1261
Running Kimi K3 on MI355X at Better Performance per Dollar Than B300
Running Kimi K3 at ~952 tok/s/node, AMD continues to prove its case as the winner in performance per dollar.
Top posts
Every blog post that took off on Hacker News, ranked by the votes of its best submission. 2,243 posts.
1262
Dulce et Decorum Est (1921)
Bent double, like old beggars under sacks, Knock-kneed, coughing like hags, we cursed through sludge, Till on the haunting flares we turned our backs, And…
1263
ClickHouse acquires Langfuse
Our goal continues to be building the best AI engineering platform
1264
Choosing an AI model: one prompt, 11 models, different results
Netlify now runs any OpenRouter model, including Kimi K3, GLM 5.2 and DeepSeek V4. We tested 11 of them on the same build prompt to see how they differ.
1265
Photon-Emission-Guided Laser Fault Injection Enables RP2350 Secure Debug
Differential photon-emission microscopy localized debug enable register activity and narrowed the laser search before SWD-guided injection set the two bits required to restore Secure debug on an RP2350 A4.
1266
Upper stage impacting the moon on 2026 August 5
On 2026 August 5, within a few minutes of 06:35 UTC, an upper stage (section) of a rocket used for a lunar mission will hit the moon. This may be of some (probably minor) scientific interest, and we may learn some things from it. It doesn't present any danger to anyone, though it does highlight a certain carelessness…
1267
Looking Ahead to Postgres 19
Discover the top PostgreSQL 19 features, including built-in REPACK CONCURRENTLY, SQL property graph queries (SQL/PGQ), and logical replication updates.
1268
Five months treating bugs like patients and coding agents like a medical team
How Cockroach Labs built MOLT Sinai, an AI coding “teaching hospital” that shipped IBM Db2 migration support in under two days.
1269
Real-time LLM Inference on Standard GPUs: 3k tokens/s per request
Today, Kog AI launches a tech preview of the Kog Inference Engine (KIE): 3,000 output tokens/s per request on 8× AMD MI300X GPUs and 2,100 on 8× NVIDIA H200 (FP16, no speculative decoding). This preview runs a 2B model, with support for large third-party MoE models coming next at similar speeds.
1270
Consistency diffusion language models: Up to 14x faster, no quality loss
Standard diffusion language models can't use KV caching and need too many refinement steps to be practical. CDLM fixes both with a post-training recipe that enables exact block-wise KV caching and trajectory-consistent step reduction — delivering up to 14.5x latency improvements
1271
Gemini Robotics-ER 1.6
Gemini Robotics ER 1.6 upgrades spatial reasoning and multi-view understanding, unlocking new capabilities like instrument reading for autonomous robots.
1272
Postmortem: Our first VLEO satellite mission (with imagery and flight data)
On March 14, 2025, Albedo's first satellite, Clarity-1, launched on SpaceX Transporter-13. We took a big swing with our pathfinder. The mission goals:
1273
Introducing Toast 1
Meet Toast 1, Mixedbread's search agent for knowledge-intensive tasks, matching or outperforming Claude Opus 5 and GPT-5.6 Sol while being up to 10×…
1274
We want Texans to know their rights
Last May, EFF reported that a sheriff’s office in Texas searched data from more than 83,000 automated license plate reader (ALPR) cameras to track down a woman suspected of self-managing an abortion. ALPRs are promoted as tools for keeping communities safe by finding missing persons and locating...
1275
Jank now has its own custom IR
The jank compiler has a new super power. With that, we push jank to the limits to see who has the fastest fibonacci in the west.
1276
The classifiers Anthropic puts in front of Fable are too zealous
COMBINE-lab develops algorithms, data structures, and software for high-throughput genomics.
1277
New copy of earliest poem in English, written 1,3k years ago, discovered in Rome
Old fashioned sleuthing and the help of modern technology leads to discovery of manuscript with poem composed by a farm labourer 1,300 years ago.
1278
Understanding the recent DDoS attack against Read the Docs
A deep dive into the massive June 2026 DDoS attack on Read the Docs, the attack patterns we observed, and the mitigations and defenses we used to defend our infrastructure.
1279
Artifacts: Versioned storage that speaks Git
Give your agents, developers, and automations a home for code and data. We’ve just launched Artifacts: Git-compatible versioned storage built for agents. Create tens of millions of repos, fork from any remote, and hand off a URL to any Git client.
1280
Training our own AI models
I really think we're on the verge of some of our best work through the next six months. Over the past year, we've started building more AI-powered…
1281
Why are there both TMP and TEMP environment variables? (2015)
Parallel evolution.
1282
Ornith-1.5: From Self-Scaffolding to Self-Improvement
Introducing Ornith-1.5, an open-source model family that extends self-scaffolding into an end-to-end self-improvement loop.
1283
We moved Railway's frontend off Next.js. Builds went from 10+ mins to under 2
We migrated http://railway.com, the dashboard, and the canvas from Next.js to Vite + TanStack Router in two PRs. Here's why, how, and what we learned.
1284
LLM Usage in Debian: Three Proposals
This proposal aims to expressly forbid any contributions to Debian written with the use or assistance of large language models (LLMs) or other generative AI tools.
1285
LibreOffice and the art of overreacting
A donation banner is not an attack to users The announcement that LibreOffice 26.8 will feature a donation banner in the Start Centre has prompted a flood of responses, ranging from positive from many FOSS supporters, who understand the need for funding, to mild apprehension to extreme alarm from others. Some articles…
1286
Instant 1.0, a backend for AI-coded apps
Instant 1.0 is out! This essay shows a bunch of demos, to explain why we think Instant is the best backend for AI-coded apps. We also cover the architecture that makes all of it work.
1287
Formatting a 25M-line codebase overnight
This is the story of how our Developer Productivity team extended and rolled out rubyfmt, our Rust-based zero-config, ultra-fast autoformatter across the world's largest Ruby codebase.
1288
Triton: DirectX 11 Driver for QEMU
In the prequel, we introduced Neptune, a Direct3D protocol forwarding layer for VirtIO. Neptune allowed us to serialize Direct3D API calls across the hypervisor boundary and this allowed us to run Wine games on a Linux guest with a Linux host faster than with DXVK directly in the guest. Admittedly, the payoff there…
1289
Big Tech ruined the cloud, so we're renaming ours
We’ve renamed Home Assistant Cloud to Home Assistant Link. Here’s why the old name never did fit, and why this one finally does.
1290
