BEFORE YOU PRESS PLAY
What is discussed
AskMonty AI editorial noteSource consulted:
Montemagno begins by saying that he used an Anthropic model identified in the episode as “Fable 5” the previous evening, then found it unavailable the following morning. He attributes the change to a US government intervention on national-security grounds, which he describes as restricting foreign users. The episode therefore starts with his personal experience of losing access and his interpretation of the political and economic context. He stresses the fragility of being able to work with a tool one day and losing it the next because of a decision outside the user's direct control. His assessment distinguishes increased capability from infallibility. He says the model helped solve problems that other systems had failed to resolve for months, while acknowledging that it still made mistakes. He describes it as a reduced version of a model called “Mythos” and says its inclusion in the subscription had already been scheduled to end on 22 June. Those details are part of the episode's account. His main evaluation is that better models can materially expand what he can accomplish without becoming magical or removing the need to recognise errors and limitations. Montemagno questions how a company making enormous investments could sustain its business if it were allowed to sell only to a sharply limited market. He points to infrastructure and computing costs to explain why restricting revenue would, in his view, create a contradiction. He also asks how nationality-based restrictions would affect people involved in developing the very systems they could no longer use. These arguments lead him to expect that the measure he describes will be temporary. However, he does not claim to know the government's next decision; his conclusion is an economic interpretation rather than a confirmed account of what will happen. The episode also discusses open models, particularly examples from China. Montemagno observes that users can install such systems on their own computers or servers and that their capabilities are improving quickly. In his own experiments, however, he has not achieved the same results as with the most advanced proprietary models he uses. His suggestion that alternatives could catch up within a few months is a forecast, not an established timetable. He adds a criticism of Anthropic's communication strategy, arguing that repeatedly emphasising AI's dangers may encourage regulatory intervention that the company itself is then unable to control. The broader practical issue is how people organise their work around AI. Montemagno argues that the answer is better understanding of the tools and a personal assistant capable of using different models. The assistant should provide continuity under the user's control, while individual suppliers may change their terms, availability or capabilities. This episode explains why he considers that goal important; it does not provide a complete technical implementation. Dependence on one company is the vulnerability he wants the reported incident to expose. Knowing how the surrounding tools fit together becomes part of maintaining the ability to continue working. He closes by connecting individual dependence with competition between countries. As models become more capable, the computing capacity and infrastructure behind them may acquire strategic importance, potentially requiring protection comparable to other crucial national assets. His final message has two parts: AI can genuinely expand a person's practical possibilities, but access to those possibilities also depends on corporate and political decisions. Learning about the tools and preserving the ability to choose between them are therefore, in his account, important elements of personal autonomy in an environment that can change abruptly. The episode also promotes his own two-day workshop on AI agents.
Key points
- The restriction is the speaker's account: Montemagno says he lost access to a model called “Fable 5” and attributes this to a US national-security measure affecting foreign users.
- Positive assessment without claims of perfection: he says the model solved difficult problems for him but still made mistakes.
- He questions the economics of restricted access: narrowing the market appears, in his view, incompatible with the investment needed to develop these systems.
- Open models are improving alternatives: he finds them weaker in his own use cases but predicts rapid progress; this is a personal forecast.
- Risk communication and regulation interact: he criticises Anthropic's relationship with government and what he sees as fear-based messaging.
- Reducing dependence is the practical lesson: a personal assistant should be able to use different models; powerful AI is framed as a strategic national asset.
