OpenAI's Code Red, Sacks vs New York Times, New Poverty Line?

The episode opens with Sam Altman's "code red" memo. According to the hosts, he sent it on a Monday and told OpenAI employees to stop working on "side quests" such as ads and to focus on the core ChatGPT experience by making it faster and better. The host who raises the topic explains the move by pointing to rivals: Gemini, Grok and Anthropic's Claude "have been crushing it." The same host calls ChatGPT 5 "a bit of a flop" that "didn't perform to expectations."

The same host adds that Anthropic has been beating OpenAI in enterprise revenue since the summer. They also say they see more startups choosing the Anthropic and Google Gemini APIs because those startups "don't essentially trust OpenAI to not steal their business." Two charts are mentioned. The first, based on July data, shows most of Anthropic's revenue coming from businesses and most of OpenAI's from consumers. The second compares OpenAI's new infrastructure deals with its revenue for 2025 alone. Code red is described as a term from the industry for the moment when "everybody reports to the office and gets focused on one thing."

One panelist splits the question into a strategic part and a tactical part. On strategy, they argue that it is "too difficult and too early to pick winners" in AI except at the silicon layer. There, they say, the die has largely been cast, with roughly NVIDIA, AMD, Google and a group of inference-chip makers, although they expect more competition to come.

For the model layer above the chips, the panelist compares the situation to Facebook's early competition with Myspace. Myspace was "an order of magnitude bigger": it had more than 100 million users while Facebook had "sort of 15." The panelist says Facebook's six-person senior team still believed its product was fundamentally better and expected to win eventually. In the panelist's view, AI models are in a similar early stage. Consumer expectations are still being formed, and as markets converge, "distribution still matters a ton, which favors Google." They add that distribution also favors Meta, although Meta is "quite behind," and favors OpenAI, which they say has 800 million monthly active users.

On tactics, the panelist suggests narrowing the company's focus and trying "to make a crisis out of every opportunity." Their reasoning is that companies tend to keep hiring at every level until "there's thousands of people that you didn't know even six months ago." They give Google as an example. The panelist says Google used the controversy over its depiction of a "black George Washington" as a rallying cry to streamline the organization and put its best people on the most important work. They credit this with Google's current "incredible overperformance" and call Gemini "incredible."

It's a three or four horse race, and Sam needs to batten down the hatches.

The conversation then treats the code red as a management tool. One host mentions that Sergey Brin called "a personal code red" and returned to the office because losing the search franchise would be existential for Google. The host notes that Google has not lost it and concludes, "The Code Red works." A question is then put to David Freiberg, described as having been at Google, about whether Google used code reds back when it had little competition. The host calls the current contest "the most competitive thing we've ever seen in our lifetimes, I think."

The reply describes an earlier Google effort. After Microsoft entered search, Google set up a code-red-style program called "Project Canada," a code word for Microsoft. It included a weekly war-room meeting, and strategy and product decisions were organized around the threat from Microsoft, a well-funded company with strong engineering talent. One concrete step was opening an office in Kirkland to recruit engineers in the Seattle area. The speaker links this kind of effort to the US race to the moon against Russia and to today's competition with China in technology and AI.

Having an impending threat is a very strong motivational tactic. It is a very focusing setting, and it drives innovation.

The same discussion covers market share. ChatGPT is described as having been basically a monopoly in LLM chat interfaces, with "only had one way to go, which was down." The figures given are a little loose. A little over a year ago, OpenAI had "90 plus market share in generative AI traffic." Google is said to have "roughly 14%" and, a moment later, Gemini is put at 15%. The speaker credits the "depth, the flywheel" that Google has built with giving it an "extraordinary advantage."

A host jokingly addresses David Sacks as "the czar of AI and crypto" and asks how the competitive field looks from Washington. The answer begins by praising Altman for calling the code red. The speaker argues that CEOs find it easy "to engage in happy talk and ignore problems," especially when talking about the problem will produce bad press. They say Silicon Valley founders treat their companies' situations as more existential than most Fortune 500 CEOs, who are "just kinda managing PR all the time."

The speaker then describes five major companies, each with its own strengths:

ChatGPT leads in consumer use, with "something like 80% market share in consumer use of LLM chatbots." Google was gaining share with Gemini 3, helped by its integration into Google Search, where people find it easily. The speaker gives this as the reason for the code red. Anthropic: the speaker dislikes what they call its "regulatory capture strategy" but acknowledges that its products are very good. They say it is widely seen as having the best coding assistant and is carving out a lucrative enterprise niche. xAI is called the best at current events because it is integrated with X. Elon Musk is described as scaling training clusters fastest, moving from Colossus 1 to Colossus 2, which the speaker says bodes well for Grok 5 because it will be trained on the largest cluster of Blackwells. Meta "sort of hit some headwinds," but the speaker expects it to come back because of its heavy AI investment and deep pockets.

The speaker says these companies keep "leapfrogging each other" with each new release, and another participant mentions leaderboards where Grok and Gemini trade places. The speaker also sees specialization, with companies building strengths in particular areas. They call this a "Goldilocks scenario": fast progress while the market stays competitive. They hope it does not consolidate into a single monopoly, as has happened in other tech markets, because a dominant company would have "too much power and control," which is generally bad for consumers and citizens.