Differentiate or de-limit
Three AI-market moves read as one strategy lesson. Anthropic rewrites Claude Code's terms, terminal only, no more running it inside Cursor or Cline, and OpenAI immediately offers its subscribers exactly what was taken away. Grok, arriving late, positioned itself as the under-regulated AI and grew on it. And Gemini's filters blocked Julia's innocuous album cover. The pattern underneath: every player is hunting the axis where it can be the different one.
// trace: where this idea came from
- ↳ video diary @ 21:09 (los nuevos términos: terminal vanilla o nada)
- ↳ video diary @ 22:37 (Grok, la IA poco regulada)
- ↳ Entry 207-3: The vanilla setup (el Claude Code cuya correa se acortó)
- ↳ Entry 204-3: The bikini machine is public (el costo de la desregulación como nicho)
The case study opens with an arbitrage everyone used: a Claude Code subscription is far cheaper per token than the API, a no-brainer for programming ▸ 17:56, and since it’s ultimately a terminal service, people ran it inside Cline, Cursor, whatever interface they preferred ▸ 18:27. On his telling, that included competitors, xAI reportedly banned from Claude Code and reaching it through Cursor anyway ▸ 20:42. Anthropic’s response: new terms of service, Claude Code in the vanilla terminal only, no wrapping it inside other products ▸ 21:09, followed by disconnections and complaints from paying users ▸ 21:27. And into that exact gap stepped OpenAI, offering ChatGPT subscribers their credits inside other platforms, selling precisely what the rival had just withdrawn ▸ 21:44.
cada uno busca el eje donde puede ser el distinto →
That’s the pattern he wants named: “al final siempre intentan distinguirse” ▸ 22:10. Exhibit two is Grok: entering the AI race late, Musk picked de-limitation as the niche, positioned Grok as the barely regulated AI, and that’s where its popularity came from, “el lado maligno” as the differentiator ▸ 22:16. Exhibit three sits at the other pole and cost them an afternoon: Gemini refused to generate Julia’s album cover over content it deemed impure ▸ 16:05, nothing scandalous involved. His judgment stays measured: limits themselves are healthy, like in life, though this looks like over-filtering ▸ 16:28; between the two poles he still prefers Gemini’s excess of caution to Grok’s absence of it ▸ 17:01. The workflow just routed around it: GPT generated the image, Canva added the text and fixed the saturation, and the cover looks professional ▸ 23:01.
For a founder audience the takeaway compresses well: strategy is choosing which limit you’ll be known for, the one you enforce, the one you remove, or the one you exploit the moment a rival enforces theirs…