Apple M5 isn’t making full use of its matmul cores yet
At the moment MLX (and Llama.cpp for Macs) run 16bit activations everywhere. Despite this, the M5 generation silicon actually does support INT8 activations - it actually…
Latest
PSA on Laguna S-2.1 – Use the updated chat template and GGUF
Link to their official GGUF repo: https://huggingface.co/poolside/Laguna-S-2.1-GGUF/tree/main All the GGUFs received this fix 5ish hours ago - correct yarn_attn_factor to 1.0 (llama.cpp derives mscale) And the…
Duane Arnold
“So maybe I should enlighten you on what happens in your absence. This selfish existence where this introvert turns extrovert and dons her social armour.” Some…
NeurIPS Reviews out for me [R]
I know it might be because we have an earlier number (in the 600's) but we got all 4 of our reviews. Even the harshest ones…
RT ℏεsam: it’s just for your safety
RT ℏεsamit’s just for your safetyPolymarket: JUST IN: Anthropic donates another $20 million to a political nonprofit demanding stricter AI regulation ahead of the midterms.
Debugging Ray Tracing Applications Using NVIDIA OptiX Toolkit
NVIDIA OptiX ray tracing engine is an application framework for achieving optimal ray tracing performance on the GPU. Applications using OptiX can fail in ways...
Google just had its first negative cash flow quarter ever due to massive AI spending
Google continues to report big quarterly revenue, but its AI spending has skyrocketed.
JEP 540: Simple JSON API (Now in Incubator)
Article URL: https://openjdk.org/jeps/540 Comments URL: https://news.ycombinator.com/item?id=49023809 Points: 22 # Comments: 5
Start Customizing NVIDIA Nemotron 3 Nano with Prime Intellect Lab in Minutes
Customization is what enables developers to take a general model and tailor it to use cases, domains, languages, and more. However, customization comes with a...
I trained a 0.5M model on 1B tokens of Fineweb-edu dataset.
Hi everyone, About a month ago I publish my very first research paper on my neural network architecture called Silia. You can look at the model…
I distilled an 8B teacher into a 0.6B student on my Mac (MLX). The 0.6B went from 36% to 100% on the task, but few-shot prompting actively made it worse.
I work with financial documents for my job, and I wanted a small model I could run locally to enrich them before they hit a RAG…
Free way to get your data out of ChatGPT Business accounts
Article URL: https://github.com/Conradqh/scrapemychats Comments URL: https://news.ycombinator.com/item?id=49023486 Points: 8 # Comments: 2
RT Google Gemma: Gemma 4 just crossed 300 million downloads. Thank you to the developers, researchers, and open-source community building with us. You…
RT Google GemmaGemma 4 just crossed 300 million downloads.Thank you to the developers, researchers, and open-source community building with us. Your work and feedback drive this…
Show HN: OneCLI – OSS credential gateway that keeps secrets out of AI agents
hey HN, Jonathan and Guy here, creators of OneCLI (https://onecli.sh/). OneCLI is an open source vault for AI Agents.Traditional vaults are used to store your secrets…
ChatGPT Sites means ChatGPT in "Work" mode can build and deploy public websites running on Cloudflare Workers, including with persistence on top of SQ…
ChatGPT Sites means ChatGPT in "Work" mode can build and deploy public websites running on Cloudflare Workers, including with persistence on top of SQLite(OpenAI do not…
RedNote has a frontier AI program
RedNote has a frontier AI programFlorian Brand: dots, the research division from xhs (Chinese Instagram), claims to have reached a perfect score on this years IMO
UK households free to install plug-in balcony solar panels from end of August
Article URL: https://www.theguardian.com/money/2026/jul/23/uk-households-free-install-plug-in-balcony-solar-panels Comments URL: https://news.ycombinator.com/item?id=49023291 Points: 11 # Comments: 2
Why A Frontier Data Agent Outperforms General Coding Agents in Quality and Cost
Conventional wisdom in agentic AI says that better answers cost more tokens: let...
RT Enrique H. Calabrés: 🚀 New tutorial just dropped! Deploy FLUX.2 on Amazon SageMaker AI with the SageMaker Python SDK This step-by-step guide wi…
RT Enrique H. Calabrés🚀 New tutorial just dropped!Deploy FLUX.2 on Amazon SageMaker AI with the SageMaker Python SDKThis step-by-step guide will show you how to:🎨 Deploy…
Agent automation controls in GitHub Issues in public preview
Agent automations increasingly label, type, assign, and close issues for you. GitHub Issues now shows the reason behind each change and lets you review them before…
Connect Amazon S3 data to Databricks with Delegated IAM Permissions
Importance of Simpler S3 connectivityConnecting Amazon S3 is one of the most important...
Introducing Impact List: a ranking of people by the expected value of their donations
TL;DR: I'm releasing a website that ranks people according to the expected value that they've produced via donations, and allows users to re-rank the list using…
The cool thing about Chinese is that they can be infinitely pragmatic and cynical, but they do have a pretty strong notion of virtue, and it activates…
The cool thing about Chinese is that they can be infinitely pragmatic and cynical, but they do have a pretty strong notion of virtue, and it…
Why Software Factories Fail (or: harness engineering is not enough)
Article URL: https://github.com/humanlayer/advanced-context-engineering-for-coding-agents/blob/main/wsff.md Comments URL: https://news.ycombinator.com/item?id=49023019 Points: 24 # Comments: 4