I'd like to give the next talk here (prior-talk.png). It's sort of a follow-on theme, riffing on the idea of the possibility of software engineers' death. And instead, it's more something like, the software engineer reborn; a vision of human AI collaboration. It's basically trying to describe how I, as a software engineer, with, anchored in neuroscience and expanding into infrastructure and back-end as part of my time in AI consulting and QuantSight, have now expanded into the full stack, and what this allows me to do. I'm sort of exploring the idea of what are the most valuable skills to have and develop as a software engineer. I think all humans are probably trying to fill out their range of skills using AI, and it's not clear what the most valuable are. And indeed, maybe none of them are the most valuable. We just have to have a proactive learning approach and a constant attitude to feedback and continual learning, which I believe software engineers are very good at. And so it naturally makes them the most, it makes them a very useful kind of employee as this vision comes to pass.
Yes. There is a very coherent follow-on talk here, and I think the strongest version is a little more specific than “AI makes software engineers more capable.”
The central thesis could be:
AI is not eliminating the software engineer so much as changing the unit of value from “the code I personally know how to write” to “the technical territory I can successfully navigate.”
That lets your own experience serve as the case study rather than the argument itself. You started with fairly deep domain expertise in neuroscience/scientific software, expanded into infrastructure and backend work, and now AI lets you operate meaningfully across much more of the stack. The interesting question is then: what actually remains valuable when the cost of acquiring usable competence in a new technical domain falls dramatically?
I would probably frame the talk around something like “The Software Engineer Reborn: A Vision of Human–AI Collaboration.” It works particularly well immediately after “Have the reports of our death been greatly exaggerated?” Another slightly less grandiose option is “After the Death of the Software Engineer: What Comes Next?”
The argument could develop roughly like this:
- The old model: Software engineers accumulate expertise in particular languages, frameworks, domains and layers of the stack. Crossing boundaries is expensive because learning is expensive.
- What AI changes: It dramatically lowers the activation energy for entering an unfamiliar domain. A backend engineer can build a credible frontend; an application developer can work on deployment; a neuroscientist can build production software.
- My own trajectory: neuroscience → scientific software → infrastructure/backend → increasingly full-stack. AI didn’t replace the knowledge accumulated along the way; it made that knowledge much more composable.
- So what skills should we optimize for? React? Kubernetes? databases? ML? Maybe that’s increasingly the wrong question. Specific technical skills still matter, but their half-life and relative importance become harder to predict.
- The durable skill is adaptation itself: decomposing unfamiliar problems, building enough of a mental model to proceed, asking useful questions, testing assumptions, interpreting feedback, recognizing when something smells wrong, and learning quickly.
- Software engineers are unusually well prepared for this world. Good engineering already consists of iterating against external feedback: compiler errors, tests, users, benchmarks, code review, production failures, documentation. AI makes that loop dramatically faster and broader.
- But breadth isn’t magic. AI can let you operate beyond your expertise while simultaneously hiding the fact that you don’t understand something. Judgment, verification, and knowing when deeper expertise is required become more important, not less.
That last point is important because it prevents the talk turning into “I have Claude Code, therefore I am now an expert in everything.” Your claim is more interesting: AI allows one person to extend their effective range, provided they have the learning habits and engineering discipline necessary to close the feedback loop.
And I think there’s an especially good conceptual distinction hiding in what you’re saying:
Yesterday’s engineer: “What do I know how to build?”
AI-era engineer: “What can I figure out how to build, and how reliably can I determine whether I’ve gotten it right?”
That gives you a much richer answer to Kelly Shaw’s question. The profession doesn’t merely survive. The boundaries around what one engineer can own start expanding. The potentially scarce employee becomes not the person with the longest inventory of currently fashionable technical skills, but the person who can repeatedly enter unfamiliar territory and become effective there.
I’d make that the intellectual center of the talk. Your full-stack expansion is then very convincing evidence for it rather than merely a personal career story.