ChatGPT gets "Computer History" memory

Welcome back, Superhuman. Typing out prompts to your LLM is becoming outdated. These days, you can verbally prompt your AI, walk it through on-screen demos, and now, thanks to a recent update, you can even let ChatGPT study your old computer screen activity to learn new skills.

Today: Microsoft Copilot gets stripped down, how to write and edit docs faster with Claude, and get the latest prompts and trending social posts.

1. ChatGPT can now remember what you worked on yesterday and learn from it: Computer History turns your computer activity into context that ChatGPT can use to give you more relevant answers. You can also reference tasks you already completed and have ChatGPT turn them into repeatable skills (see how it works). In separate news, OpenAI rolled out Ultrafast mode for GPT-5.6 Sol that works at 14x its normal speed. It’s launching to a limited group of companies. Request access here.

2. Microsoft is stripping features from Copilot ahead of a major release: The tech giant is merging its consumer and business Copilot apps into one single tool, while also eliminating Group Chats, Podcasts, Deep Research, Copilot Labs, and Mico — the company’s Clippy-like mascot that was announced last October. The redesign creates a cleaner foundation for Microsoft’s upcoming “super app,” which is expected by the end of September.

3. AI infrastructure stocks recover June’s losses after a strong earnings week: CoreWeave, SMCI, and Nebius all reported strong earnings and upbeat revenue projections this week, signaling that demand is still booming for AI compute capacity. In particular, Nebius’ stock rose roughly 40% on the week after it revealed $582M in Q2 revenue, a 450% year-over-year jump.

Yesterday’s most-clicked story: Anthropic’s Claude got even better at using the internet, thanks to an update that lets it take over any Chrome internet browser.

Half of your important conversations happen outside the office or off-screen. By the next morning, the details are gone.

Memoket Gem is the AI wristband that catches them all. A 0.4 oz wearable that records your conversations with one press and turns them into instant summaries, action items, and full context across every meeting. Ask what happened, what changed, and what's next. All the details can be synced with ChatGPT, Claude, Google Calendar, Notion, or your own AI agent.

Pre-order is now open — get the device with a free year of the app. Ships in August.

The lab's Imagine Image 2.0 specializes in precise editing and workflow templates, already ranking #3 on Arena's text-to-image leaderboard. A day later, SpaceXAI launched Grok Bot, an always-on computer agent that can work across your desktop or the internet. Then came Grok 4.6, a frontier model matching OpenAI’s GPT-5.6 Sol and only trailing Anthropic’s top models.

Moonshot became the fourth lab to have a model escape from a cybersecurity testing environment, joining Anthropic, OpenAI, and Meta. Separately, OpenAI declared Astra its first "critical" model for cybersecurity, meaning the lab believes it could potentially carry out a real-world cyberattack. Read OpenAI's full statement.

In a personal essay, Meta's CEO argues that AI's potential to help people invent new products outweighs job-automation risk, saying access to AI should be open-source and broadly distributed. The essay landed the same day Meta released Muse Glimmer, an open-weight model small enough to run on a Mac or PC.

MatrAIx is an AI simulation infrastructure of 8.3 billion agents built from real public-record human profiles supplemented with synthetic personas, letting researchers simulate human behavior at scale. Watch how it works here.

Readers also loved this new game that drops you into a random historical scene and challenges you to guess when and where you are. Test your knowledge.

⚽ Hogwarts Soccer Team: A viral clip blends Harry Potter characters and famous soccer players into one person using AI. The random mash-ups actually work surprisingly well, causing the video to rack up over 11M views. Watch it here.

📝 Snitch Mail: OpenAI has a unique strategy to eliminate bottlenecks. The AI lab reportedly gives all employees a direct email to contact Sam Altman, mirroring an approach popularized by Jeff Bezos. This article breaks down the strategy.

🧑‍⚖ Prompt Objection: There’s no historical precedent for this. A person representing themselves in court tried to prompt inject the court’s AI systems into taking his side. They later claimed it was a joke, but the court didn’t agree. Read the full story.

🚽 The Compiss: This vibe coded app might be a lifesaver during your next road trip — it’s a digital compass to the nearest public bathroom. The app’s gone viral Reddit, with commenters praising its clever name (5K upvotes).

🤖 Auto Continue: Claude Code just got a new update that makes it automatically continue working once your usage limit resets. No more setting a timer to check back in hours later.

Skip the lectures. 70% of the 2,200+ sessions are interactive: workshops, code talks, AWS Jams, GameDays, and the Architecture Rodeo. Run demos against real workloads with sandbox access in small groups. Las Vegas, Nov 30–Dec 4.

Register by August 25 and save $1,200.

Grow customers & revenue: Join companies like Amazon, HubSpot, and Salesforce. Showcase your product to our 4 million+ readers and followers on socials. Get in touch.

Until next time — Zain, Theodore, & the Superhuman AI team.