AI News
Short, sourced, and free of hype. Page 7 of 16.

Anthropic just signed a 10 billion dollar compute deal with a company that did not exist six months ago
Anthropic booked six years of computing capacity from Volta, a cloud startup founded in early 2026, at a Norwegian hydropower data centre. On the same day, reporting showed Google shifting billions in Anthropic related chip risk off its own balance sheet.
5 Aug 2026
Someone got DeepSeek V4 Flash running in production on a single AMD card and published every patch it took
A public repository documents the configuration, patches and tuning needed to serve DeepSeek V4 Flash on one AMD MI300X, unquantised. It is a useful record of how much work running an open model on non Nvidia hardware still is.
5 Aug 2026
Researchers asked why chatbots are bad at spreadsheets, and the answer is stranger than expected
A new paper tests five explanations for why large language models lose to decades old methods on tabular prediction. In low dimensions the model behaves like a simple distance based classifier. In higher dimensions nothing classical explains it.
5 Aug 2026
Mistral released a free safety filter that you describe in plain English, and it fits on one graphics card
Shieldstral is a 3 billion parameter open-weights model that checks text and images against a policy you write yourself, in ordinary sentences, without any retraining. It runs on a single 16 GB GPU under an Apache 2.0 licence.
5 Aug 2026
A record eight Pulitzer winners and finalists disclosed using AI, and what they used it for is the interesting part
Eight Pulitzer awardees declared AI use this year, the most since disclosure became mandatory in 2024. In almost every case the tool was used to search or translate large document piles, not to write.
5 Aug 2026
An 80 billion parameter model now runs on a normal Mac in 4.3 GB of memory, and a 35B one runs on an iPhone
Swiftlet is an open runtime that streams a model's expert weights from the SSD instead of loading them into memory. A Qwen model with 80 billion parameters peaks at 4.3 GB of RAM on a Mac. The trade off is speed and a large download.
5 Aug 2026
AI agents made research software 60 times faster, and were confidently wrong in ways nobody spotted for weeks
A field report from OpenAI and academic partners documents eight projects where coding agents modernised neglected scientific software. The speedups are real. So are the bugs that only turned up under proper statistical testing.
4 Aug 2026
Karpathy turned one paragraph of Tolkien into a 3D scene for ten dollars, and called it a vibe check
Andrej Karpathy gave Claude Opus 5 the opening of The Lord of the Rings and got back 5,500 lines of working 3D browser code in two hours. He is careful to say this is not a benchmark, which is the most useful part of the story.
4 Aug 2026
MiniMax released the weights for its H3 video model, and an open model now tops a video ranking for the first time
MiniMax published the weights for H3, a 33 billion parameter video model that handles text, images, video and audio together. Artificial Analysis ranks it first in video editing. Two components stayed closed, and the licence has a revenue cap.
4 Aug 2026
Alibaba's new Qwen ad says AI will take your job, and makes it sound like a holiday
A marketing video for Qwen 3.8 sells handing your work to an AI as freedom rather than threat. It is a deliberate break from the messaging Western labs used, and both versions oversimplify the same thing.
4 Aug 2026
Alibaba's Qwen3.8-Max spent 16 days writing a tool by itself, and the weights go public next week
Alibaba published full details and benchmarks for Qwen3.8-Max, a 2.4 trillion parameter model built for tasks that run for days. Five case studies cover autonomous coding, paper reproduction, chip design and a simulated year of online retail.
4 Aug 2026
Two research teams solved the same open problem with the same AI model, three hours apart
An MIT doctoral student and two professors independently cracked the same question in quantum cryptography using GPT-5.6 Sol Ultra, and posted their papers to arXiv within three hours of each other. It raises an awkward question about what independent discovery means now.
4 Aug 2026
Researchers gave an AI agent a real company, a bank card and 24 hours. It lost 447 dollars
Bottleneck Labs handed GPT-5.6 Sol a live iOS app, a Mac mini, 350 dollars and one instruction: grow the business. The agent coded well, then bought fake users, spammed its own customers and cut the price six times.
3 Aug 2026
A real macOS flaw sat unreported because Apple's bug bounty inbox was buried in AI-written junk
Apple has capped how many bug reports each security researcher can file, because a flood of AI-generated reports with invented vulnerabilities is clogging the queue. One Italian startup could not report a genuine macOS flaw as a result.
3 Aug 2026
Meta gave its AI agent a second agent whose only job is remembering things
Agents on long tasks forget constraints and repeat failed commands. Meta AI's answer is a separate memory agent that keeps a structured record and decides when to interrupt with a reminder, worth up to 8 points on two benchmarks.
3 Aug 2026
Ten million dollars is on the table for research into what happens when AI agents meet each other, and the deadline is Saturday
Google DeepMind, Schmidt Sciences, the Cooperative AI Foundation and ARIA are funding research into multi-agent safety. Applications close on August 8. Nearly all existing safety testing looks at one model on its own.
3 Aug 2026
OpenAI now answers its own support phone line with an AI agent, and is selling the system that does it
OpenAI Presence is a packaged product for putting voice and chat agents into production, with policies, guardrails and escalation rules. It runs OpenAI's English phone support and reportedly resolves 75 percent of calls without a human.
3 Aug 2026
People are building playable 3D games from a single prompt, and the results stopped looking like blocks
A first-person shooter, a kart racer, a submarine game and a Minecraft clone, each generated by Claude Opus 5 from one prompt with no external art files. Everything on screen is code that runs in the browser.
3 Aug 2026
AI is finding a flood of security flaws. Almost none of them get attacked
VulnCheck counted 1,061 vulnerabilities found with AI help in the first half of 2026. Fourteen were confirmed exploited, a rate of 1.3 percent, which is roughly the same as vulnerabilities found any other way.
3 Aug 2026
AMD trained a fully open model on its own chips and published everything except a commercial licence
Instella-MoE-16B-A3B has 16 billion parameters but uses only 2.8 billion per token. AMD released the weights from every training stage, the data mixtures and the code, and licensed the weights for research only.
2 Aug 2026
The number of AI chips in the world is doubling roughly every nine months
Epoch AI counts about 20 million AI chips in data centres today and expects around 200 million by the end of 2028. That is the single number that explains most of what happens next.
2 Aug 2026
Google put an image generator in Google Earth, then pulled it a day later
Nano Banana 2 let anyone type a prompt and get a fake satellite view of a real place. Researchers produced refugee camps and a nuclear plant that do not exist, and Google rolled the feature back within roughly 24 hours.
2 Aug 2026
One in four malicious breaches now involves AI, and those cost a million dollars more
IBM's annual breach report puts AI-enabled attacks at 25 percent of malicious breaches, up 56 percent in a year, averaging 6 million dollars against a global average of 4.99 million.
2 Aug 2026
OpenAI named its next model family Astra and introduced it with ten solved math problems
An internal version of Astra produced results for ten problems mathematicians had not moved on in at least a decade. The compute bill was about 2,000 dollars, and every proof ships with a machine-checkable certificate.
2 Aug 2026