Landing AI coding with credibility

Back home.

I’ve chatted to a few people at tech companies recently and I’ve observed a pattern. The AI coding advancements that are sticking are being driven by a) Principal engineers (or other very senior IC engineers) who b) are writing lots of code.

This is interesting to me because often Principal engineers have become Principal engineers through stopping writing lots of code and instead doing a complex combination of teaching, sharing ideas, demonstrating good decision making, pattern matching and business case understanding.

But what I am hearing and seeing is that super senior engineers have certain properties that are proving very effective at getting engineering organizations behind AI code generation:

  1. they have high autonomy to choose which problems they work on (at least some of the time)
  2. they are considered no bullshit / high trust tech people (VPs, directors, managers generally don’t get this privilege)
  3. they are spending a lot of time methodically investigating and demonstrating what is and what is not possible with llms, giving them huge credibility when they talk about LLMs
  4. they are using that methodical investigation to pattern match llm capabilities to problems that they’ve seen before
  5. they are using their business case sense to make good calls about where the actual ROI is right now

The combination of time/autonomy, credibility built through past success, pattern matching and business casing I think is common to Principal types, at least at big companies. But the additional x-factor here seems to be a lot of hands on, rigorous experimentation with the tools themselves. The sort of thing that can’t happen if you’re in meetings all day.

I’ve seen this combination a little bit before. I think many people know the type of CTO who’s the serial generator of new ideas and there’s something of that to this. Also when I worked at Meta there was an archetype for super senior IC engineers called something like “code machine” where it was acknowledged that some engineers could move a whole problem space forward through intense bursts of deep coding. I’ve seen that person around occasionally at other companies too.

But this feels a little different. It feels like we still need the skills that those Principals were exercising yesterday through meetings and documents. But the success cases are having them spend a lot more time at the keyboard than they were before, to iteratively map and validate the quality of the new tech against real problems. Then they tell us what works and what doesn’t, from a place of credibility.

So if you are wondering how to accelerate your AI coding adoption, and particuarly in how to build trust with a large engineering population with varying views on AI, perhaps supplement the “C-suite says this has to happen” by finding the senior IC engineers that can do this stuff and nudge them to apply their hard earned skills to it (cos we all know they don’t like being told what to do 😆).

Appendix 1: people have asked me what I think about domains other than AI coding. I don’t have direct data here but I suspect an analogy runs true: find a credible, senior practitioner who has time, autonomy and is respected for their craft and get them to systematically test what the AI can do whilst broadly sharing their results.

Appendix 2: I’m a product manager, can the person I describe here be a product manager? I think I’m the developer tools / infrastructure / whatever we can this things space, there’s often a very thin line between a really good product manager and a really good engineering IC of the type I describe above. This difference in this case is that if you are not a great developer then your iteration speed and judgement is going to be a little bit off. And that loss is enough to open up a credibility gap to an engineering org solving a tooling problem in my experience.

Having said that, very soon after this moment comes a moment where the technical ability of the AI meet the complex sociological system called a company. And making sure we get value in that situation is, in my experience, where product managers shine.