In space travel, there exists the concept of the Wait Equation [0]. It basically means that if you consider radical propulsion innovations and the resulting speed improvements, it may be better to wait before starting your journey. Excited travelers—early adopters—will lose time flying, while their pragmatic friends will wait on Earth and still arrive at the destination faster.
A similar thought experiment is relevant when it comes to adopting AI. Developers who heavily invest in AI too early spend their time working on tools and harnesses. They build complex agent orchestration frameworks whose sole purpose is to work around weaknesses of the current generation of models (I will have another post on how structuring teams of agents after human teams is dumb). They need to tune and tweak their workflows each time a new model is released. They nervously say that "coding is solved," while at the same time producing enormous and buggy codebases that outrun the models' reasoning capabilities [1]. Other, more pragmatic engineers may still use AI, but only to the extent that allows them to work on their products and features without the agentic mental masturbation. They still read their code, write it in small chunks, and steer the models to build proper, human-compatible abstractions. They learn, they build products, and they focus on customers. They are not rewriting stuff in Rust just because it's possible. They still benefit from all improvements in LLMs and coding harnesses, but at the same time, the products they build and the experience they gain are what pull them ahead. Unless your work is to build agentic tooling, you shouldn't.
Working in Big Tech, I believe this thought experiment accurately describes what's happening. Several friends of mine spent all of 2026 working hard on delivering agentic tooling with a half-life of hydrogen-7 while delivering nothing for customers.
I wrote before that you shouldn't be worried about being left behind. Now I've changed my mind. You should be worried about being in front. As William Wallace would say: Hold!
[0] https://en.wikipedia.org/wiki/The_Wait_Equation
[1] https://x.com/Steve_Yegge/status/2094947586373505122?s=20