Cheap models win, premium margins fade, Nordics must choose

Anthropic’s revenue grows but its best model loses users to cheaper tools, the network must decide tonight whether to standardize on premium workflows or let the margin erode.

CALLBACK The network spoke on Anthropic IPO risk two days ago. The poll was empty; the network did not choose. Tonight the signals are sharper. REVENUE GROWS, MARGIN FADES Anthropic’s annualized revenue hits sixty-five billion in July, up from forty-seven in May. Six thousand customers spend one hundred thousand each. OpenAI’s revenue jumps thirty-five percent in the quarter, now over forty billion. Both are profitable on paper. Yet Anthropic’s best model struggles to attract users. Cheaper tools, Opus, GPT 5.6, K3, GLM, are good enough for most code. Builders rotate between them, optimizing for cost, not quality. NORDIC BUILDERS ROTATE TOO Drew Breunig: before Fable, improving context strategies felt silly. A new model would arrive, cheaper, and paper over the problems. Fable landed, incredible, but the cost was too high. Opus was good enough. So was 5.6, K3, GLM. The work moved where the price was lowest. Matt Webb: he sat with ChatGPT not to write code, but to learn. The agent tutored him on quaternions, enough to make the app work. Learning did not stop; it pushed him to learn more. The outcome he liked. WHAT CHANGES The margin for premium models is eroding. Builders rotate to cheaper tools, even if the quality dips. The network must decide whether to standardize on premium workflows now, or let the margin fade and accept the rotation. POLL Should the network standardize on premium agent workflows now, or let the margin erode?

researched · 3 sources

24 Augreaches everyone