The Software GardenIt was never the typing ← Contents
The Software Garden · Why · 01

It was never the typing

Grief is part of the rollout.

For a lot of engineers, writing code isn't a task. It's who they are. Tell them they may never write another line by hand, and some cheer, some freeze, and some go quiet and wait for it to blow over. That isn't a training problem. It's a loss, and it needs handling like one.

Scroll ↓
Name it

It's a loss, not a misunderstanding.

Some resistance is about beliefs: sentences people say that were true of the old process, and evidence moves those (chapter 17, The constraint is mindset). Some isn't. It's people losing something they valued, and no amount of evidence argues anyone out of a loss.

Writing code gave people things beyond the code: the pleasure of the craft, hours of flow, a visible measure of their worth, and the status of being the one who knows how it works. Say that out loud. Then look at what carries over, because most of it does.

The spread

Every team splits. That's normal.

Some people go all in in the first week. Some are curious and careful. Some wait it out in the hope it's a fad, and some stop showing up in any real sense. People move between these at different speeds, and not always forwards. A rollout plan that assumes the average engineer fits nobody.

All in“Why would I ever go back?”Needs guardrails and a way to teach others. Can outrun the structure and leave weeds.
Trying“It's slower than me right now.”In the dip. Needs protected time and someone to pair with, or they'll quit at the bottom.
Waiting it out“Let's see if it lasts.”Needs evidence from people they trust, on code they know, not from a vendor demo.
Hiding“…”Usually afraid, not lazy. Needs safety first: a private first loop, on a small thing, with no audience.
Pressure

Fear doesn't move people forward.

The tempting move is the cost comparison: an agent loop runs for a few dollars an hour, so keep up or be replaced. It's often true, and it's almost useless. People who are already curious speed up a little. People who are waiting or hiding go further into hiding, and some of your best leave.

What moves people is safety, time and examples: protected hours to get through the dip, someone fluent to pair with, leaders who go first in public, and recognition for the new kind of work. Try it on a team of 24.

HidingWaitingTryingFluentLeft
Week 0
 

Illustrative model, not measured data. Same random seed every run, so only your choices change the outcome.

Identity

You were never paid to type.

Typing was the slowest, most visible part of the job, so it's easy to mistake it for the job. The job was always deciding what to build, how it should be shaped, and how you'd know it worked. Agents take the typing. They make the rest of it matter more.

That's where engineering moves: to the edges, vision and validation; into the loops that make code climb to correctness; into rules that enforce themselves. None of that is a lesser job. Most engineers who get through the dip say it's the more interesting one.

⌨️ WasWriting the code, line by line.
🧭 IsDeciding what good looks like, and making it checkable.
🔍 WasReading every PR.
🔁 IsEngineering the loop that reviews them, and what escapes it.
🧠 WasKnowing how the module works.
📜 IsWriting that knowledge down where every agent can use it.
🏆 WasThe fix at 2 a.m.
🛡️ IsThe check that makes that class of bug impossible.
Leading through it

Lead it like a change people are living through, not a tool rollout.

A tool rollout has a launch date and a training session. This has stages, setbacks and people who need different things at different times. It's slower to start and much faster to finish.

Give each person their own way in. A first loop on a small, safe codebase they know, with no audience, does more than any all-hands.

  1. Go first, in publicShow your own agent work, including what failed and what you changed because of it.
  2. Name the lossSay that this changes something people valued. Pretending it doesn't makes it worse.
  3. Protect the dipSet aside time to learn, and don't judge anyone on output while they're in it.
  4. Pair across the spreadPut fluent people with people who are waiting. Peers convince where leaders can't.
  5. Count the new workRecognise rules, checks, skills and constraints removed, not lines written. → 17
  6. Be honest about what you don't knowRoles will change. Don't promise they won't. Do promise how you'll decide, and who you'll ask.
So

Treat it as a loss, give people a way through, and show them what's on the other side.

Engineers aren't losing engineering. They're losing the typing, and the identity that grew around it. Name that, expect the team to split, skip the threats, and give each person time, a pair and a first loop of their own. The work on the other side needs everything they know.

Next: The role of humans → ← From Dark Factory to Software Garden Contents