OpenAI has launched a major memory upgrade for ChatGPT, allowing the chatbot to reference every conversation you've ever had — even ones you didn't explicitly ask it to save. CEO Sam Altman announced the update on X, calling it a step toward "AI systems that get to know you over your life."
The update builds on last year's "Memory" feature, expanding it with two tiers: "saved memories" that users explicitly asked ChatGPT to retain, and "reference chat history" — insights the system mines from past conversations on its own. For now, the upgrade is rolling out to Pro subscribers ($200/month), with Plus access coming soon. It will not be available in the EU or UK due to strict AI regulations.
Why this matters
Separately, Anthropic published a wide-ranging piece titled "When AI Builds Itself" — a serious warning about recursive self-improvement (RSI). The company reveals that Anthropic engineers now ship 8x more code per quarter than they did from 2021-2025, thanks to autonomous AI coding agents. The trend, Anthropic argues, points to a future where AI systems can fully design and develop their own successors.
The numbers are staggering. In March 2024, Claude Opus 3 could complete four minutes' worth of software tasks. One year later, Claude 3.7 Sonnet handled tasks that take a human about 90 minutes. Fast forward another year: Claude Opus 4.6 now manages 12-hour tasks. If this trajectory holds, AI systems could be handling multi-day tasks by the end of this year.
What experts are saying
This week also saw major moves in the AI compute arms race. Google signed a "short-term" deal with SpaceX to tap compute capacity at the Colossus 1 data center in Memphis, following Anthropic's similar partnership with SpaceX last month. Google cited "surging customer demand for our agent platform, Gemini Enterprise, which has been even higher than we expected."
Meanwhile, reports emerged that Sam Altman discussed the possibility of the Trump administration taking a stake in OpenAI — pitched as a way to bring the economic benefits of AI to the public.
On the regulatory front, New York passed a first-of-its-kind bill barring AI chatbots from acting as companions to minors. The legislation follows lawsuits against Character.AI and OpenAI alleging their chatbots encouraged suicidal thoughts among teenagers.
The bottom line
The market is moving fast on every front: OpenAI is pushing AI deeper into personal territory, Anthropic is raising red flags about how fast the technology is accelerating, and tech giants are racing to lock down compute capacity — including SpaceX-backed data center deals and the controversial Stratos Project in Utah. The next decade may be the most consequential (and the most concerning) in the history of artificial intelligence.
