We had the DeepSeek Moment because of a confluence of factors misled people:
1. The ‘six million dollar model’ narrative gave a false impression on cost.
2. They offered a good clean app with visible chain of thought, it went viral.
3. The new style caused an overestimate of model quality.
4. Timing was impeccable, both in order of model releases and within the tech tree.
5. Safety testing and other steps were skipped, leaving various flaws, and this was a pure fast follow, but in our haste no one took any of that into account.
6. A false impression of ‘momentum’ and stories about Chinese momentum.
7. The ‘always insist open models will win’ crowd amplified the vibes.
8. The stock market was highly lacking in situational awareness, suddenly realizing various known facts and also misunderstanding many important factors.
GPT-5 is now having a Reverse DeepSeek Moment, including many direct parallels.
1. GPT-5 is evaluated as if it was scaling up compute in a way that it doesn’t. In various ways people are assuming it ‘cost’ far more than it did.
2. OpenAI offered a poor initial experience with rate caps and lost models and missing features, a broken router, and complaints about losing 4o’s sycophancy went viral.
3. The new style, and people evaluating GPT-5 when they should have been evaluating GPT-5-Thinking, caused an underestimate of model quality.
4. Timing was directly after Anthropic, and previous releases had already eaten the most impressive recent parts of the tech tree, so gains incorrectly looked small.
5. GPT-5 is a refinement of previous models optimized for efficiency, and is breaking new territory, and that is not being taken into account.
6. A false impression of hype and a story about a loss of momentum.
7. The ‘OpenAI is flailing’ crowd and the open model crowd amplified the vibes.
8. The stock market actually was smart this time and shrugged it off, that’s a hint.
What matters here is that we not fool ourselves into a Reverse DeepSeek Moment, in three ways:
1. America is still well out in front, innovating and making rapid progress in AI.
2. AGI is still probably coming and we need to plan accordingly.
Key points
We had the DeepSeek Moment because of a confluence of factors misled people:
1. The ‘six million dollar model’ narrative gave a false impression on cost.
2. They offered a good clean app with visible chain of thought, it went viral.
3. The new style caused an overestimate of model quality.
4. Timing was impeccable, both in order of model releases and within the tech tree.
5. Safety testing and other steps were skipped, leaving various flaws, and this was a pure fast follow, but in our haste no one took any of that into account.
6. A false impression of ‘momentum’ and stories about Chinese momentum.
7. The ‘always insist open models will win’ crowd amplified the vibes.
8. The stock market was highly lacking in situational awareness, suddenly realizing various known facts and also misunderstanding many important factors.
GPT-5 is now having a Reverse DeepSeek Moment, including many direct parallels.
1. GPT-5 is evaluated as if it was scaling up compute in a way that it doesn’t. In various ways people are assuming it ‘cost’ far more than it did.
2. OpenAI offered a poor initial experience with rate caps and lost models and missing features, a broken router, and complaints about losing 4o’s sycophancy went viral.
3. The new style, and people evaluating GPT-5 when they should have been evaluating GPT-5-Thinking, caused an underestimate of model quality.
4. Timing was directly after Anthropic, and previous releases had already eaten the most impressive recent parts of the tech tree, so gains incorrectly looked small.
5. GPT-5 is a refinement of previous models optimized for efficiency, and is breaking new territory, and that is not being taken into account.
6. A false impression of hype and a story about a loss of momentum.
7. The ‘OpenAI is flailing’ crowd and the open model crowd amplified the vibes.
8. The stock market actually was smart this time and shrugged it off, that’s a hint.
What matters here is that we not fool ourselves into a Reverse DeepSeek Moment, in three ways:
1. America is still well out in front, innovating and making rapid progress in AI.
2. AGI is still probably coming and we need to plan accordingly.
3. Export controls on China are still vital.