OpenAI Bets Everything on One Agentic Platform

Greg Brockman is now running OpenAI's products. Not as a stand-in for Fidji Simo, not as a temporary coordinator - as the permanent head of a company that has decided it can't build four things at once anymore. Last week's internal memo, reported by TechCrunch and confirmed by multiple outlets, is the clearest statement OpenAI has made about what it actually wants to be: one unified agentic platform, not a portfolio of experiments.

TL;DR

Brockman permanently assumes product leadership as Simo recovers from medical leave
ChatGPT, Codex, and the developer API merge into one unified "agentic platform"
Sora paused; OpenAI for Science shut down; adult content mode shelved
ChatGPT's AI chatbot traffic share dropped from 86.7% to 64.5% over the past year
OpenAI's Q4 2026 IPO needs a cleaner product story - this is it

A House Divided

The Code Red That Started It

Sam Altman declared a "code red" on ChatGPT in December 2025. The message to staff: the core product had stalled while rivals caught up, and the company needed to stop spreading itself thin. At the time, OpenAI was running three largely separate product lines - ChatGPT under Nick Turley, Codex under Thibault Sottiaux, and the developer API under its own engineering leadership. Each one competed for GPU time and engineering headcount.

That structure had a visible cost. Codex, launched as a standalone cloud coding agent, had grown into one of OpenAI's fastest-expanding products. But ChatGPT - the product that made OpenAI a household name - was losing ground. Web analytics data cited by multiple outlets shows ChatGPT's share of AI chatbot traffic fell from 86.7% to 64.5% across 2025, while Google's Gemini climbed from 5.7% to 21.5%.

Simo Steps Back

Fidji Simo, hired as CEO of AGI deployment to coordinate OpenAI's product lines, went on medical leave before the restructuring concluded. OpenAI said she participated in designing the new org structure despite her absence and is expected to return. Brockman had been covering her responsibilities informally. He now holds them permanently.

His memo was direct:

"We're consolidating our product efforts to execute with maximum focus toward the agentic future, to win across both consumer and enterprise."

Everything Into One

What the Memo Actually Said

The plan is to build a single agentic platform that combines ChatGPT's conversational layer, Codex's code execution and task-running capabilities, and the developer API into one product with one team. The platform will include a built-in browser, a code execution layer, and an interface built to complete multi-step tasks without the user switching tools. OpenAI is calling this a "super app" internally.

Sottiaux - who built Codex into a product that drew enterprise contracts away from GitHub Copilot - leads the combined product organization. Turley moves to enterprise products and critical industries. Ashley Alexander, formerly VP at Instagram, takes consumer product responsibility.

Who Leads What Now

The new org concentrates authority in fewer hands. Brockman sits above Sottiaux on strategy. Turley's move to enterprise removes one competing voice from the consumer product side. The structure mirrors what Anthropic did earlier this year by collapsing its research and product functions under a narrower leadership group.

Two notable departures accompany the restructuring. Kevin Weil, who oversaw products including Sora and GPT-4o, has left. Bill Peebles, who led Sora's technical development, is also out. The exits remove the people most associated with the experiments now being ended.

OpenAI ChatGPT logo displayed on a computer screen ChatGPT's declining market share - from 86.7% to 64.5% over 2025 - is part of what drove the consolidation decision. Source: unsplash.com

What Got Cut

Sora's Long Goodbye

Sora, OpenAI's video generation model, is paused. Its shutdown had been signaled earlier this year when a planned $1 billion Disney investment collapsed over concerns about reliability and cost. High GPU usage relative to revenue made Sora a poor candidate to survive an IPO process that rewards simplicity. Video generation hasn't found a path to sustainable margins at any of the labs building it, and OpenAI is no longer willing to subsidize the search.

The Other Casualties

OpenAI for Science, a research tool aimed at academic and life-sciences workflows, has been shut down. The adult content mode, which launched quietly and generated quick internal and external pushback, has been shelved. Together with Sora, these three products represent a significant reduction in OpenAI's surface area - fewer teams, fewer compute allocations, fewer roadmaps to coordinate.

Brockman described the logic in a separate staff note cited by The Decoder: "computing power is 'not enough for even a personal assistant and the Codex line.'" The consolidation is partly about focus and partly about resource allocation. OpenAI doesn't have infinite GPUs.

The Competitive Pressure Behind the Pivot

OpenAI doesn't need to explain its market share problem in the memo - the numbers explain it. Cursor, the AI coding assistant built by Anysphere, surpassed $2 billion in annualized revenue in February 2026 and is now pursuing a $50 billion valuation in a new fundraise led by Andreessen Horowitz. Cursor got there largely by doing what Codex does, but as a standalone tool embedded in the editor. OpenAI is catching up, but Cursor captured that revenue first.

On the consumer side, Gemini has taken real ground. Anthropic's Claude continues to grow in enterprise. Every major tech company now ships a chatbot. The period when ChatGPT could hold 87% of AI chat traffic by default is over.

Fragmentation doesn't cause that competitive pressure, but it made responding to it harder. A product org split across three product leaders competing for GPU time wasn't set up to move quickly.

Abstract AI technology visualization The unified platform will combine ChatGPT, Codex, and the developer API under Thibault Sottiaux's product leadership. Source: unsplash.com

The IPO Frame

OpenAI is targeting a Q4 2026 public listing at a valuation of approximately $852 billion. The company raised $122 billion in its most recent private round at a similar level. Public market investors tend to reward clear product narratives, and a company running four parallel experiments across video generation, consumer chat, enterprise agents, science tools, and adult content isn't a clear narrative.

The restructuring removes most of those experiments before anyone has to ask about them. A single unified agentic platform - chat, code, browser, task execution - fits on one slide. The Musk lawsuit verdict last week cleared one governance question before the IPO. This clears another.

What the memo doesn't answer is whether users actually want one AI app that does everything rather than a set of specialized tools. Sottiaux has until the Q4 roadmap to show what the unified platform looks like in practice, and until the IPO to show that it works well enough to justify the valuation.

Sources: