A new era in AI starts Today. You're not gonna like it.

Summary of the video

• AI pricing changes on July 7, 2026 are framed as a major shift in how people will access frontier AI models. The speaker warns that viewers may not be prepared for the economic impact, describing it as a “boiling the frog” moment—slow, gradual discomfort that still ultimately changes behavior.
• Fable/Mythos is reportedly leaving subscription plans. After the stated date, users won’t be able to use it under the subscription, whether they pay $20/month or significantly more.
• Access is moving to “pay per usage.” The speaker claims it will be about twice the price compared with Opus, likely raising costs for typical users and even more for heavy users.
• Prompting tip (to avoid undesired model behavior):
  • Avoid using words like “security,” “hacking,” or “vulnerability” in prompts, since it may trigger a reversion to Opus.
  • Using alternative wording may help keep the request on Fable and may also “fix” certain security-related issues.
• Broader business pattern: the end of subsidies/subscriptions.
  • The speaker argues people will initially switch to cheaper models (Opus), then gradually move to paying per token when higher capability is needed.
  • Over time, frontier models may become “dumber” or more constrained, including Fable itself, but more slowly than models during the subscription era.
• Rationale: AI companies need profits.
  • The speaker claims providers are not making money or profits, while investors are impatient.
  • Competition pressures include Chinese models and open-source/open-weight offerings.
• Limits of local/open models (in the near term).
  • The speaker suggests local/open models could become strong enough eventually, but doubts they will widely commoditize frontier-level capability soon.
  • If open models truly threaten frontier providers, the speaker implies they may respond with bans/restrictions.
• Impact on real work (Squareblock’s perspective):
  • The speaker and their team at Squareblock evaluate AI based on ROI, reporting about 25% more output/quality when guidance is added to their “Squarebot blueprint.”
  • They say they would respond to unsustainable AI economics by reducing usage or raising client prices if models stop delivering valuable results.
• Local AI is viable but constrained.
  • They describe local models as useful for specific workflows (e.g., large-scale website processing) with heavy guardrails and automation.
  • Local models are positioned as best for narrow tasks, not as a general replacement for API-based frontier models.
• Market-dominance analogy: platform pricing behavior.
  • The speaker compares the shift to a platform pricing strategy (e.g., Uber lowering fares initially, then raising them after saturation), suggesting providers may move toward profitability as adoption deepens.
• Call to action: build before the cost jump.
  • Viewers planning complex AI builds—especially AI coding tools—are urged to do it now, before higher per-usage costs make those projects harder.

Presenters / contributors

• Unspecified speaker (single narrator/author): The video appears to be narrated by one person. No other presenters are named in the provided subtitles.