Mo BitarJune 17, 202613m

Chinese OPEN SOURCE model better than FABLE

TL;DR

GLM 5.2 is presented as a shockingly strong open model: Mo highlights benchmarks claiming the Chinese open source model beats Claude Opus 4.8 and GPT-5.5 on front-end coding while being 82 percent cheaper than Opus 4.8.
Token economics are becoming impossible to ignore: Using his $3,000 comparison, DeepSeek V4 Pro buys about 3.45 billion output tokens, GLM 5.2 buys 682 million, and Claude Fable 5 buys just 60 million.
Cursor's reported sale to SpaceX would be a major strategic hit to Anthropic: Mo says Cursor grew from $100 million to $2 billion ARR in 18 months, has 1 million-plus paying users and 67 percent of the Fortune 500, and once accounted for nearly half of Anthropic's enterprise revenue.
Cursor's CEO imagines a post-code world, but Mo thinks that vision collapses back into code: Michael Truell describes software logic becoming more like editable English or pseudocode, while Mo argues any attempt to make software precise and predictable just recreates programming languages.
The real AI market is being distorted by subsidies: Mo says Anthropic and OpenAI's $200 plans effectively hand out roughly $8,000 and $14,000 of usage, which hides true costs and delays a reckoning on what these models are actually worth.
Tim Ferriss offers a concrete example of AI replacing information products: Ferriss says his nonfiction sales are down sharply and could be 80 percent below 2022 print sales by 2026, because a chatbot is now a faster interface to the same advice his books used to package.

The Breakdown

An open Chinese model, GLM 5.2, is framed as beating GPT-5.5 and Claude Opus 4.8 on front-end coding while costing a fraction as much, and Mo Bitar argues that open source pricing pressure is finally exposing how heavily subsidized the closed-model market has been. He also ties that shift to bigger fallout, from Microsoft's model-swapping strategy to Tim Ferriss saying AI is crushing sales of his once-reliable nonfiction catalog.