March 2020. I'm on a Zoom call with 47 engineers. Half of them are about to lose their jobs. I know who. They don't.

That was my second recession as a staff engineer. The first one (2008) I got laid off. The third one (2023) I helped decide who stayed.

I originally wrote this in October 2023. The core is still true. But the AI boom between 2023 and now changed enough of the calculus that this needed a real update. What follows is the original playbook plus a new section on what's different now.

---

What the AI Boom Changed (2024-2026 Update)

I'll get to the evergreen advice in a moment. But first: the 2023 version of this article missed something important, because it didn't exist yet.

The headcount math shifted. Engineers using AI tools effectively are doing more work per person than before. That's great until it isn't — it means companies can sustain the same output with fewer people. The question "is this engineer revenue-critical?" now has a second dimension: "is this engineer productively amplified by AI tools, or are they getting lapped by peers who are?"

AI skills are the new revenue proximity. The 2023 version of this article said to get close to revenue systems. That's still true. But there's a new category of protection: engineers who own the company's AI capabilities — the evals, the agent workflows, the prompt infrastructure — are increasingly hard to cut. That work is new enough that few people understand it deeply, and it touches everything.

Agency beats intelligence. This is the most underrated hiring and survival insight I've seen in the last two years. The engineers thriving in the AI era aren't necessarily the most technically brilliant ones. They're the ones who can direct AI systems toward useful outcomes — who set clear success criteria, catch when models are heading the wrong direction, and iterate deliberately. That's agency: the ability to pursue goals, act on feedback, and self-correct. It's more scarce and more valuable than raw technical knowledge, and it's observable in how someone works.

The side project signal changed. In 2023, "has a deployed side project" meant you were keeping your skills current. In 2026, the more specific signal is: have you shipped something using agents or AI workflows? The former shows you can build; the latter shows you understand the new paradigm. Both matter, but the second gets attention.

The time-vs-timing distinction. Some career investments compound over time regardless of when you start them: deep domain expertise, relationships, foundational skills. Others depend heavily on timing — catching a wave when it's peaking. AI skill-building right now is one of those rare things that is . The compounding value of becoming technically fluent with AI systems is high. And the window where this is differentiated rather than table stakes is closing.

Related Articles

Your Support Team Is Claude Code on a Timer

An L2 support engineer that's just Claude Code on a /loop, with all its state living in Slack reactions. No app, no database, no deploy — and it only pings you when it's real.

What an AI Audit Actually Finds

Most teams asking for AI don't have a model problem. They have a "which problem is even worth it" problem. Here's what a real AI audit surfaces — and why "don't build this" is often the most valuable line in the report.

Inside a Production Voice Agent: How the Stack Actually Ships

Production voice-AI has converged on a pattern: graph-based conversations, separated decision and response prompts, synthetic-call regression testing, and per-component latency budgets. Why the stack looks the way it does — and what most teams are still missing.