives on Dollar Street. The poorest to the left and the richest on the right. Every else live somewhere in between. Where would you live? Visit Dollar Street and see homes from hundreds of homes from all over the World.

We reverse all of the core design principles behind the Macintosh human interface guidelines to arrive at the characteristics of the Internet desktop.

Owen Flowers and Thalha Jubair were convicted for their roles in the attack, which led to large costs for Transport for London.

Failures, asynchrony and all kinds of effects can be represented by continuations

An experimental role-reversal chatbot: you play the AI and answer users’ prompts.

Reinforcement learning with verifiable rewards without human-annotated data, often referred to as zero RL, has emerged as a powerful paradigm for eliciting chain-of-thought reasoning. However, due to computational constraints, existing studies are largely restricted to small models, leaving the training dynamics and emergent capabilities at a large scale unexplored. To meaningfully explore this frontier, we aim to elicit high-quality reasoning behaviors from the model. However, we find that naive scaling often suffers from poor readability, token redundancy, and a lack of adaptive reasoning depth. To address these challenges, we present a stable and efficient training pipeline, incorporating algorithmic and system optimizations such as clipped importance sampling, training-inference ratio correction, and mixed-precision control. Our experiments offer three key findings that validate the "bitter lesson" of scaling: (1) scaling to 1T parameters significantly enhances sample efficiency and performance ceilings; (2) the training process progresses sequentially through an initial discovery phase followed by a sharpening phase; and (3) the model spontaneously develops advanced cognitive behaviors, including anthropomorphism, structured formatting, self-verification, parallel reasoning, and context anxiety, rendering hand-crafted heuristics redundant. Evaluated on seven mathematical benchmarks, Ring-2.5-1T-Zero achieves competitive performance. Additionally, to assess CoT quality beyond final-answer correctness, we propose a structured evaluation framework across three dimensions: comprehensibility, reproducibility, and efficiency, where our model demonstrates clear advantages in producing structured and concise reasoning traces. By sharing our observed emergent phenomena, we hope to provide the community with deeper insights into scaling behaviors, particularly at the 1-trillion scale.

Hi. Lo. Solve. Guess the 7-letter word in 5 tries.