Skip to main content
matthewpurdon
← All Field Notes

#AI-Assisted Engineering

Building software when a model is in the loop for every commit.

Core themes explored

Model-in-the-loop workflowsBespoke context maps and stubsEvaluations & quality pipelinesAgentic developer harnesses

6 posts

AI-Assisted Engineering
AI-Assisted Engineering

Postman was a workaround

I set out to explore a sparsely documented API in Postman and lasted about an hour before I gave up and built the client I actually wanted. The generic tool was only ever a workaround for a cost that just evaporated: building the specific thing yourself. On why buy-over-build is dying, why most of your code is now disposable, and the access-patterns-to-spec-to-code ritual I build with now.

Teams & Process

What is a principal engineer?

Reaching Staff means specializing into one archetype. Reaching Principal means mastering the transition between all four. Why you cannot lead a platform (or an AI transition) from a single gear, and how we ran the engine on our largest multi-agent initiative.

Hiring

Interviewing the AI-assisted engineer

For a decade, the coding interview measured whether you could produce syntax under pressure. That proxy is broken. If we want to find engineers who can actually ship, we have to stop asking them to write code and start asking them to judge it. On the review round, seeding quiet landmines, and hiring for taste.

AI-Assisted Engineering

Judgement is the job now

Hand ten engineers the same model and the work comes back in two piles. AI removed the effort filter, and judgement is what is left. On taste, the dependency graph that wasn't, and the audit my vacation is about to run.

AI-Assisted Engineering

Build the model a map

An agent will rediscover your whole database every session, and pay for it. The fix is not a better prompt; it is writing the map down. On indexes, stubs, skills, and the documentation your teammates never got.

Teams & Process

Summaries all the way down

An engineer writes the truth. Four summaries later, an exec reads a guess. Field notes on the compression chain, and the layered artifact I build instead.

Related topics

#The New SDLC#Product#Hiring#Teams & Process#Industry#Sovereign AI