The latest blog posts to take off on Hacker News. 2,240 posts.
A storage appliance that shipped before AVX2 existed now runs Google's Gemma 4 26B in my basement, no GPU. Here's how we got a modern MoE model onto a CPU older than the instructions its inference engine assumed were there.
This is a quick update about the status of Briar. Short version: the project is still active but we’re only making essential security updates and bugfixes for now. Long version: For several years we’ve been trying to find solutions to some of the longstanding issues with Briar, such as high battery usage and…
/prewalk hands the cheap model a trajectory, not a plan, retaining 92–97% of frontier pass rate at roughly half the cost on SWE-Bench Pro.
Dependabot now waits until a new release has been available on its registry for at least three days before opening a version update pull request. This cooldown is now the…
The first long-term support release in 5.x series. Supported until July 2028.
A seven-month-old arbitrary code execution vulnerability remains present in Cursor, leaving 1 million daily users to rely on public disclosure after months of unsuccessful attempts to obtain remediation updates.
Today we're announcing Bonsai 27B, our multimodal flagship: ternary at 5.9GB for laptops, 1-bit at 3.9GB for an iPhone 17 Pro, with a 262K-token context.
Raffael Hickisch spent years dreaming up conservation tools he couldn't afford to build. Then he built them himself, with an AI agent and a country's worth of public data.
AI built for autonomy crowds people out, making us passive observers of what’s coming. We’re building toward a different future.
The Solar Sharer Offer launches 1 July 2026. Here's what changed, how the cap works, and how to maximise free daytime electricity for your home.
Real token cost of frontier models on the same TypeScript input - Claude Fable 5.1, Opus 5, GPT-5.6, Gemini 3.7 Flash. Tokenizer counts, effective cost.
Compare Apple's new SpeechAnalyzer API to Whisper and SFSpeechRecognizer. Real accuracy benchmarks on 5,559 utterances with raw transcripts for rescoring.
Precursor, our new continuous behavioral validation engine for bot management, offers visibility into how humans and bots actually interact across the full user journey. By turning session-level behavior into bot detection signals, it identifies advanced automation with higher precision — while reducing friction for…
Math.tanh, every CSS trig function, and the Web Audio compressor all route through the host libm, so the rounding of a cosine betrays the OS a browser actually …
Claude Code vs OpenCode token overhead measured at the API boundary. Out-of-the-box baselines, instruction file weight, MCP schema tax, subagent multipliers, and cache-write behaviour.
For four months, no frontier model beat Claude Opus in our production evals. GPT-5.6 did. This is what we learned while migrating.
Virginia Tech scientists at the Fralin Biomedical Research Institute show that resistance training outperforms endurance exercise in improving insulin sensitivity in obesity and Type 2 diabetes models.
How Mesh LLM pools existing GPU resources across machines into a single OpenAI-compatible API, built on iroh.
Neoclouds are one of the more hotly debated AI business models, with CoreWeave and Nebius being the two most widely recognized names. These companies have seen their sales, backlog, and share prices soar. Yet, supporting their growth is extremely expensive, and neoclouds do not have the same cash nor operating cash…
PgBouncer is single-threaded, so a single process caps out at one CPU core no matter the box size. See how ClickHouse Managed Postgres runs a peered fleet of PgBouncer processes with so_reuseport to scale pooling across every core — with benchmarks showin
Researchers have shown the first direct experimental evidence that the textbook triple bond structure breaks down in heavy elements, where relativity makes the rules.
GPT-5.6's new Sol, Terra, and Luna tiers go head-to-head with Grok 4.5, Claude, Meta's Muse Spark, and the open-weights crew on a raycaster, a Rubik's cube, a calculator, and Game of Life. Here's every build, with cost and latency.
Open source software is the foundation of much of our digital infrastructure. However, is AI undermining the incentive structure? What needs to change?
The Cambridge Programme on AI Science & Policy (CASP) is an interdisciplinary research programme on frontier AI at the University of Cambridge.
One of the best things about LLMs is that they'll write code for you, all day long. Who cares about DRY? You don't have to be the one updating the same long con...
We set out to build the first AI tutor to teach math and reading to kids ages 4-9. For AI to actually teach a five-year-old, pedagogy must be baked into the engineering. A child can't wait for a slow reply, can't read a chat interface, and can't unhear anything a model gets wrong. We wanted to share some of the…
A real quarter of UK SME books, 59 transactions, 68 minutes, $2.73 of tokens — and a VAT return within 7 pence of the human-prepared one.
