Reading Tanya’s Staff Engineers path1 and also yesterday hearing Take Two CEO interview2 on Gen AI. Strauss Zelnick saying, from his game company perspective, hits are always novel. And Gen AI produces what is derivative. Dont expect to get creativity out of an LLM basically.

This helps to better direct the plasma cutter, the coding harness, to a tighter subset of problems. Or the better analogy is what are the nails, when youre holding the code gen shaped hammer. ( I also discussed this here 3).

And on page 88, of Tanya Reilly’s book, she refers to, Amazon’s “tenet of respect what came before”, to say that, from a problem solving perspective, you often dont need to reinvent the wheel, because most problems arent new.

And her point here is that your output isnt the code, it is the problem solving. I think more broadly , it sounds like she has spent a lot of time in big companies so her point of view is likely about how big companies have lots of mini silos of solutions you can leverage . So far, her book seems to be a lot about how the staff engineer+ roles she describes, help tear down the walls of these silos.

But she may also refer to not reinventing the wheel in gejeral. Like in research, it is expected that you cite prior work first. Prior art.

But her book is more about companies than research I think, referring to faster paced environments, that ship code to production.

This is my first note here on her book, so I should note I find it fascinating that she writes so much about how the staff+ roles deal less and less with code than with connecting people, and aligning perspectives of teams, like with her three maps (chapter 2). But she wrote this in 2021, a few years before the arrival of generative AI taking the world by storm. And introducing generative coding tools , coding harnesses that let you CRUD code through prompting. So many people have stopped writing code or even reading code. They have Tanya’s 50,000 foot view sort of, but without the big picture thinking the same as she writes about it.

Of course all generated code is derivative. If you are solving problems solved before , you dont need to reinvent the wheel. Unlike creating a hit at a video game company. But funny thing is Gen AI does something in the middle. It will reinvent the wheel indeed each time.

References

  1. interview take two ceo, Strauss Zelnick , https://youtu.be/-AgwE_4_slM , reaction , https://x.com/MarioNawfal/status/2056170401831796778

  2. Staff Engineer Path, Tanya Reilly

  3. https://michal.piekarczyk.xyz/note/2026-05-12--merging-is-hard/#the-mundanity-vs-the-novelty

  4. https://michal.piekarczyk.xyz/note/2026-05-14--mine-the-gaps/

  5. https://michal.piekarczyk.xyz/note/2026-05-01-odsc-closing-notes/