OpenAI is merging its flagship products into a single desktop application while simultaneously pursuing the ambitious goal of building an autonomous AI researcher capable of independently advancing science.
The Superapp Strategy
The San Francisco company announced plans to combine ChatGPT, the Codex AI coding tool, and the Atlas AI browser into one unified application, according to a memo from Fidji Simo, OpenAI's CEO of Applications, cited by The Wall Street Journal. Simo acknowledged that internal fragmentation "has been slowing us down and making it harder to hit the quality bar we want."
This consolidation marks a significant strategic shift. OpenAI is abandoning what internal documents call "side quests" — notably winding down its Shopping initiative that partnered with Walmart, which reportedly saw conversion rates around one-third of click-outs. Instead, the company is doubling down on enterprise applications and coding tools.
The Astral Acquisition
The superapp merger comes alongside OpenAI's acquisition of Astral, a coding startup that joins OpenClaw and other open-source projects in the company's growing portfolio. This acquisition completes a pattern established when Google acquired the Antigravity team last July, and Anthropic purchased Bun last December — major AI labs are racing to own developer tooling infrastructure.
The move signals AI companies prioritizing coding and enterprise applications as core competitive advantages. OpenAI's decision to unify ChatGPT and Codex apps into a single "superapp" reflects this prioritization at the highest levels.
Cursor, meanwhile, has emerged as a formidable competitor in the AI coding space. The company's Composer 2 launch demonstrated impressive benchmarks: $0.50 per million input tokens and $2.50 per million output tokens, with scores of 61.3 on CursorBench, 61.7 on Terminal-Bench 2.0, and 73.7 on SWE-bench Multilingual. Third-party engineers noted Cursor's unique position as a company combining API models with domain-specific in-house models.
Building the Automated Researcher
Perhaps most ambitious is OpenAI's announcement that building a fully automated AI researcher will serve as its "north star" for the coming years. Chief Scientist Jakub Pachocki outlined the timeline: by September 2026, OpenAI aims to deploy "an autonomous AI research intern" capable of handling specific research problems independently. By 2028, this will evolve into a full multi-agent system designed to tackle problems too complex for human researchers in fields including mathematics, biology, and policy.
This goal reflects a broader industry recognition of the recursive importance of agentic coding in improving AI systems. As one analysis noted, AI labs have increasingly understood that improving AI requires AI — agentic systems that can code and iterate on themselves represent a key acceleration mechanism.
Competitive Pressures
The timing of these announcements comes amid intensifying competition from Anthropic, which has gained ground in the enterprise market. OpenAI's consolidation strategy and focus on coding tools can be read as direct responses to Anthropic's momentum, following last year's high-profile launches of the Sora video app and the acquisition of Jony Ive's AI hardware startup.
The strategic pivot away from experimental consumer products toward core infrastructure — developer tools, enterprise solutions, and autonomous research capabilities — suggests OpenAI believes sustainable competitive advantage lies in deepening its technical moat rather than expanding into new consumer verticals.