GPT-6 grew buttons. Your job grew nervous.
GPT-6 for everyone, ads in ChatGPT, Claude moves into your Docs
Hey, sapiens.
Good news: the robots still need us. Mostly to look at ads. This week ChatGPT put GPT-6 in front of 1.2 billion people, started selling space next to your AI pictures, and Claude moved into your Google Docs without asking about the rent. Grab a coffee while coffee is still made by humans.
GPT-6 reaches all of ChatGPT, and it brought buttons
The gistOpenAI is pushing GPT-6 to everyone on ChatGPT, paid and free. That is more than 1.2 billion people a week, which is roughly every human who has ever lost an argument with autocorrect. The headline feature is Intelligent UI. Instead of a wall of text, answers can come back as buttons, forms, charts and small custom calculators. Ask about a mortgage and you might get a slider, not a lecture. The rollout is staggered. Paid tiers get GPT-6 Sol first, and Free and Go users get GPT-6 Luna a day later. Why it matters: the chatbot is turning into an app that builds itself on demand. Every small tool you used to open separately, from the unit converter to the budget sheet, now has a rival living inside a chat box. What to do: try one task you normally do in a spreadsheet and see what GPT-6 hands back. Then decide which of your tabs you still need.
Claude moves into your Docs, Sheets and Slides
The gistAnthropic opened a public beta of Claude for Google Workspace for paying Claude users. A sidebar in Docs, Sheets and Slides reads the open file and edits it in place. New connectors also let claude.ai create and edit Google files.
The assistant now lives where the work already is, so the copy-paste dance is over. Admins may need to switch the connectors on first, so ask yours before you get excited.
ChatGPT starts selling space next to your pictures
The gistOpenAI is testing visual display ads beside ChatGPT image results, in the US, starting later this month. The ads sit apart from the generated image, and OpenAI says they will not shape answers.
Plus, Pro and Enterprise subscribers see none, so free users get the billboard. The free chatbot era now has a business model, and it looks familiar.
OpenAI does maths, cryptographers check their locks
The gistOpenAI posted new results on open maths problems from an internal frontier model, with Lean proofs on GitHub. Scott Aaronson says AI labs have quietly started testing whether models can break crypto protocols.
Vitalik Buterin says the risk deserves attention, and a mathematicians group urged members to stop working with OpenAI. When the maths gets good, the locks get nervous.
Meanwhile in Europe
OpenAI will quietly sign ChatGPT text in the EU
The gistAn invisible textGrain watermark is coming to ChatGPT and Codex output in the EU, to meet the AI Act. Short text is detected around 80 percent of the time, and editing makes it worse.
Click the new GPT-6 buttons before they start clicking you. Ciao, sapiens.