Daniel Olshansky | olshansky.eth π¦π
@olshansky
CTO @BuildWithGrove βοΈ @POKTNetwork Ex @waymo @magicleap @twitter @google NΞ¨ 1T4 @UofT π¨βπ»π§βππ€Έπ₯Ύπ²π§ βοΈπππΊπ
ChatGPT & Copilot: Concept & Context
#ChatGPT & @GitHubCopilot have, by far, been the most significant innovations to developer (et al) productivity I've seen and exceeded my prior imagination of what's possible.
Copilot is contextual while ChatGPT is conceptual. When the two merge, it'll be a game changer.
π οΈ Tooling
I come across tons of tools daily that can make my work more productive, but there's not enough time to explore everything.
Getting actionable tool suggestions will be a boon to productivity.
E.g. You're working on optimization X, consider profiling it using Y.
𧽠π¦ Rubber Duck
I'm already directing more questions to ChatGPT than any other single human, but it doesn't ask questions in return unless I prompt it to.
E.g. You've been stuck on this for a while... Let's take a step back and re-evaluate what we're trying to accomplish.
ββπ» πΈ Tech debt
The balance of identifying, documenting and resolving tech debt is an art more than a science. There's a hidden data-driven opportunity.
E.g. You and your team have spent a lot of time changing/reading this part of the code. Consider refactoring like so...
π§βπ Onboarding
Introducing a new engineer to a codebase is an ever-changing process based on the stage/maturity of the project and seniority of the individual. Having contextual understanding is living documentation
E.g. Your task is X; look at Y and ask if you need any help.
β³πππ Long-term
I'm sure @OpenAI is looking at a BYOE (Bring Your Own Embedding) business model that involves transfer learning to customize ChatGPT to your data. This is an extremely difficult task that, in my opinion, is 5+ years out.
But...
π was a couple of decades too early, but it's time to make comeback!