AI News Today

Howdy all. I'm Zack :wave:. I've been thinking about the problem of misguided AI pull requests and figured I'd throw a possible solution out there for feedback. Basically, CleverCrow lets supporters give tokens to a GitHub repo (or set of issues in that repo) for the maintainers to use to build/fix stuff. The fun implementation challenges have been around implementing the pooling dynamics and keeping the maintainers in charge while the backers are motivated to support their work. Comments URL: h

By AI I mean Image Generation models like stable difussion, but I had to make a short title. They look like Video Games sometimes too. This reminds me of that case of a scanner used in an architecture studio that was changing numbers for others, and it turns out they were using content aware compression and decompression. Maybe the commercial pressure to use AI caused them to apply an AI filter to upscale the images and they end up looking like videogames? To be precise I'm looking at the Belgiu

This is my markdown viewer for the terminal. It's most notable feature is that it displays headers Bigger™ then other text. This was my main motivation for writing this, as big headers make markdown really more readable and was the only thing making me reach outside of a terminal for reading markdown documents. I did this by leveraging ratatui-image, which I also had written for another thing. MdFried tries to get your terminal's font, or otherwise prompts you with a font picker. The headers are

Duno what Id call it exactly? A boggle-like smushed together with crossword style riddles and some board re-roll mechanics. Bit of an LLM experiment in spinning off daily, theme-based word games easy enough to bang out over breakfast or a coffee break. Anyway, fire it up in a mobile browser and have a poke at it - appreciate any thoughts as I keep tinkering with it. No account or login required. Comments URL: https://news.ycombinator.com/item?id=48621571 Points: 1 # Comments: 0

Article URL: https://slashdot.org/story/26/06/20/0556251/tech-pundit-cringely-co-founds-startup-2brains-inc-to-solve-llm-hallucinations Comments URL: https://news.ycombinator.com/item?id=48621381 Points: 2 # Comments: 0

Claude Guillemot, who founded Ubisoft with his four brothers, has died at the age of 69.

<!-- SC_OFF --><div class="md"><p>We just did a big revamp of <strong>WeightsLab</strong> and wanted to share it here.<br/> If you’ve ever spent hours debugging a training run only to discover it was a data problem all along, this is for you.<br/> WeightsLab lets you pause training mid-run, inspect your live loss signals, and catch mislabels, class imbalance & outliers before they tank your model.</p> <p>Open source, PyTorch-native, built for CV engineers working with images, videos & LiDAR poin

Article URL: https://staffengineer.dev/ Comments URL: https://news.ycombinator.com/item?id=48620903 Points: 2 # Comments: 0

Article URL: https://vexyn.app/ Comments URL: https://news.ycombinator.com/item?id=48620808 Points: 1 # Comments: 0

Article URL: https://tenureai.dev/writing/llm-as-judge-became-the-default-for-agent-evaluation/ Comments URL: https://news.ycombinator.com/item?id=48620731 Points: 5 # Comments: 0

Understanding ow LLMs interact with the world around them, from returning data to taking action The post Tool Calling, Explained: How AI Agents Decide What to Do Next appeared first on Towards Data Science .

Amazon will no longer release "Artificial," which stars Andrew Garfield as OpenAI CEO Sam Altman.

<!-- SC_OFF --><div class="md"><p>Hi guys</p> <p>Does anyone know of papers where EMA on LoRA adapters has been used successfully?</p> <p>Im interested in cases where the EMA adapter acts as a self-teacher generating soft labels for the trainable adapter.</p> <p>On-policy self-distillation [1] uses ema for the teacher. However, they seem to fully fine-tune. Any empirical results showing the idea is working on lora/ left models?</p> <p>[1] <a href="https://arxiv.org/abs/2601.19897">https://arxiv.

Article URL: https://www.youtube.com/watch?v=WNiSUGCC-gY Comments URL: https://news.ycombinator.com/item?id=48620402 Points: 20 # Comments: 7

<p><em>Hello, I'm Shrijith Venkatramana. I'm building git-lrc, an AI code reviewer that runs on every commit. <a href="https://github.com/HexmosTech/git-lrc" rel="noopener noreferrer">Star Us</a> to help devs discover the project. Do give it a try and share your feedback for improving the product.</em></p> <p>Machine learning engineers spend countless hours optimizing models, tweaking architectures, and squeezing performance out of hardware.</p> <p>Yet many developers who train large models toda

Well, calling it a browser is a bit of a stretch I suppose. It's really just chromium + playwright with a terminal screen bolted on. But damn does this thing actually work. I've tried every other kind of browser use agent and they all try to get the model to use websites by taking screenshots and moving the mouse around. Which is slow and inefficient and expensive too. By contrast I've used this browser to do things like scrape thousands of pictures or automatically configure fly.io billing for

Article URL: https://zither-zeta.vercel.app/ Comments URL: https://news.ycombinator.com/item?id=48620328 Points: 1 # Comments: 0

Article URL: https://bartoszlenart.com/blog/bonfires-in-the-dark Comments URL: https://news.ycombinator.com/item?id=48620314 Points: 3 # Comments: 0

Article URL: https://github.com/TheCulliganMan/crystal-llm/ Comments URL: https://news.ycombinator.com/item?id=48620210 Points: 1 # Comments: 0

I write Abaqus UMAT subroutines as a graduate student in computational mechanics. These are complex multi-physics simulation models using Abaqus Fortran subroutines (UMAT ~ mechanical behavior, UMATHT ~ heat/diffusion and a lot more) that simulate how different material systems fail under high temperature or manufacturing processes. However, the entire process has been quite challenging, 80-90% of time is actually spent on how make Abaqus CAE simulate a physics you already know on paper or what

Article URL: https://akarouter.dev Comments URL: https://news.ycombinator.com/item?id=48620183 Points: 2 # Comments: 1

Article URL: https://electrek.co/2026/06/21/tesla-megapod-ai-data-center-hardware/ Comments URL: https://news.ycombinator.com/item?id=48620163 Points: 8 # Comments: 0

Comments

<pre style='white-space:pre-wrap;width:81ex'>fix(gateway): preserve owner MCP tools for agent RPC</pre>

Article URL: https://atizar.io Comments URL: https://news.ycombinator.com/item?id=48620103 Points: 2 # Comments: 0

Article URL: https://portal.neuralwatt.com/ Comments URL: https://news.ycombinator.com/item?id=48620100 Points: 2 # Comments: 0

Article URL: https://github.com/GoogleCloudPlatform/knowledge-catalog/tree/main/okf Comments URL: https://news.ycombinator.com/item?id=48620068 Points: 2 # Comments: 2

Article URL: https://github.com/usefulsoftworks/ai-agent-control-checklist Comments URL: https://news.ycombinator.com/item?id=48620058 Points: 2 # Comments: 0

Welcome back to TechCrunch Mobility — your central hub for news and insights on the future of transportation.

Article URL: https://www.askmaps.ai Comments URL: https://news.ycombinator.com/item?id=48620028 Points: 3 # Comments: 3

Hello HN! I built a pre-commit code scanner that checks your staged changes for security vulnerabilities every time you run 'git commit'. I am an inexperienced programmer and this is my first personal project. Any feedback, positive or negative, big or small, relevant or not, would be greatly helpful and appreciated! The tool requires Gitleaks and Python to run, you will also need to use your own AI API key, which I understand is a big setup friction. You can check out my demo video instead if t

I'm Dustin Kirkland, author of Byobu[1]. I built Trustmux[2] to solve a personal problem: secure, encrypted access to my tmux[3] and Byobu sessions from my phone, without the pain of other mobile shell approaches. I run long-lived shells, Claude, and other sessions in the background on my workstation, and I needed to check in from mobile — securely, with real TLS certificates and Tailscale-backed privacy, not some insecure workaround. Why not SSH / Mosh / Claude Mobile? - SSH/Mosh: Synchronous c

On the new episode of Equity, we discussed what actually prompted the administration's latest moves against Anthropic, and what this might mean for the AI ecosystem.

Article URL: https://smartdino.dev/features Comments URL: https://news.ycombinator.com/item?id=48619681 Points: 1 # Comments: 0

Article URL: https://agent-departures.vercel.app/ Comments URL: https://news.ycombinator.com/item?id=48619665 Points: 1 # Comments: 0

I was writing a blog when this hit me and couldn't find too many convincing arguments other than incentivizing heavier use, would love to hear takes on this. Closest parallel I could think of is gift cards, which too do not expire in many places. Comments URL: https://news.ycombinator.com/item?id=48619635 Points: 4 # Comments: 5

Article URL: https://www.ai-convert.xyz Comments URL: https://news.ycombinator.com/item?id=48619605 Points: 1 # Comments: 0

Article URL: https://brooker.co.za/blog/2026/06/18/my-blog-and-ai.html Comments URL: https://news.ycombinator.com/item?id=48619544 Points: 2 # Comments: 0

Enterprise Document Intelligence [Vol.1 #5septies] - When a PDF prints a contents page but exposes no outline, two ways to turn it back into structure, plus the page-alignment step everyone forgets The post Reconstructing the Table of Contents a PDF Forgot to Ship, So RAG Can Scope by Section appeared first on Towards Data Science .

Siri’s AI overhaul may have grabbed the headlines at WWDC, but some of Apple’s most useful AI features are arriving elsewhere in iOS 27.