The Junior You Didn't Hire

Your most expensive engineer is going to send a resignation email in October.

You will read it and feel ambushed. He just got promoted. His comp is at market. His one-on-ones with you have been fine. Two of his teammates complimented his work in the last all-hands. The note will say he's leaving for "an opportunity to build something new," and you will spend the rest of the morning trying to remember when he stopped sounding like himself.

The chain of decisions that produced this resignation began with a budget you closed last September, when you decided to skip the junior hire.

You ran the numbers and they were clean. A senior engineer with the Cursor and Copilot and Claude subscriptions you'd already approved could ship more than a two-person team from 2023. Your internal benchmarks pointed that way. Your CTO did too. The thread in your founder Slack ran with it, posting a chart from McKinsey's developer survey showing a 46% reduction in time on routine coding tasks.

The senior cost $220K all-in. Adding a junior would have run another $130K. With AI subscriptions at $400 a month, the math came down to a $225K-versus-$350K choice. That's a Series A burn rate problem solved with one calendar invite, and your CFO sent a thumbs up.

The hire you skipped never showed up on any dashboard. Most decisions that cost a company a year of velocity hide that way. The line item gets cut from the budget. The role never gets opened. The cost shows up two summers later, in a one-line resignation note.

He has 28 people and $4.2M ARR. Eleven of them are engineers. He stopped hiring under the senior layer in late 2024. He describes the choice as the one every other founder he respected was making.

When we started talking, he had a different concern. His senior engineers, the three he was counting on for the company's hardest work, were taking longer than they used to. The architecture decisions he expected to be quick were dragging. Code reviews were thorough and slow. His top performer was now spending what felt like every Friday writing internal documentation that he would have handed off to a midweight a year and a half ago.

I asked him to walk me through his most recent calendar. The exercise was the one I ran with the founder from last week's piece, mapping every 15-minute block.

The hours his seniors used to spend on architecture and tricky integration work had been replaced by something else. They were doing first-pass review on AI-generated code. They were writing the prompts that fed the model in the first place. They were debugging tooling. They were reading pull requests from the rest of the team, including the ones the assistant had effectively authored for their teammates. They were on call. They were taking customer support tickets that used to route to a midweight engineer. They were running standups. One of them had been the only person who could onboard a contractor for a six-week engagement that should have run for two.

61% of marketers say this is the biggest marketing shift in decades.

Get the data and trends shaping growth in 2026 with this groundbreaking state of marketing report.

Inside you’ll discover:

Download your 2026 state of marketing report today.

Get Your Report

The AI was making the routine code go faster. That same code had been the midweight's job before, and the junior's first six months before that.

What the model had done was eat the middle layer. The midweight role had vanished. So had the apprenticeship spread. The flow that used to run from senior down to midweight to junior had collapsed into a single point: him, with a stack of reviewable artifacts to handle. He had the same number of hours and a much wider pile of things to do in them.

He told me, in February, that two of his three seniors were "fine" and one was "showing signs." By April, the diagnosis had widened. The original "fine" pair were trending the same way; he just couldn't see the shape of it from inside the company.

You have probably read this article half a dozen times in the past 90 days. Senior engineers with AI assistance are radically more productive. The McKinsey number on routine coding time. The HackerNoon piece in April on cost parity between automated tools and junior developers. The engineering reorg announcement at every major SaaS company through Q1, every one of them framed as the workforce pivot.

All of that is true at a scale you do not operate at.

The companies in the headlines run engineering orgs of thousands. When a senior gets bored or burned out or recruited away, a lateral pool of mid-level engineers sits in adjacent teams ready for internal promotion. That bench was filled five and seven years ago, when every public software shop was hiring juniors at scale. The pipeline they are using right now to backfill what they keep losing was built before any of us had ever heard the phrase "vibe coding."

You have 11 engineers, three of them senior. There is no adjacent team to promote from. If one leaves, you go to LinkedIn Recruiter and post a six-figure role for a "senior backend engineer with five years of experience and AI tooling fluency." So does the company across the street. So does Stripe, whose comp band starts where yours tops out. You will spend four months looking, find someone you can afford, learn after eight weeks that the fit is wrong, and start over.

The big-company playbook works because they have a hiring market built from their own alumni. The closest thing you have to that bench is a recruiter who charges 25% of first-year salary.

If you stopped hiring juniors, the company across the street did too. So did Stripe. The share of developer postings labeled entry-level now sits under 3% nationally, down from above 15% in 2021. That is the apprenticeship layer the entire software sector has been running on for 30 years. You participated in pulling it down, one quarter at a time, by closing the requisition every September.

Three things are happening at once, and from inside the founder's chair they do not look related.

The senior who is leaving is doing it because his job stopped being his job. He signed up for a role where his judgment compounded across people. He sits in one now where his judgment compounds across machine outputs and 2 AM debugging sessions. The ladder he expected has the third rung sawed off. There is nobody to bring up. The headcount budget says he is the team.

The midweights you do have are sending their resumes out on Sunday nights. The two engineers in the company who used to be junior and are now mid are watching the senior layer get overloaded and seeing there is no one for them to delegate to either. They are next.

The customer is paying for it. Your most expensive engineer is the one writing the apology email for the integration that broke at 2 AM Eastern, because the rotation of people who could have handled it does not exist. Your second-most-expensive engineer is reviewing the AI-generated PR a contractor submitted, because nobody more junior sits between her and that contractor. Your ACV is going to take a hit at renewal because two of your enterprise customers have started comparing notes on response times in DMs you do not see.

Every one of these is a story I have watched play out at three different growth-stage companies in the last six months. The founders did not see any of it coming, because the metric they were watching said the senior+AI bet was working.

From AI and cloud to DevOps and security — WeAreDevelopers World Congress brings the entire modern stack to San Jose. 500+ speakers. 10,000+ developers. One epic September. Use code GITPUSH26 for 10% off.

Secure Your Pass

It buys you nine months. Sometimes 12.

Then the senior starts showing the signs. Slower output. Less patience with reviews. More Slack messages that read as terse where they used to read as crisp. A wedding-related vacation request that turns out to be about something other than the wedding. The recruiter ping he used to forward to the team chat for a laugh is now answered on a Sunday night from his personal phone.

Then he leaves. The replacement search starts. You spend the equivalent of his salary again on the recruiter, the time, and the productivity loss while the search runs. The new hire ramps for four months. During those four months, your other two seniors absorb his work, on top of theirs, on top of the AI surface area that keeps arriving.

By the time the replacement is fully ramped, one of those other two seniors has signaled. Now you are running the search again.

The $130K you saved last September pays five different bills this fall: the recruiter fee on the senior replacement, the four months of ramp on the new hire, the work that lands on your other two seniors during those four months, the slip in your top customer's response times, and the second resignation that follows 90 days after the first.

The reason I am writing this in May is that you still have time before October. Five things to put on your calendar in the next 90 days.

Run the senior-week audit honestly. Sit with each of your seniors for an hour. Take their last completed week and walk it through in 15-minute blocks. Architecture-critical code, code review, mentorship, AI tooling work, customer escalations, on-call, standups, recruiting, internal documentation. Tag every block. If more than 40% of your senior's week is going to code review and AI-output triage, your senior is doing the midweight job that the model was supposed to replace. The replacement landed on the wrong layer.

Hire one junior this quarter. One is the number. One is what your seniors have hours to teach. One is what fits the budget without becoming a project. The role exists to receive delegation from your seniors. The point of the hire is to convert their judgment from a bottleneck back into a multiplier. Pay them at the top of the market for the role. Title them honestly. Make it clear to your seniors that mentorship is part of the senior role itself, weighted in their performance review and paid for in time on their calendar.

Set the AI up to teach. The teams I have watched that are still hiring juniors are using the AI tools the same way the best seniors are. Pair programming with the model. Code review with the assistant as first pass and the senior as second. Architectural decision sessions where the junior writes the prompt and the senior watches the conversation. The model accelerates the junior's growth curve. The supervised year of work it takes to become a midweight still has a senior sitting next to a junior, three afternoons a week, working through the code together.

Stop performing AI-native for your board. Half of the founders I am working with have told me their lead investor wants a slide on "AI workforce strategy" at the next board meeting. Make the slide. Walk through it confidently. Then make decisions inside the company based on the operating reality your seniors describe to you. The slide will look different from those decisions, and that is fine. The investor sending you the template is sending it to a dozen other founders. Most of those founders will break their teams trying to act on it. Be the founder who built something durable.

Talk to your seniors about it by name. Get on a call with each of your seniors. Tell them you noticed the pattern. Tell them you are hiring a junior to take work off them. Their direct-report count will go up by one. The hours they are spending on midweight-level work will go down by half. Ask them what they need to stay another two years. Listen to the answer. The seniors I have watched leave in the past six months almost all left before this conversation happened. The ones who stayed had it.

Run those five moves in the next 90 days and the October resignation does not happen.

That sentence is going to sound like a promise. Treat it as a return on running the audit and making the hire. The founder I described in March did the audit at the end of April, brought in a midweight in early May, and put a junior recruiting plan in motion last week. His top senior, the one I had marked as the closest to leaving, sent him a Slack message on Mother's Day saying he was excited to mentor again. That message is the kind of thing a founder remembers next March. The version of the founder who waited until October to run the audit did not get the message at all.

The founder I was on a sales call with last Tuesday told me his engineering velocity had been "weirdly soft" for the past six months. He had not connected the soft velocity to the junior pipeline he killed in 2024 until we were 40 minutes into the call. He is at the front edge of the window I have been describing in this piece. He has eight weeks.

The press is going to keep writing about the seniors who are 10 times more productive with AI. The October resignation rarely makes the trade outlets.

You have time to write a different story.

The senior+AI math works on a one-year horizon and breaks on a two-year horizon. Almost every decision that breaks a growth-stage company looks like that. Your job is to make decisions that survive the horizon you are operating on, which is measured in years.

The junior you skipped last September can still get hired in June. Hire one. The senior who was going to quit in October will be the person who interviews him.