AI News Today

Hi everyone. We're AI researchers at Harvard and Carnegie Mellon working on a project to advance the state of agent security. Currently, many systems rely on static sandboxing, which in long-running sessions enables agents to understand the safeguards holding them in place and break out of them. We've found vulnerabilities across over a dozen agent providers and frameworks (practically every one we tested) displaying this behavior (eg. a model fraudulently splitting payments to avoid a company-s

Article URL: https://mtok.market/ Comments URL: https://news.ycombinator.com/item?id=48837658 Points: 1 # Comments: 0

Explore how the Aspire team turns merged product changes into SME-reviewed docs pull requests, closing the gap between release and documentation. The post Automating cross-repo documentation with GitHub Agentic Workflows appeared first on The GitHub Blog .

At the Open Source Summit in Mumbai, Torvalds discusses the pain and power of AI in the kernel, and why Linux no longer supports 'museum' technology.

Lionel Messi and Cristiano Ronaldo are betting on AI, health tech, and startups. Mohamed Salah is taking a more traditional route beyond football.

Article URL: https://openai.com/index/separating-signal-from-noise-coding-evaluations/ Comments URL: https://news.ycombinator.com/item?id=48837396 Points: 202 # Comments: 71

Article URL: https://app.shevtoneaudio.com/ Comments URL: https://news.ycombinator.com/item?id=48837380 Points: 1 # Comments: 1

Netflix engineers detailed how they handle wide partitions in Apache Cassandra for the TimeSeries Abstraction. Two approaches work together: Time Slice re-partitioning tunes future partitions at the table level, while dynamic partitioning detects and splits oversized partitions per TimeSeries ID on the read path. Detection runs via byte counting and Kafka, splits are checksum-validated, and Bloom filters route reads to parallel child partitions. Average read latency dropped from seconds to low d

Treasure in the trash. ScienceAlert stories are written, fact-checked, and edited by humans, never generated by AI. Don't miss a story, subscribe here.

You would expect only people who have a very high opinion of themselves to not feel the need to use AI. Comments URL: https://news.ycombinator.com/item?id=48837332 Points: 5 # Comments: 7

Earlier this week, a picture seemed to show Kentucky Senator Mitch McConnell covered in tubes in a hospital bed in a state of extreme distress. It turned out to be an AI-generated fake.

Article URL: https://www.washingtonpost.com/technology/2026/07/07/most-nurses-say-ai-isnt-good-enough-trust-with-patient-care-survey/ Comments URL: https://news.ycombinator.com/item?id=48837096 Points: 7 # Comments: 0

Like its U.S. counterpart, the European Chips Act aims to foster the semiconductor industry — in part thanks to state subsidies. One of the beneficiaries is QuantumDiamonds, a German startup that applies a novel approach to inspecting chips.

As someone who doesn't use any of the more "advanced" features on sites like ChatGPT or Claude (agentic mode, memory, image generation, deep research, etc), I found the bloat of these services, both in UX and in performance, to be pretty tedious. There are a million AI chat interfaces out there, but I could never find one that just did simple messaging and history in the most minimal and lightweight way possible. So I made my own. Nully is written in Go and Vanilla HTML/CSS/JS. It literally just

Article URL: https://confighub.com/blog/gitops-in-the-age-of-ai-c73d0bd506ba Comments URL: https://news.ycombinator.com/item?id=48836971 Points: 3 # Comments: 1

Article URL: https://danluu.com/ai-coding/#llm-variance Comments URL: https://news.ycombinator.com/item?id=48836951 Points: 11 # Comments: 1

Third-party testing shows heterogeneous compute platform combining H200s and SN50 RDUs churning out 763 tok/s in MiniMax M2.7

On May 23, 2026, fresh off the stage at Google I/O, our Google Developer Experts (GDEs) converged on...

Article URL: https://www.theatlantic.com/international/2026/07/xi-jinping-censorship-ai-training/687696/ Comments URL: https://news.ycombinator.com/item?id=48836825 Points: 4 # Comments: 3

Experiments in using AI to build AI show that the future doesn’t just belong to the frontier labs.

Manna is launching a U.S. operations and manufacturing facility in Tulsa, Oklahoma that will eventually employ 1,000 people.

<blockquote cite="https://twitter.com/kentonvarda/status/2074924213983740233"><p>I just declared a moratorium against AI-written change descriptions (e.g. PR and commit messages, also issues/tickets) from my team.</p> <p>AI was writing change descriptions that were worse than useless to me as I tried to review PRs: outlining details of the code that could easily be seen by looking at the code, but omitting the higher-level framing needed to understand broadly what the code is doing.</p></blockqu

<p>In the world of Deep Learning, there is a fundamental tension that keeps researchers and mobile developers awake at night. On one side, you have the mathematical idealism of high-precision deep learning models, born in the realm of <code>float32</code> (32-bit floating point). On the other, you have the brutal physical reality of mobile hardware: limited RAM, finite battery life, and the need for instantaneous inference.</p> <p>If you try to deploy a massive, high-precision model directly to

Article URL: https://dibi8.com Comments URL: https://news.ycombinator.com/item?id=48836687 Points: 2 # Comments: 0

As GitHub struggles to manage AI load, challengers take aim

<p>Same outputs, same judge, two runs, two scores. The gate flickered red then green on a branch with zero code changes, and that flapping cost me more trust than any real regression.</p> <h2> The flap </h2> <p>I had a faithfulness gate on merge: judge scores every case, the mean has to clear 0.80. One Tuesday it failed at 0.79. I re-ran the identical job, no code change, no prompt change, and it passed at 0.82. Ran it a third time: 0.80 exactly. Nothing in the repo had moved. The judge was disa

<h2> Hook </h2> <p>A public GitHub issue, a hidden instruction, and one word changed in a prompt was enough to get an AI agent to leak private repo data. No stolen credentials required. If that doesn't make you nervous about what you've plugged into your CI pipeline, it should.</p> <h2> Context </h2> <p>This isn't a new category of bug — it's the oldest bug in the book wearing a new costume. Confused deputy problems have existed as long as we've had systems that act on behalf of users with eleva

Today, we're announcing the Claude apps gateway for AWS, a self-hosted control plane that gives organizations a single point of control over access, cost, and policy for Claude Code and Claude Desktop. In this post, we show how to set up and run Claude apps gateway for AWS with Amazon Bedrock and Claude Platform on AWS.

AI-generated football memes are taking over the 2026 World Cup, with Haaland, Mbappé, Ronaldo, Messi, and more becoming fan-edit material

Article URL: https://runtimewire.com/article/startup-spotlight-sourcerer-wants-to-turn-ai-procurement-into-a-supplier-not-saa Comments URL: https://news.ycombinator.com/item?id=48836477 Points: 2 # Comments: 0

Article URL: https://www.americanbanker.com/news/ai-is-starting-to-replace-humans-in-operations-analyst-jobs Comments URL: https://news.ycombinator.com/item?id=48836460 Points: 2 # Comments: 0

Article URL: https://blog.larah.me/test-slop/ Comments URL: https://news.ycombinator.com/item?id=48836454 Points: 2 # Comments: 0

Article URL: https://www.dwarkesh.com/p/blog-prize-winners Comments URL: https://news.ycombinator.com/item?id=48836384 Points: 2 # Comments: 0

Elon Musk's tech company released the newest version of Grok on Wednesday, promising a cheaper, more efficient alternative to other powerful AI models.

Article URL: https://github.com/column-tax/tax-calc-bench Comments URL: https://news.ycombinator.com/item?id=48836352 Points: 2 # Comments: 0

What, did someone get some bad news during their IPO process or something?

TLDR: Do you have a pattern or structure for doing long-lived research or non-coding projects using LLMs? I was kind of inspired by this hacker news article where someone collected data to solve their fatigue problem: https://news.ycombinator.com/item?id=48605117 A similar but smaller I did was to collect nut-free restaurants to eat in Chicago. So I had articles, reviews, emails, phone calls, all sorts of stuff. I found it very easy to just use Copilot as my interface and I just data into it ove

More fun with AI jailbreaks, this time at the workflow level

General Intuition is betting millions of hours of video game data can train the foundation models for physical AI, making it easier to build smarter robots with minimal real-world data.

Article URL: https://techstackups.com/articles/ai-model-naming-is-a-mess/ Comments URL: https://news.ycombinator.com/item?id=48836200 Points: 3 # Comments: 0