Mark Seemann: you can build above your own level of understanding, and that is the trap

ploeh blog (Mark Seemann)·medium signalXBlueskyLinkedInCopy link

In a 2026-09-16 post, Seemann separates doing from learning and argues LLMs accelerate the first without touching the second, since "the bottleneck is hardly the teachers nor the materials, but how fast a human brain can absorb new knowledge." His concrete framing is the risk of having built a system above your own level of understanding, which is fine until something breaks. His recommendations are usable: learn fundamentals systematically if your foundation is weak, deliberately understand the abstraction layer directly below and above where you work, and restrict LLM questions to falsifiable ones where code either runs or does not, because the models will confabulate on open-ended learning guidance.