Grok 3 Did Not End the ChatGPT Era—It Showed That the Race Now Runs Every Day
What did Grok 3 reveal about a race in which model leadership can now change almost every day?
Evaluate Grok 3 not by a momentary lead in one test, but by stability, access, price, and readiness for everyday work.
What to watch for
Key takeaways
The practical meaning of “Elon Musk got into U.S. government database systems” is that this is Elon Musk’s style — demonstrate execution speed and force competitors to react; but the amount of compute alone does not tell us how much better the model is at real work.
In the context of “JD Vance (Trump's right hand) and his speech in Europe on AI,” this criterion applies: Vance talks about the freedom to develop AI and new investment in the U.S. — a signal that the administration is choosing minimal restrictions, while lawmakers simultaneously discuss bans on Chinese models.
The “Chinese open source vs. the US: a question of military superiority?” issue should be assessed with one constraint in mind: the comparison comes down to the task — Grok is strong on fresh information and convenient for an X user, OpenAI on ecosystem, Anthropic on code and long-running work, and open Chinese models on price and local deployment; there is no single winner.
For the “JD Vance will allocate a trillion dollars for AI development” scene, the decisive point is this: Vance speaks of partnership and names eight hundred billion to a trillion dollars — the U.S. invites the world into its technological development and expects to export its own software, but only without restrictions, which fits poorly with Europe's GDPR.
The “DeepSeek banned in the US” issue should be assessed with one constraint in mind: lawmakers propose restricting DeepSeek over data transfers to China, but banning the web service and banning the open model itself are different things that public debate keeps mixing together.
The working conclusion from “Copyright and AI training: what will happen to legal services?” is that Thomson Reuters won the first major AI-training lawsuit — the court found Ross Intelligence unlawfully used protected content; the question now is how to prove who trained on whose data, and what happens to the legal services AI assistants promised to replace.
The working conclusion from “Grok 3: Comparison with other AI” is that models have become very similar in quality and users lack a framework for choosing; on the desired cases — unbiased news analysis, comparing colleges, shopping, and trip planning — the systems still fail: a hand-made product table from Amazon is more useful than their analysis.
The boundary of the “OpenAI response to Grok 3: ChatGPT 5” case is defined by this point: OpenAI said it will not release a standalone o3 and will focus on GPT-5, which consolidates everything — the tool itself will decide when to reply with voice, an image, deep search, or analysis; in effect a new working interface instead of the old chat.
The working conclusion from “Perplexity released Deep Research” is that a longer answer can increase rather than reduce the number of confident errors: the more persuasive a report looks, the harder a hallucination is to notice — ChatGPT already saves time on everyday tasks, but professional analysis requires source verification.
The “Donald Trump's statement: “We must win the electricity race”” scene leads to a working conclusion: Grok 3, Stargate, and new models depend on power and data centers, so the ChatGPT era has not ended — it has turned into an endless infrastructure race in which updates ship faster than users can understand what changed.
What this episode is about
xAI promises to update Grok 3 continuously and is demonstrating an enormous amount of compute. At the same time, the United States is relaxing some restrictions, discussing a DeepSeek ban, and building Stargate. In this race, the model, politics, electricity, and open source can no longer be separated.
The Grok 3 presentation was built around scale: more compute, rapid cluster growth, and a promise to update the model literally every day. That is Elon Musk’s style—demonstrate execution speed and force competitors to react. But the number of computers alone does not tell us how much better the model is at real work.
Comparisons with ChatGPT, Claude, and DeepSeek again come down to the task. Grok may be good at finding current information and convenient for an X user, while OpenAI is strong in ecosystem, Anthropic in coding and long-running work, and open Chinese models in price and local deployment. There is still no single winner.
Politics is intervening directly. JD Vance talks about freedom to develop AI and new investment in the United States, while lawmakers simultaneously propose restricting DeepSeek because data may be transferred to China. Banning a web service and banning the open model itself are different things, but public debate often mixes them together.
Perplexity and OpenAI are developing Deep Research, yet a longer answer can increase rather than reduce the number of confident errors. The more persuasive a report looks, the harder a hallucination is to notice. ChatGPT already saves time on everyday tasks, but professional analysis still requires source verification.
Trump’s phrase about an “electricity race” is more precise here than much of the benchmark debate. Grok 3, Stargate, and new models depend on power and data centers. The ChatGPT era has therefore not ended; it has turned into an endless infrastructure race in which models are updated faster than users can understand what changed.
Grok 3 did not end the ChatGPT era. It confirmed that leadership now changes faster than users can understand the differences, while the race depends increasingly on infrastructure.
Episode transcript
The episode is in Russian; below is an English reading guide to the transcript (the full EN transcript is a machine translation). Voice matching applied to 60 segments: 47 identified, 5 mixed, 7 probable, and 1 unresolved.
Read transcript on a separate page
Loading…