1741
Tau² benchmark: How a prompt rewrite boosted GPT-5-mini by 22%
We expected small models to be fast, but our benchmarks revealed a common reliability trap. Here’s our deep dive on finding and fixing it.
Top posts
Every blog post that took off on Hacker News, ranked by the votes of its best submission. 3,240 posts.
1742
Facebook to open-source AI hardware design
Visit the post for more.
1743
Popping the GPU Bubble
Photon, Moondream's inference engine, achieves near-realtime VLM inference (~33ms on NVIDIA B200). This is a peek into how it delivers up to 35% higher decode throughput by optimizing how the GPU works.
1744
Ollama is now available as an official Docker image
Ollama can now run with Docker Desktop on the Mac, and run inside Docker containers with GPU acceleration on Linux.
1745
Open-Source Speech Recognition: Simon 0.4.0 Released
After years of hard work, the Simon team is proud to announce the new major release: Simon 0.4.0. New in Simon 0.4 This new version of ...
1746
Researchers find evidence of ChatGPT buzzwords turning up in everyday speech
Within five days of ChatGPT’s release in 2022, the artificial intelligence chatbot gained more than a million users. Today, more than half of all adults
1747
1748
AI hallucinations: Why LLMs make things up (and how to fix it)
Nine Technical Strategies for Reducing Hallucination
1749
Context is the bottleneck for coding agents now
Ship production-quality code at superhuman speed.
1750
Fable 5 On Vending-Bench: Misbehaving, With Plausible Deniability
Claude Fable 5 is a partial step back in alignment from Opus 4.8 on Vending-Bench, returning to price collusion and deceptive negotiation while rationalizing misbehavior it knows is wrong.
1751
Quasar 438B: Europe's Leading AI Model
Quasar sets a new benchmark for European AI, outperforming comparable European models on seven of eight selected Artificial Analysis evaluations. ...
1752
Remote MCP Support in Claude Code
Connect your favorite development tools and data sources securely to Claude Code through remote MCP servers. Today, we’re announcing support for remote MCP servers in Claude Code.
1753
1754
Video Stabilization on YouTube
Posted by Matthias Grundmann, Vivek Kwatra, and Irfan Essa, Research at Google One thing we have been working on within Research at Google is deve...
1755
Historical weather data API for machine learning, free for non-commercial
Get weather data for the past days, weeks or months
1756
1757
Beyond automatic differentiation
Posted by Matthew Streeter, Software Engineer, Google Research Derivatives play a central role in optimization and machine learning. By locally app...
1758
Bad Actors Are Grooming LLMs to Produce Falsehoods
Our research shows that even the latest "reasoning" models are vulnerable
1759
Turn ideas into music with MusicLM
Starting today, you can sign up to try MusicLM, a new experimental AI tool that can turn your text descriptions into music.
1760
GPT-5o-mini hallucinates medical residency applicant grades
Since the opening of the residency recruitment season, we’ve heard your feedback on the Cortex Core Clerkship Transcript Normalization feature. As a physician-founded organization born out of our lived experience as applicants and program leaders, we understand the many emotions tied to the start of the season. We…
1761
Turn your best AI prompts into one-click tools in Chrome
Skills in Chrome let you discover, save and remix AI workflows — and repeat them instantly.
1762
LlamaCloud and LlamaParse
LlamaIndex introduces LlamaParse and LlamaParse for production-grade RAG. Unlock complex PDF data to improve LLM response quality. Discover how today.
1763
OpenAI Five: Goals and Progress
OpenAI Five is the first AI to beat the world champions in an esports game, having won two back-to-back games versus the world champion Dota 2 team, OG, at Finals this weekend. Both OpenAI Five and DeepMind’s AlphaStar had previously beaten good pros privately but lost their live pro matches, making this also the…
1764
1765
What's in a GGUF, besides the weights – and what's still missing?
What extra stuff is needed to properly run a language model? Besides the weights of a language model, what is the gguf metadata that we need to parse and use?
1766
1767
TensorFlow Fold: Deep Learning with Dynamic Computation Graphs
Posted by Moshe Looks, Marcello Herreshoff and DeLesley Hutchins, Software EngineersIn much of machine learning, data used for training and inferen...
1768
WARC-GPT: An open-source tool for exploring web archives using AI
The Library Innovation Lab is growing knowledge and community by bringing library principles to technological frontiers.
1769
Google open sourced GPipe, a library for training large ML models
Posted by Yanping Huang, Software Engineer, Google AI Deep neural networks (DNNs) have advanced many machine learning tasks, including speech rec...
1770
A Year of Telepathy
Over the past year, three people with paralysis have received Neuralink implants. This blog post explores how each person is using Telepathy in everyday life.