AI as infrastructure, not magic. Limits, costs, and fragility.
We separate capability from narrative. While the models are impressive, the systems wrapping them are often fragile, non-deterministic, and economically unproven.
Reality: AI is replacing the learning process for junior developers, creating a "seniority gap" that will cripple organizations in 5 years.
Why the unit economics of generative AI don't scale like traditional SaaS, and what that means for the current subscription pricing models.
Building reliable systems on top of probabilistic components is a solved problem (ask a physicist), but software engineers are trying to solve it with "prompt engineering."