The news in brief

OpenAI has hit the big red button. In an internal memo sent on December 1, 2025, CEO Sam Altman declared a “code red” and ordered teams to pause or slow adjacent projects so the company can refocus on the core ChatGPT experience—speed, reliability, and personalization—amid stiff new competition from Google’s just-launched Gemini 3. Multiple outlets confirmed the memo and the reprioritization, including reports that ad experiments, shopping and health agents, and the proactive assistant Pulse will be delayed while ChatGPT gets a rapid polish. AP, Financial Times, Reuters Breakingviews, MacRumors.

An urgent, late-night product war room with red dashboards labeled ‘Code Red’, ChatGPT UI elements on big screens, and engineers reprioritizing tasks.

What triggered the urgency

  • Google introduced Gemini 3 on November 18, 2025, calling it its “most intelligent” model and pushing it into Search’s AI Mode on day one, plus the Gemini app, Vertex AI and AI Studio. Early independent testing has shown strong reasoning and multimodal scores. Google blog, Google Search blog, Simon Willison.
  • High-profile users amplified the momentum. Salesforce CEO Marc Benioff publicly said he’s switching to Gemini 3 after years of daily ChatGPT use. Fortune.
  • OpenAI’s leadership acknowledged the moment. In the memo, Altman urged a “surge” on ChatGPT’s day‑to‑day experience; ChatGPT head Nick Turley posted that the focus now is making ChatGPT “more intuitive and personal,” and called out search as a major opportunity. AP, Nick Turley on LinkedIn.
800M
ChatGPT weekly usersSource: TechCrunch, 2025-10-06
650M
Gemini app monthly usersSource: Google blog, 2025-11-18

What OpenAI is changing right now

OpenAI’s near‑term roadmap shifts (December 2025)

AreaWhat’s prioritizedWhat’s delayed/paused
ChatGPT core UXFaster responses, higher reliability, better personalization, broader question coverage
Ads/monetizationAdvertising experiments in ChatGPT (multiple reports of in‑app ad tests now on backburner)
AgentsAI agents for shopping and health tasks
Proactive assistantPulse (personalized morning briefs)
Org/processDaily coordination, temporary team reassignments to accelerate ChatGPT improvements

Sources: AP, FT, Reuters Breakingviews, MacRumors. On Pulse: OpenAI blog.

The strategic backdrop: search, scale, and spend

  • Search is in play. Turley says search is one of ChatGPT’s biggest opportunities; he also claimed ChatGPT accounts for “roughly 10% of search activity,” though he didn’t define the denominator. Regardless, OpenAI now has a browser—ChatGPT Atlas—to bring assistant workflows deeper into the web. Nick Turley on LinkedIn, OpenAI: Atlas.
  • Google’s distribution advantage. Gemini 3 landed in Google Search’s AI Mode on day one and ships across the Gemini app and enterprise tooling. That puts a capable model in front of billions of search journeys and millions of developers with minimal friction. Google Search blog, Google blog.
  • Economics and infrastructure pressures. OpenAI is valued around $500B after a fall secondary sale but remains unprofitable and has taken on more than $1T in long‑term cloud and chip obligations as it races to build out compute. Its Stargate buildout with Oracle points to multi‑hundred‑billion dollar data‑center investments. CNBC, AP, Reuters: Stargate.

A quick look at Gemini 3’s claims

Google positions Gemini 3 as “state‑of‑the‑art in reasoning,” with better depth and multimodal performance, plus a forthcoming Deep Think mode for tougher problems. Early, independent spot‑checks (e.g., benchmark summaries collated by developer Simon Willison) generally corroborate improved scores versus prior Gemini models and strong standing against peers—though results vary by task and tool use. Google blog, Simon Willison.

Split-screen, stylized comparison: on the left a ChatGPT interface focused on speed and personalization toggles; on the right a Gemini 3 panel showing Search AI Mode and benchmark badges.

What this means for teams using AI today

  • Expect rapid, user‑visible ChatGPT changes. Improvements in latency, stability and personalization are the stated goals. If your workflows depend on specific model behaviors, build lightweight evaluation checks into your pipelines so you can detect regressions fast. AP.
  • Plan for tool pluralism. Gemini 3’s strong early showing and deep Search integration make it a credible “daily driver” for many tasks. Keep an “adapter layer” in your stack so you can route by task, data policy, or cost across vendors.
  • Watch monetization experiments. Reporting in late November surfaced ad hooks in ChatGPT’s Android app; OpenAI has since deprioritized ads to focus on quality, but the business need hasn’t gone away. If/when ads return, expect differences between free and paid tiers. MacRumors.
TipBuilder checklist
  • Add vendor‑agnostic interfaces for prompts, tools and memory.
  • Track latency and refusal rates per provider, not just accuracy.
  • Maintain a human‑in‑the‑loop for high‑stakes automations.
  • Budget for compute volatility; keep an alternative inference path hot.

What to watch next

  1. OpenAI’s next model update. Reports suggest a near‑term “reasoning” model is queued up; if it lands, compare long‑horizon tasks (coding, planning) rather than single‑shot Q&A. MacRumors.
  2. Gemini 3 “Deep Think.” Google says it will roll out after additional safety evaluations. Evaluate trade‑offs between speed and quality for your use cases. Google blog.
  3. Search experience shifts. As ChatGPT leans into search and Google doubles down on AI Mode, the default “learn something / find something” flow is changing. Expect new ad formats and SEO dynamics.
  4. Compute buildouts. Stargate-scale spend will shape availability, pricing, and regional footprints for training and inference. Reuters.

Sources

  • Associated Press: OpenAI CEO declares ‘code red’ to improve ChatGPT (Dec 2, 2025) — confirms memo, focus areas, and delayed projects, plus >$1T in cloud/chip obligations. Read
  • Financial Times: Altman declares “code red,” delays ads/agents/Pulse. Read
  • Reuters Breakingviews commentary on the “panic button” and the need for actual reprioritization. Read
  • Google: Introducing Gemini 3 (model overview) and Search AI Mode rollout. Read, Read
  • Simon Willison: Early benchmark spot‑checks on Gemini 3. Read
  • Fortune: Marc Benioff says he’s “not going back” to ChatGPT after trying Gemini 3. Read
  • OpenAI: Introducing ChatGPT Atlas (browser) and ChatGPT Pulse (proactive briefs). Read, Read
  • CNBC: OpenAI secondary sale at ~$500B valuation. Read
  • Reuters: Stargate data‑center expansion plans with Oracle. Read