Staff engineer · AI products
Michael Nakayama
Eleven years shipping full-stack software, the last few putting AI into products people actually use. I write about building with AI, engineering craft, and the job hunt.
Writing
11 posts- New · AI
System One model: Jev
Typesafe's first System One model, Jev, is built not for chat but for fast, structured decisions. So I tested it.
- AI
Code building AI
Google's AlphaEvolve verifies, runs and scores proposed programs using automated evaluation metrics.
- Job hunting
Who do you know?
The job search is never about how good you are on paper. It's about who you know.
- Job hunting
On interviewing
Most non-coding interviews break into two parts: STAR questions and your chance to ask questions. Here are common ones and my favorites to ask.
- Code style
Some random nits
A few easy wins that can save you time in the long run.
CareerUpdating my resume
I updated my resume, and now we can get some A/B testing done on resume formats.
- ML
Evaluating random forest classifiers
Large language models get the buzz, but they're not the only way to build AI-enhanced products.
RustRust on Arduino Uno
Rust balances ease of use with being a powerful, fast, low-level language. Today I'm going to put that to the test.
- Go
What makes Go special?
Go took the experience of a large-scale organization, Google, and distilled key tenets of good software engineering.
- Go
Foray into Go
AKA how many projects can I work on concurrently? I don't know anyone who doesn't go through phases.
- JavaScript
Yarn resolutions
TIL about yarn resolutions, a simple way to manage the sub-dependencies of a project.