Cloud is awesome! Just hosted an MLFlow server on AWS EC2, with experiment artifacts stored in S3. And then tracked an ML experiment on that server from my Google Colab notebook.
The best use of an AI coding tool like @cursor_ai is to write massive amounts of throw-away code. I can just generate 3k lines of code in less than 5mins, get some ideas and discard the whole thing away without thinking twice. And then actually get to solve the problem.
Upgrading from @nextjs 14 to @nextjs 15 is certainly an experience.
Some open source libraries don't have compatibility with react -19. Obviously can't blame the maintainer but it's unfortunate nonetheless.
Turbopack is a lifesaver for sure. Idk what I was doing without it.
@MLflow integration with @zenml_io is just so nice. The orchestration of ZenML with the experiment tracker of MLFlow - the single source of truth for all your experiments, artifacts, tracking and versioning. Now I wish I should have discovered it sooner.