Presented by
Good morning, AI enthusiasts. The world's most-used chatbot is about to stop being a chatbot. OpenAI is planning its biggest ChatGPT overhaul ever, turning it into a full superapp built around coding tools and AI agents.
Six hundred million monthly users is a pretty big canvas. With an IPO on the horizon and Anthropic steadily gaining enterprise ground, is ChatGPT's identity crisis finally over?
In today's recap:
OpenAI's biggest ChatGPT overhaul, superapp plans
Trump eyes a US government stake in OpenAI
Build AGENTS. md files that actually help coding agents
4 new AI tools, prompts, and more
AI
OpenAI plots biggest ChatGPT overhaul yet
Getty Images
Recaply: OpenAI just outlined plans to rebuild ChatGPT into a superapp combining coding tools and AI agents into one platform, targeting a jump in business revenue from 40% to 50% of total income.
Key details:
Codex, OpenAI's AI coding assistant, sits at the center of the new interface, allowing developers to run multiple instances at once and track progress in real time.
ChatGPT currently serves 600 million monthly active users. Two million businesses are on the platform, generating 40% of total revenue, a share OpenAI wants to raise to 50%.
The revamp is timed ahead of OpenAI's planned IPO. Sources told the FT the company wants to show strong enterprise growth before going public.
The new interface rolls out within weeks, starting with developer-facing features before expanding to all users.
Why it matters: ChatGPT started as a chatbot, but OpenAI always had bigger plans. With 600 million users and an IPO approaching, the timing for a full platform push is right. Anthropic is gaining ground in enterprise, and OpenAI needs ChatGPT to be more than a chat window. The superapp bet is a race to lock in business customers before rivals do.
PRESENTED BY MINTLIFY
AI agents now read your docs almost as much as humans do.
Mintlify analyzed 790 million requests across its documentation platform. The finding: AI coding agents account for 45.3% of all traffic, nearly tied with traditional browsers at 45.8%.
Two tools are driving almost all of it:
Claude Code: 25.2% of total traffic, more requests than Chrome on Windows
Cursor: 18% of total traffic
Together they account for 95.6% of all identified AI agent traffic
The rest of the field, OpenCode, Trae, ChatGPT, and NotebookLM, is showing up but nowhere close.
One caveat: OpenAI's Codex doesn't send an identifiable user-agent header, so the real agent percentage is likely even higher.
The takeaway for anyone maintaining developer docs: your documentation now serves two audiences. Structure and machine-readability matter as much as clarity for human readers.
AI
Trump eyes government stake in OpenAI
Getty Images
Recaply: OpenAI CEO Sam Altman just confirmed ongoing White House talks about a potential US government stake in OpenAI, with discussions reportedly centering on equity of 1% to 5% of the company.
Key details:
The proposed deal would use a Public Wealth Fund model to hold any government shares, similar to sovereign wealth funds used by states like Alaska.
OpenAI is valued at $850B, meaning a 1% stake would be worth around $8.5B. Axios reports talks range between 1% and 5%.
The proposal has drawn unusual cross-party support, with both the Trump administration and Senator Bernie Sanders reportedly backing the idea, though for different reasons.
No formal deal has been announced yet. OpenAI is also restructuring as a Public Benefit Corporation, which makes any government ownership more complex.
Why it matters: A government stake in a private AI lab would be a first in the US. The fact that both Trump and Bernie Sanders back it shows how much AI power has shifted. If a deal happens, the US would hold direct equity in its leading AI company, changing the accountability picture for everyone from OpenAI's board to its customers.
GUIDES
Build AGENTS. md files that actually help coding agents
Recaply: In this tutorial, you will learn how to build AGENTS. md files that actually improve coding agent performance, drawing on research that separates helpful context from expensive noise.
Step-by-step:
Identify the details your agent can't infer from the codebase. Common examples include API rate limits, naming conventions, environment constraints, and dependency choices not visible in the code.
Run 3 to 5 test tasks without AGENTS.md first. Note every failure, repeated question, or incorrect assumption. This becomes your baseline for comparison.
Write a short markdown file with only facts the agent can't determine from training. Skip general advice and anything it already knows. Target under 500 words.
Re-run the same tasks with the file active and compare outcomes. ETH Zurich research found that well-written context files improve task success rates by around 4%, while adding about 19% more tokens per session.
Remove any entries the agent didn't use or that didn't reduce errors. LLM-generated AGENTS.md files cost about 20% more tokens and lower success rates by 3%, so unused lines carry a real cost.
Pro tip: Keep one test task in rotation for the most error-prone area of your workflow. Run it before and after each update to spot regressions early.
TOGETHER WITH WISPR FLOW
10x the context. Half the time.
Speak your prompts into ChatGPT or Claude and get detailed, paste-ready input that actually gives you useful output. Wispr Flow captures what you'd cut when typing. Free on Mac, Windows, and iPhone.
TOOLS
Trending AI Tools
🔒 Lockdown Mode - OpenAI's enhanced security mode for ChatGPT sessions handling sensitive data, restricts external connections, free with all plans
📧 Gmail Sync - Personalizes ChatGPT responses using real email context from your Gmail inbox, free with ChatGPT
🎵 Magenta RealTime 2 - Google DeepMind's open-weights live music AI, runs locally on Apple Silicon MacBooks at 200ms latency, free
💬 Poke - First AI agent approved for Apple Messages for Business, runs personal AI assistance via iMessage, SMS, and Telegram
NEWS
What Matters in AI Right Now?
ChatGPT crossed 600 million monthly active users for the first time, according to a Similarweb report shared on X, marking the fastest user growth since the platform launched in 2022.
Apple held a top-secret executive meeting in early 2025 to address the company's AI crisis, Bloomberg reports, with the session setting the agenda for significant AI announcements expected at WWDC 2026 this week.
Anthropic's unreleased Mythos model has started appearing in developer testing environments, according to tracking account @testingcatalog, suggesting Anthropic is preparing the new model for a broader reveal.
OpenAI introduced Lockdown mode, a new security setting for ChatGPT that restricts external connections and disables data-sharing features, available for free to all users handling sensitive sessions.
Moonshot AI is seeking a $30B valuation in a new funding round, up from its previous $20B valuation, driven by growing demand for its Kimi AI assistant in China and globally.
NVIDIA and SK hynix announced a multiyear partnership to develop memory systems optimized for AI factory workloads, combining NVIDIA's Vera Rubin AI platform with SK hynix's next-generation HBM memory chips.
🧡 Enjoyed this issue?
🤝 Recommend our newsletter or leave a feedback.
How'd you like today's newsletter?
Cheers, Jason
