Skip to content
Deep Research · Apple · OpenAIEpisode 050 · 23 March 2025 · 58:14

Chinese Models Are Advancing Through Freedom of Choice, Not the Quality of a Single Answer

Central question

Why are Chinese models expanding their influence through freedom of choice and openness rather than through one best answer?

What you take away

Compare the strategy around Deep Research and OpenAI across the whole technology chain and the conditions of market access. The decision requires the reader to account for export controls, jurisdiction, access to infrastructure, and the independence of the ecosystem.

Main threads

What to watch for

1Compare “New Chinese Baidu model is cheaper than DeepSeek” with “Corporate espionage: a court between US companies”: they provide different criteria for judging the same issue.
2Test the conclusion from “Perplexity and competition with OpenAI” in your own use case—what actually changes in the process and what remains a promise.
3Before choosing a product or approach, record the constraint identified in “Retraction from Limitless, Jusikes”.
4Define the owner of the outcome and the quality metric for the situation described in “Apple button collapse”.
Signals to track afterwards
Watch for actions by Apple and Baidu that confirm or challenge the episode’s central claims.
Compare new launches and policy changes with “Perplexity and competition with OpenAI”: have access, quality, price, or constraints changed?
Check whether the scenario in “Apple button collapse” becomes repeatable practice rather than a one-off demonstration.
Most useful for
InvestorsEntrepreneursExecutives and managersStrategy teamsAI usersProduct teams

Key takeaways

01:35The most frustrating moment in paid AI is when the requests run out

The boundary of the “New Chinese Baidu model is cheaper than DeepSeek” case is defined by this point: this section clarifies the mechanism behind the topic and preserves a constraint that would otherwise be lost in an overly simple conclusion.

02:59What determines the outcome: personal experience with Deep Research

The “Personal experience with Deep Research” scene leads to a working conclusion: the important signal is not one funding number: the next round, available runway, and closure rate show whether a company can survive the new cost of capital.

06:17American companies respond by talking about security and data leakage to China

The “Corporate espionage: a court between US companies” scene leads to a working conclusion: the key test is enforceability: who must prevent the risk, who records the violation, and who is accountable for the consequences.

08:09The market tests it through use: deep Research in the Work: Real Cates

The boundary of the “Deep Research in the Work: Real Cates” case is defined by this point: a benchmark measures a narrow capability; working value requires repeatability, a clear cost, and control over errors.

10:28The boundary between value and constraint: a negative experience with Deep Research

In the context of “A negative experience with Deep Research,” this criterion applies: the issue turns on whether the rule can be enforced and who carries responsibility, not merely on the existence of a new requirement.

13:03Who owns the outcome: chatGPT for $20 and $200

The “ChatGPT for $20 and $200” topic becomes clearer once this point is included: the important signal is not one funding number: the next round, available runway, and closure rate show whether a company can survive the new cost of capital.

18:26What changes in real work: should we pay $200 for Deep Research's signature

For the “Should we pay $200 for Deep Research's signature?” scene, the decisive point is this: the issue turns on whether the rule can be enforced and who carries responsibility, not merely on the existence of a new requirement.

19:48Deep Research shows OpenAI’s value when a large market overview is needed

The “Perplexity and competition with OpenAI” topic becomes clearer once this point is included: this section clarifies the mechanism behind the topic and preserves a constraint that would otherwise be lost in an overly simple conclusion.

38:52Limitless adds memory to the competition

The practical meaning of “Retraction from Limitless, Jusikes” is that this section clarifies the mechanism behind the topic and preserves a constraint that would otherwise be lost in an overly simple conclusion.

52:15ChatGPT will not disappear because of one Chinese “monster.” It can gradually lose parts of the workflow: local models will take private data, Perplexity will take search, Google and Apple will take system functions…

The “Apple button collapse” issue should be assessed with one constraint in mind: this section clarifies the mechanism behind the topic and preserves a constraint that would otherwise be lost in an overly simple conclusion.

What this episode is about

DeepSeek, Gemma 3, Perplexity, and local models are forcing users to choose among quality, price, privacy, and limits. OpenAI and Anthropic are responding with regulation and expensive modes, while wearable devices such as Limitless collect ever more personal data. The end of ChatGPT’s monopoly begins not with one killer, but with dozens of alternatives.

The most frustrating moment in paid AI is when the requests run out. The user has already built the model into a workflow, and the service suddenly says to wait or buy another plan. Chinese and open models attack precisely that dependence: they can be deployed more cheaply, run locally, and used without accepting the rules of a single provider.

American companies respond by talking about security and data leakage to China. The risk is real, especially for corporate information. But it does not exist only with Chinese services. Inside Silicon Valley itself, engineers move among companies, startups sue over corporate espionage, and knowledge continually travels with people.

Deep Research shows OpenAI’s value when a large market overview is needed. On a simple local search, however, the result may be no better than Google or Yelp. Gemma 3 and other Google models expand the choice, and Perplexity lets users work with several sources, but no system wins every task.

Limitless adds memory to the competition. The pendant records conversations and lets a user ask what was discussed yesterday or at a morning meeting. The utility is obvious, but the audio is continuously sent over the internet, and the people nearby may not know they are being recorded. Ray-Ban glasses with a camera create the same conflict.

ChatGPT will not disappear because of one Chinese “monster.” It can gradually lose parts of the workflow: local models will take private data, Perplexity will take search, Google and Apple will take system functions, and specialized agents will take work. The user wins if providers can be changed without losing memory and process. Portability, not another benchmark, will become the real sign of a mature market.

Portability will define a mature market: users should be able to change providers without losing memory, data, or workflow. A single benchmark cannot provide that freedom.

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 113 segments: 61 identified, 8 mixed, 27 marked with ✓, and 17 unresolved.

Loading…