Descriptions:
Matthew Berman argues that a single infrastructure decision made by Anthropic roughly two years ago is the root cause of its current competitive disadvantage against OpenAI — and that the gap is widening. The decision, attributed to Anthropic’s Dario Amodei, was to avoid betting the company on aggressive compute infrastructure investment in case AI demand growth came in at 5x rather than 10x per year. OpenAI took the opposite approach, and as demand exceeded even optimistic projections, Anthropic now lacks the capacity to match OpenAI’s user-facing generosity.
The video maps out how this plays out practically. OpenAI has repeatedly reset subscription quotas through executive Tibo, temporarily removed rolling usage limits for Plus and Pro subscribers, and banked resets that users can draw on over 30 days. Anthropic, by contrast, has been extending access to its Claude Fable 5 model week-by-week rather than committing it to subscriptions long-term — an approach Berman argues is ceding developer mindshare to OpenAI and GPT-5.6 Soul, which he describes as nearly as capable as Fable at substantially lower cost-per-task.
Berman also quotes an Anthropic executive confirming the capacity constraint directly: Fable will return to subscriptions “as soon as capacity allows.” The analysis frames this not as a quality or talent failure — Anthropic is credited with having the best researchers and highest revenue — but as a structural disadvantage stemming from that one strategic divergence, which Sam Altman is now actively exploiting through public messaging positioning OpenAI as the more developer-friendly option.
📺 Source: Matthew Berman · Published July 15, 2026
🏷️ Format: Opinion Editorial
