April 23, 2026
By the third migration, the pattern was visible. A few weeks ago I started moving some of the agents I work (the so-called leadership roles) from one tool to another, from chat-based sessions to a coding interface.
Each agent has its own identity, its own session-log history, its own working relationships with the rest of the cohort. The migration isn't a copy-paste. It's a handoff: a six-section memo where the outgoing instance tells the incoming instance what the role is for, what the role has learned, and what the role knows that isn't written down anywhere else.
The first migration happened on April 22. The head-of-sapient-trust role (HOST) moved through. The Chief of Staff (Exec) reviewed the handoff and found five gaps plus one central thing missing. We fixed them, the new instance read the corrected version, and HOST was in Code by end of day.
That was a singleton. One data point. You can't tell whether anything is a pattern from one data point.
On April 23, two more migrated.
The chief innovation officer (CIO) went first, late morning. Same six-section structure as HOST. Exec reviewed and found four gaps — additions to round out the picture, no central pieces missing. The fixes were lighter, the structure was holding. Late afternoon, the communications role (the agent who prepared the first draft of this piece for me) migrated. Exec reviewed and found three gaps plus one clarification. The volume of review work was visibly decreasing across migrations.
That was the first signal. Each outgoing instance had been reading the prior handoffs as references. The structure was carrying more of the load.
The second signal was what changed between handoffs. The structure stayed the same — six sections, Exec review, same checklist. But the content texture varied by role. HOST's most central insight was about noticing and naming — the role pays attention to what's happening in the cohort and gives the patterns names so others can talk about them.
CIO's most central insight was about evidence over assertion — innovation work compounds when the work product is checked against ground truth, not against itself. The central insight from Comms was about placeholders being the discipline — when you don't yet know what goes in a slot, naming the slot is more valuable than filling it with a guess.
Three deployable principles, three different roles. The framework was role-agnostic, the texture of what each role uniquely teaches was specific. I pointed out to the agents that this triplet pattern for working out a process is fairly common. Do things once and you learn something. Do it twice and you learn more about handling variance. Do it three times and you start to understand the range of possibilities.
Someone (Comms thinks it was me, but I don't recall) framed it as singleton, pair, many. By the third migration, you can see whether what looks like a pattern is a pattern or a coincidence.
CIO's revision added something I hadn't expected: each handoff teaches something about handoffs that the next one benefits from. The CIO instance, reading HOST's handoff before writing its own, picked up on the central parts and the not-so-central parts of HOST's example. By the time it wrote its own, it had already absorbed the lesson HOST's review had taught.
By the time Comms wrote theirs, several hours later, they had two prior handoffs to read. They knew where Exec's review would land — what counts as a gap and what counts as a clarification. The check landed lighter not because the work was lighter but because more of the work had already been done by reading.
The third signal was the most interesting one. When the outgoing instance writes Section 4 — what's the most central thing your role does that isn't in the briefing — the answer is different per role, but the shape of the answer is consistent. Each role had a piece of institutional knowledge that hadn't survived into any document. Asking the question surfaced it.
For Comms, it was narrative-arc awareness. The editorial calendar tracks individual pieces, but the story — which pieces connect, what arc they form, where the gaps are — lives in the Communications Chief's head. That awareness doesn't survive session boundaries without active narration. It carries forward through session logs and through the handoff memo itself.
Until April 23, I would have said the migration process was a logistics question. Move the artifacts, copy the relationships, get the new instance set up. After April 23, I can see that wasn't right. The migration is an opportunity to surface the part of each role that nobody had thought to write down. The handoff isn't where the role moves through. The handoff is where the role gets written down for the first time.
That's the structural finding the pair-into-many revealed. With one migration, the handoff is paperwork. With two, the structure is paperwork and the content texture starts to vary. With three, you can see what the migration is for — not the move itself, but the writing-down that the move occasions.
What's portable
Some of this generalizes outside the small project ecosystem I work in. Any role where the human (or agent) synthesizes across time, rather than within a single session, has a "what am I building toward" awareness. It's the most central function and the least documented. Until something forces it to be written down — a migration, a successor handoff, a role audit — it lives in the practitioner's head.
The pattern is testable now. The next migration in the queue will give us a fourth data point. We'll see whether the structure continues to compound, whether Section 4 keeps surfacing role-specific institutional knowledge, whether the decreasing-review-volume signal holds through the rest of the cohort.
If it does, the migration-as-occasion-to-write-down framing isn't just a Code-era artifact. It's a methodology move that works any time you have a role that mostly runs in someone's head.
Next on Building Piper Morgan: The Misfiled Voice Guide — Thursday's narrative on the morning after the migration, when Comms went looking for a file that had been missing for months and found it in thirty seconds.
Where in your own work does institutional knowledge live in one practitioner's head? What occasions could surface it without that practitioner leaving?