Fable model's high cost prompts rethinking of AI coding harnesses

Simon Willison’s weblog carries a quoted passage from Drew Breunig, writing on his own site on August 23, addressing the arrival of a new coding model called Fable.

Breunig writes that before Fable, investing heavily in refining a coding harness or context strategy seemed a poor use of time, since a new model would typically arrive at the same price, or cheaper, and cover over most existing shortcomings. Fable, in Breunig’s account, changed that assumption. He describes it as incredible in capability, but says its cost was high enough that other models, Opus, and what he refers to as 5.6, K3, and GLM, remained good enough for most coding tasks despite being less capable than Fable itself.

The excerpt, as quoted on Willison’s site, breaks off at this point, without further elaboration on what conclusion Breunig draws about harness design or context strategy going forward.

Willison’s post frames the quotation as commentary on the economics of AI coding tools, noting the pattern Breunig identifies, that pricing and capability trade-offs are prompting renewed attention to the surrounding tooling rather than reliance on the next model release to resolve inefficiencies.

No further detail on Fable’s pricing, release date, or publisher is given in the material provided.