Open source

On ToTheMoon, open source is treated less as a licensing label than as a question of strategy: who gains from giving models away, and what the label really permits. It surfaces in talk about DeepSeek, Llama and the Chinese model makers, about what banks can deploy, and about the US–China standoff. As early as 2024 the hosts asked whether open source means open to everyone or comes with strings attached (episode 004).

The early framing was that releasing models openly is a catch-up tactic (episode 013): being cut off from Google once forced China to build its own platforms, and the same learned independence now shows in AI. By early 2025 the assessment had hardened: the best open models are Chinese, and businesses find them the easiest to adopt (episode 046), while banks would rather run a slightly weaker open model inside their own perimeter than send data to an outside vendor (episode 044). Episode 048 walked through the MIT license on DeepSeek R1 and pushed back on the idea that all open models are alike.

From spring 2025 the channel reads openness as a political tool. Open source is often not an ideology but a challenger's strategy, and China may close what it once released, much as OpenAI did once it led; among American companies only Meta was seen as seriously competing in the open (episode 053). Later the hosts noted that OpenAI did ship open models after all, so as not to cede that slice of the market (episode 070). By 2026 the conclusion is calmer: closed labs will hold the frontier, while open models settle in as infrastructure inside corporate systems (episode 150).