The AI Rug Pull
Frontier AI companies are currently operating at a significant loss per user, effectively using paying customers as training data during what the author calls an 'apprenticeship' phase. Three key forces—synthetic data, agentic self-play, and diminishing returns to human feedback—are expected to end this phase within three to five years. When the apprenticeship ends, consumer access to top-tier AI capabilities will likely be restricted, with premium features reserved for enterprise clients and casual users excluded.
- ▪Frontier AI providers sell their services at a 4–7x loss per user because human interactions are used to train models, not because users are the primary customers.
- ▪The apprenticeship phase is expected to end in three to five years due to synthetic data, agentic self-play, and saturating returns to reinforcement learning from human feedback.
- ▪When the phase ends, the $20 consumer tier is likely to disappear, with top capabilities gated behind enterprise contracts and labs taking over operational roles.
- ▪Enterprises and operators of open-weight models are best positioned to survive the transition, while casual users and small operators relying on subsidized access will be negatively impacted.
- ▪Users are advised to build for portability by using closed models for paid work, incorporating open-weight fallbacks, and avoiding mission-critical reliance on free AI tiers.
Opening excerpt (first ~120 words) tap to expand
Essay · AIThe ApprenticeshipFrontier AI is sold at a structural loss because users are still teaching the models. Three predictions for what happens when the apprenticeship ends — and who gets locked out of the workshop afterward.By Shaun WarmanMonday, April 27, 20269 min readTL;DR — TakeawaysFrontier AI is sold at a 4–7x loss per user because the human is the training set, not the customer.Three forces — synthetic data, agentic self-play, and saturating returns to RLHF — are closing the apprenticeship window in three to five years.When it closes, expect the $20 tier to vanish, top capabilities to gate behind enterprise contracts, and the labs themselves to step in as operators.Enterprises and owners of open-weight model capacity survive cleanly.
…
Excerpt limited to ~120 words for fair-use compliance. The full article is at Warman.
Discussion
0 commentsMore from Warman
