Jason Fleetwood-Boldt is a software engineer and agentic engineer based in New York City. For two decades he’s helped startups and tech ventures ship product across Rails, React, TypeScript, GraphQL, and NestJS — and the infrastructure decisions that quietly decide whether a company survives its second year.
In January 2026 he pivoted into agentic engineering, not as a trend but as a thesis: the job has shifted from syntax to structure. AI handles the typing now. What it can’t do is hold a system in its head, anticipate how an architectural choice plays out in 3, 6, or 12 months, or catch the patterns that quietly destroy companies. That judgment is what Jason brings to a project.
On the side, he’s building Helios Flow — an AI-native marketing automation platform, part experiment, part product, and the thing he’s proudest of right now. It’s where the agentic engineering thesis meets real-world use: email tooling that adapts to your voice instead of making you adapt to it. If you need marketing automation that doesn’t feel like software from 2015, it’s worth a look.
His thesis on software in the age of AI:
“Code quality” was always a proxy. What matters is architectural quality — whether the seams are in the right place, whether the data model survives the fifth feature, whether the security model assumes the right adversary. AI now produces code that’s lint-clean, idiomatic, and architecturally catastrophic. The proxy and the thing it measured have finally divorced.
What’s left is the work that was always the actual job: structure, judgment, and rigor — automated testing as a structural argument about whether code means what it appears to, not a nice-to-have. That’s where Jason delivers value: not more code faster, but the right code that holds up.
Sane defaults Jason still believes in:
Automated testing remains foundational, not optional. Tests are a domain design exercise first and a regression catch secondarily. Hermetic, targeted, expressive. Factories over fixtures. CI on every commit and every PR. Use TDD to expose complected code and de-complect as you refactor. Get rid of flakes and brittleness; do not tolerate them. These principles do not get less important in the age of AI — they get more important, because AI can now produce a great deal of code very quickly, and tests are the structural argument about whether that code actually means what it appears to mean.