One new card from Drew Breunig. The Aug 17-23 window was covered in depth by six consecutive daily runs. This scan found nothing that cleared the bar outside Breunig's August 23 post, which fell outside the prior coverage window and carries a clear mechanism.
The end of passive model improvement
Every prior frontier model release arrived cheaper and more capable than the one before. Developers who had not optimized their workflows still improved automatically. Fable broke the pattern: more capable, more expensive. The free optimization cycle stopped.
Breunig names the consequence precisely. Before Fable, it made no economic sense to invest heavily in harness or context tooling, because the next release would paper over the inefficiency anyway. After Fable, harness quality is the only reliable path to lower per-task cost.
His workflow splits tasks by what the model constraint actually is. Fable handles design and architecture: reasoning quality is the constraint, cost is justified by decision value. GLM handles execution: context quality is the constraint, and GLM 5.2 costs roughly one-ninth of Fable's price. A well-structured brief routes most rote coding to the cheaper model.
"Prior to Fable, it felt silly to waste _too_ much time improving your coding harness or context strategies."
"I frequently chat with Fable to interrogate and shape a design, before handing off a brief to GLM"
The broader pattern connects to what Tunguz found in the intelligence-per-watt research (Routing everyday queries to local models cuts energy 80%, compute 77%, and cost 74% while matching frontier quality on 89% of tasks) and in his local model benchmark (Small local models match cloud model quality by reasoning from first principles rather than retrieving from memory): the default should be the cheapest model that meets quality requirements, not the best available model. Breunig adds the forcing function: frontier premium pricing turns this from a best practice into a financial necessity.
- Drew Breunig, When frontier model pricing rises rather than falls, harness quality becomes the primary investment. Fable's premium pricing ends the passive benefit of frontier model releases; teams must route tasks across model tiers, and harness quality determines how much cheaper models can absorb.