So OpenAI wants to build a super app. Again. Still. Whatever tense applies when a company keeps announcing the same strategy until it becomes real.

According to the Financial Times, a significantly overhauled ChatGPT is coming in the next few weeks — one that bundles coding tools, AI agents, and what OpenAI's core product lead Thibault Sottiaux describes as "your own personal agent that is capable of helping you across everything in your life, be it personally or at work." Bold words. Let's unpack what's actually going on here.

The "Chat Is Dead" Take

An unnamed senior OpenAI employee reportedly declared, "Chat is dead." Which is a fascinating thing to say about the product that put your company on the map and currently pays a non-trivial portion of the bills. But they're not entirely wrong — raw chat interfaces are becoming table stakes. Every competitor has one. The differentiation battle is moving elsewhere: into agentic workflows, deep tool integrations, and the kind of persistent, context-aware assistance that actually justifies a monthly subscription.

The real translation of "chat is dead"? Chat alone doesn't convert free users into paying customers anymore. That's the business problem OpenAI is trying to solve before it heads toward an IPO. Turning ChatGPT into a funnel — one where casual users get pulled toward products like Codex (the coding assistant that enterprise teams might actually open their wallets for) — is the play.

The Super App Strategy, Explained Without the PR Gloss

The concept isn't new. Reports of OpenAI's super app ambitions surfaced last year, and the Wall Street Journal documented this strategic pivot back in March. What changed? OpenAI launched a flurry of standalone products in 2025 and apparently learned a hard lesson: fragmentation kills retention. Users don't want ten separate apps — they want one interface that does everything without making them context-switch constantly.

So now OpenAI is consolidating. The targets include:

• Coding tools — Codex, positioned for developers and enterprise customers willing to pay premium pricing
• AI agents — autonomous task runners that can operate across your digital life, not just answer questions
• A unified platform layer — think less "chatbot" and more "operating system for personal productivity"

The casualties of this focus shift are telling. Sora, the video generator that generated enormous buzz and then mostly silence, has been shuttered. Executives are publicly calling out "side quests" — their term for the product experiments that diluted engineering focus without building durable revenue.

What Could Actually Go Wrong Here

Super apps are notoriously hard to pull off outside of specific ecosystems. WeChat works in China partly because of regulatory moats and deeply embedded social infrastructure. Western attempts to replicate that playbook have a mixed-to-poor track record. The graveyard of "everything apps" is well-populated.

The specific challenges OpenAI faces are worth naming directly:

• Inference costs don't disappear because your UI is prettier. Bundling more agentic features means more compute per user session. The unit economics need to work at scale.
• Agents that operate "across everything in your life" create real security and privacy exposure. Enterprise customers especially will want answers on data handling before they hand over access to their workflows.
• Anthropic isn't standing still. OpenAI explicitly wants to close the gap with Anthropic among business customers — which implies they currently lag there. That's a competitive gap worth taking seriously.
• The IPO timeline creates pressure that can warp product decisions. Building for profitability metrics and building the best product aren't always the same objective.

The Genuine Opportunity

None of that means this strategy is doomed. If OpenAI can genuinely deliver a coherent, capable agent layer — one where context persists, tasks complete reliably, and integrations don't break constantly — there's a real product here. The demand for "AI that actually does things, not just talks about doing things" is enormous and largely unmet.

The gap between what AI agents are demoed doing and what they reliably do in production is still embarrassingly wide. If the revamped ChatGPT closes that gap meaningfully, the super app framing becomes secondary — it'll just be called "useful."

We'll see what ships in the next few weeks. Announcements are cheap. Reliable agentic execution at scale is not.