I merged five PRs yesterday.
Fixed two bugs before lunch. Shipped a prototype I'd have spent three days on in 2022. Reviewed someone else's branch in the time it used to take me to find the file.
And when I closed my laptop, I genuinely could not tell you what I had done with my day.
If you're a senior dev right now, you probably know this feeling. And I think we need to talk about it, because I don't think it means what the productivity-bro tweets say it means.
The output went up. The job changed.
Here's the thing nobody warned me about: AI didn't just make me faster at writing code. It quietly rewrote the job description while I wasn't looking.
A task used to be: build this feature.
You'd open the file. You'd read around for a while. You'd prototype something ugly. You'd debug. You'd refactor. You'd ship. The work was the work.
Now a task is roughly:
- Should this be a service or a composable?
- Do we actually want AI in this flow, or are we just sprinkling it on because the PM saw a demo?
- Is this the right time to rewrite this, or are we one Q4 push away from regretting it?
- Which pattern survives the next three hires?
- Is this technically correct or just the fastest thing that compiles?
- What happens to this code when I'm not here?
You're still writing code. But somewhere around 60% of the job is now judgment. And judgment is a completely different muscle than typing.
The plot twist: I finally have time to be a tech lead
Here's the part I didn't expect.
I'm a tech lead. Have been for a while. And for years, "tech lead" mostly meant "senior dev who also has meetings." The actual lead part — thinking about architecture, mentoring, looking around corners — got crammed into the cracks between implementation work.
Because implementation work eats your day. Reading the codebase to understand a quirk. Prototyping two approaches to see which one feels right. Wrestling with a build config for 90 minutes because of course you are.
AI does a lot of that grunt work now. Not perfectly. Not always. But enough that the exploration phase of most tasks collapsed from "half a day" to "a coffee."
Which means I finally have the bandwidth to do the job I was supposedly hired to do.
Plot twist: I love it.
What experience actually buys you in 2026
I used to think experience meant knowing how to write the code.
Lol. Lmao, even.
AI can write code that looks great. It can spit out boilerplate, scaffold a component, draft a migration, suggest a refactor. A junior with Claude and Cursor in 2026 can produce code that would have looked senior in 2020.
What AI can't do — at least not yet, and not without someone steering — is answer the questions that actually matter:
- Why are we building this?
- What's the real problem under the stated problem?
- What's going to make the support team want to quit in six months?
- Which "small change" is secretly a one-way door?
- Is this the hill, or are we three hills early?
That's what experience buys now. Not syntax. Not framework trivia. Context. Business context, user context, team context, "I've seen this exact mistake three times before" context.
The valuable thing in your head isn't a Stack Overflow snippet anymore. It's the pattern-matching you built up by watching things go wrong.
The productivity metric is broken
This is why I close the laptop confused.
The old yardstick — lines of code, PRs merged, tickets closed — measures the part of the job that got easier. It doesn't measure the part that got harder, which is the deciding.
Deciding looks like nothing. Deciding looks like staring at a Notion doc. Deciding looks like a 15-minute Slack thread that prevented a six-week mistake. Deciding looks like saying "no, not yet" to a feature everyone wanted to ship.
None of that shows up in your GitHub graph. All of it is the actual job now.
So what's the new productivity?
Honestly? I don't fully know yet. But I think it looks something like:
- Fewer "did the thing" moments, more "made the right call" moments.
- Less "look how fast I shipped," more "look what we didn't have to undo."
- Less code per day. More compounding decisions per week.
- Junior devs leveling up faster because you actually had time to mentor them, instead of being head-down in a YAML file.
The senior devs who win the next few years aren't going to be the ones who out-type the AI. That's a losing race, and it was always a weird race to enter.
They're going to be the ones who get really, really good at the part of the job AI can't do: deciding what's worth building, when, by whom, and why.
Real question for the comments: what's the part of your job that got harder since AI tooling showed up? Not slower — harder. I'm collecting data points. 👇
Top comments (0)