In the 2010S, Technical Gaps Meant Relying on Colleagues or Hoping for Online Answers

LL Ms reward expertise. In the 2010s, technical gaps meant relying on colleagues or hoping for online answers. Today, LL Ms make everyone a generalist, but many assume prompting requires no skill.

That's wrong. The key skill is domain expertise. Terence Tao's ChatGPT conversation about the Jacobian Conjecture shows this: his mathematical understanding lets him steer the model, ask specific questions, and identify flaws.

Similarly, knowing your codebase lets you push an LLM further than generic principles would. Without domain knowledge, you can still get something from an LLM, but with it, you extract far more value. Human expertise remains crucial as models improve; the bottleneck is communicating the desired solution, not the model's capability.