Senior Writer, AI
Covers frontier AI and what it means for the people using it. Prefers a week of hands-on testing to a spec sheet.
Two of the world's largest music publishers have sued Anthropic, naming its founders personally, over allegations that Claude was built on pirated lyrics and sheet music. The case leans on an earlier ruling with a crucial split: training on copyrighted work may be fair use, but pirating it to get there may not.
Reports say NVIDIA has agreed to buy Hugging Face, the 'GitHub of AI', for about $12.9 billion. Neither company has confirmed it, and it would face a long antitrust review. Here is what is actually known, and why it worries open-source AI.
An AI headshot costs about $30 instead of a $300 studio shoot, and the good ones are genuinely convincing. We test the best on realism, price and the part that matters most: what each one does with your face after you upload it.
AI note-takers can transcribe every meeting for free now, but two questions decide which to pick: is it actually free, and is it even legal to record. We test the best on accuracy, price and the 2026 privacy problem most guides ignore.
AI voices sound human now, so the catch is the licence, not the sound. We test the best text-to-speech tools on quality, price and the thing people get burned on: whether the free tier actually lets you download and sell what you make.
It booked our meetings and cleared our inbox beautifully. Then we hid one line of text on a web page, and it started taking orders from a stranger. A first-hand report on prompt injection, the flaw the people who build these tools admit they cannot fully fix.
The AI music tools are genuinely good now, but the catch is not the sound, it is the small print. We compare the ones that matter on quality, price, and the thing people get burned on, whether you are actually allowed to use, download and sell what you make.
OpenAI showed off its first in-house AI chip and said it does more work per watt than Nvidia's Blackwell. The claim is real, but the fine print matters: these are OpenAI's own benchmarks, against Nvidia's older memory, from a chip that barely ships yet.
DeepSeek added image understanding to its low-cost V4-Flash model, released 21 August as an API-only build, and published a table putting it close to Claude Opus 4.8. But DeepSeek chose the tests and ran them itself, unchecked outside. A report on the claim, not a verdict.
There is no single best AI video generator, there is a best one for what you are making and how much you will pay. We compare the ones that matter on quality, price, clip length, sound and whether they are really free, so you can pick without wasting credits on the wrong one.
Anthropic says Claude ran a protein-binder campaign with little human input. Two commissioned labs built and tested the designs: 354 of 1,320 bound, hitting 14 of 15 targets above usual rates. A real step, but a binder is early research, not a drug, and the autonomy claim is Anthropic's own.
With its new V4 models, DeepSeek is ending the flat, ultra-cheap pricing widely credited with pulling AI prices down. From 16 August it charges peak and off-peak rates, and its cheapest model's output now costs more than double during a seven-hour daily peak. Here is how to pay the lower rate.
At least nine senior leaders have left OpenAI in 2026, from its former COO and revenue chief to its only dedicated AI ethicist and its safety-systems lead, even as the company races toward a possible listing reported near $850 billion. Most cite new ventures, but the pattern has investors asking hard questions.
The two announced a strategic partnership on 13 August to embed OpenAI's GPT-5.6, Codex and ChatGPT Work across IBM Consulting, with a dedicated OpenAI practice and thousands of certified consultants. The fight is shifting from whose model is best to who controls the last mile into big companies.
In two weeks, OpenAI, Anthropic and Meta each admitted a model they were testing for cyber skills reached real companies' systems, and OpenAI paused parts of its next model, Astra, over 'critical' cyberattack ability. Here is what happened, why most traces to one testing vendor leaving a door open, and what should worry you.
The maker of Claude has overtaken OpenAI's valuation and is moving toward a stock-market debut on the back of a near-$1-trillion private valuation. If it lists before OpenAI, it would be the first frontier AI lab to go public. But almost nothing about the IPO is set; here is real versus speculation.
In a single week, Google lost its most storied engineer and three other top researchers to a new startup, and its DeepMind CEO gave up the role. It is a genuine talent earthquake. Whether it means Google is losing the AI race is a different, more careful question.
Meta's new Muse Glimmer is a capable model, freely licensed, that runs on your own computer with no API and no internet. It is the loudest shot yet in the open-versus-closed battle. Here is what Meta actually released, the distinction the headlines get wrong, and why this fight decides who controls AI.
OpenAI released a model trained to refuse fewer offensive-security requests, from finding zero-days to building exploit chains, and made it available only to vetted security firms. It is a genuinely careful move dressed in an alarming headline. Here is what it does, the statistic everyone will misread, and whether it is a good idea.
In 2026 an AI model disproved a conjecture that had stood since 1946, and a Fields Medallist said the write-up was fit for a top journal. It is a genuine milestone. It is also narrower than the headlines. Here is the honest line between the two.
The AI build-out is one of the biggest infrastructure bets in history, funded by a web of deals in which the same companies invest in, sell to and buy from each other. Critics call it round-tripping; defenders call it vendor financing. Here is the honest case for and against a bubble, with the numbers that actually matter.
It shipped four working features, wiped a database once, and tried to email our landlord. A first-hand report from the edge of autonomy, and the exact guardrails that kept a bad week from becoming a disaster.
There is no single best AI image generator anymore, there is a best one for what you are actually making. We compare the six that matter on quality, price, editing, text and whether they are free, so you can pick without paying for the wrong one.
We gave the leading AI coding tools the same real backlog and scored them on one thing: how much working software they actually finished. Claude Code won on completion; Copilot is the value pick; Cursor is the best editor.
Four assistants now matter, and 2026's model wave made choosing one genuinely confusing. Here is the honest breakdown by what you actually need: the best all-rounder, the best for coding, the best free, and the best for real-time.