Crafting the Next Tech Lead (Without Breaking Them First)

My youngest tech lead came to me a few weeks into the job and asked, with real feeling, why nobody had the decency to warn him the role sucks. He's not entirely wrong, and the part that is wrong is my fault.

We did what most companies do. We took someone happy and good at writing code, handed him a title, and acted surprised when the title turned out to be a different job. A tech lead still writes code, just far less of it. The rest of the time they're a project manager, a part-time architect, a mentor, and the news anchor who has to translate the project's reality up and down at once. All of it carrying influence and zero formal authority: they can't make anyone do anything. They can only be right often enough, and decent enough, that people choose to follow, which is the hardest currency in the building to earn and the easiest to spend to zero with one bad month.

When I first did this in 2008, my coding time fell from roughly 95% to 50%, my meetings went the other way just as hard, and the slice of my day I actually controlled dropped to about nothing. Nobody warned me those numbers were coming. I learned it slowly, and at other people's expense.

The thing that actually breaks them isn't the work

What sinks a new tech lead isn't the tasks. It's an identity problem nobody names. For a decade this person has been winning on a scoreboard they can see and trust: lines shipped, hard bugs killed, the gnarly thing nobody else could crack. The title quietly confiscates that scoreboard and hands them one they can't read yet, where their best day might be one where they wrote no code at all and instead unblocked four people and killed a doomed project before it started. Most new leads spend their first quarter privately convinced they're failing, because they're still measuring themselves on the old scoreboard while being paid for the new one. Prep them for the calendar and the meetings all you like; if you don't name the grief of losing the old scoreboard, they'll quietly spiral while telling you they're fine.

The prep nobody does

Not a training course. A few unglamorous weeks before the title ever lands.

Spend the run-up in extra 1:1s, telling them what's coming, the good and the genuinely irritating, weighted maybe three to one so you don't scare them off. Do not promise it'll be easy. You will eat that promise in front of witnesses.

Rehearse the conversations they've only ever been on the receiving end of. They've gotten hard feedback; they've never had to deliver it to a friend. They've muttered about vague requirements; they've never had to look a product manager in the eye and say "not like this." Run those out loud while the stakes are pretend, because the first time someone does it for real, with the adrenaline up, is the worst time to be improvising.

Pick the first project like it matters, because it does, and make it winnable. The cursed legacy system held together by one person and a prayer is not a starter project, however satisfying it would be to finally hand it off.

Then show up, hardest when they're flailing. First quarter, more 1:1s, not fewer. Let them vent at you, because your office is a better place for it than their peers or your manager. Let them make the mistakes you can still quietly clean up behind, and never bench a panicking new lead. Coming back from the bench is worse than the panic was, because now there's a story about them.

One rule sits over all of it: you have authority and they don't, so never fix their problem by pulling a string only your seniority can pull. You'll have helped them once and undercut them twice, because now the team has seen that the real power still lives with you.

The bottom line

You don't make a tech lead by handing over the title and wishing them luck. You make one in the weeks before it lands, and the truth you owe them is about the new scoreboard. Do it badly and you find out, somewhere around month three, that you've quietly traded one of your best engineers for one of your worst-supported managers.