cinnavun

No paywalls just likely:

1. More funneling people towards cloud compute
2. More aggressive Data creep and telemetry infused into everything because they sell- er i mean "share non-sensitive data" with their partners

• Nvidia already routes contact info through identity-enrichment vendors like 6Sense and Leadspace to append job title, company, and industry, and shares hashed identifiers with ad providers - add that onto HF's user graph and you have the highest-value B2B lead-gen database in existence. Paying $13 billion for a company doing $150 million a year in revenue makes more sense as an acquisition of 18 million developer accounts whose activity constitute a map of who is building what in AI.

3. The slow erosion and deprecation of formats and runtimes that don't require or privilege the hardware they're selling.
   No obvious sabotage needs to happen, the decisions will be be as subtle as llama.cpp ignoring pull-requests that improve compatibility and performance with AVX, while merging PRs that boost performance and reduce friction for CUDA (just a theoretical example)