Amodei on the AI Backlash: 'This Is a Crisis of Trust'
An investor accuses Dario Amodei of fueling the backlash against AI with his warnings. Amodei fires back: the problem isn't his messaging, but a distrust that's been building for decades.
Anthropic Made $11.5 Billion in Revenue – in One Quarter
Anthropic is showing prospective investors its numbers: more than $11.5 billion in second-quarter revenue, a 14-fold jump in a single year. The autumn IPO is getting closer.
Anthropic's Risk Report: Higher Risk – and a Secret Model 2
In its second risk report, Anthropic bumps the risk of catastrophic misalignment from 'very low' to 'low'. Along the way, the document mentions an internal Model 2 that's stronger than Mythos 5 – and stays locked in the vault for now.
The Old Workbench Goes Dark Today – and Three Prompt APIs With It
As of today it's over: the legacy Workbench in the Claude Console is gone, and the experimental prompt endpoints for generating, improving, and templatizing now return nothing but an error. If you still use them, it's moving day.
ChatGPT Will Quiz You Now – and Let You Switch a Project's Memory
OpenAI handed ChatGPT a few solid little upgrades: interactive quizzes right inside the chat, switchable memory per project, and the Think button now on Free too. Nothing earth-shaking – but you'll feel it day to day.
Google Retires Imagen 4 – and the Switch Isn't a Simple Model Swap
As of today, three Imagen 4 models in the Gemini API can be shut down. The successor is nicknamed 'Nano Banana' – but moving over means touching your code too: the old generate_images method is gone.
Claude Code 2.1.233: The Todo Tools Are Gone – and a Windows Leak Is Sealed
On the newer models there are no more todo lists in the terminal, GitLab merge requests now show up in the agents view, and an NTLM leak on Windows is closed. A release with one change you'll notice right away.
Claude Tag Now Reads the Whole Slack Channel – and Decides for Itself When to Help
Claude in Slack used to see only the one message right in front of it. Now it reads the whole channel, uses its memory, and jumps in when it has something to add – or stays quiet. At no extra cost.