💬 OpenAI folds ChatGPT's Chat and Work tabs together 💬
One workspace changes how users access ChatGPT's tools.

OpenAI is preparing to merge the separate Chat and Work tabs in ChatGPT into a single workspace, according to the latest reports. Instead of jumping between modes depending on the task, users would find everyday conversations and longer, work-oriented requests in one place, with usage limits shared across the unified experience. The change is aimed at simplifying access to ChatGPT's growing set of tools, which has become harder to navigate as new features keep landing. For individuals and teams on paid plans, a shared quota also means one budget to track rather than several. A cleaner interface matters, as friction now decides which assistant people open first. No firm rollout date has been confirmed yet.

🚀 GPT-6.2 Astra spotted in Codex before launch 🚀
A staged rollout signals OpenAI's next coding model is close.

References to GPT-6.2 Astra have started appearing inside Codex, OpenAI's coding agent, suggesting the model is being tested through a staged rollout before a public launch. Sightings like this often signal a release is close, as OpenAI tends to expose new models to a small group of users first to monitor quality and load. If Astra follows the pattern of previous coding upgrades, developers could see gains in speed, reliability on longer tasks and tool use inside Codex. Timing matters here: coding agents are the most contested market in AI right now, with every major lab shipping updates. OpenAI has not yet confirmed specs, pricing or an official release date.

⚠️ Anthropic IPO filing warns Claude nears self-improvement ⚠️
Self-improving models could bring tighter regulation and faster research.

In its IPO filing, Anthropic warns investors that Claude is approaching a point where it could meaningfully contribute to improving itself. Risk disclosures in a listing document are written for regulators and shareholders, so the language carries legal weight rather than marketing spin. A self-improving Claude could accelerate research speed inside the company, compressing the time between model generations. At the same time, it is exactly the scenario that could invite closer regulatory scrutiny, as policymakers weigh how much autonomy frontier systems should have. For investors, the warning frames regulation risk as a core business factor. It also puts pressure on rivals to explain how they would manage similar capabilities as their own models advance.

INTERESTING TO KNOW

🏦 Claude now reads your bank accounts 🏦

Anthropic is opening access to bank account data in Claude, with the integration starting as a beta. Once connected, Claude could read account information to help users review spending, spot recurring charges or answer questions about their finances in plain language. It is a distribution play into personal finance, where trust decides adoption, so expect privacy controls to be the main point of scrutiny as the rollout widens.

🧬 DeepMind open-sources AlphaEvolve on GitHub 🧬

Google DeepMind is widening access to AlphaEvolve by publishing it as open source on GitHub. AlphaEvolve is the agent that pairs Gemini models with automated evaluators to discover and refine algorithms, previously available only to selected partners. Researchers and developers can now inspect the code and run it on their own problems, an adoption signal that could speed up algorithm research outside Google's labs.