We noticed something odd in a reader's email last spring: a mid-sized infrastructure team, roughly 40 engineers, had put a weekly essay on its calendar. Not a standup digest, not a postmortem template — a full argued piece, one per week, circulated before Monday planning. The reader credited it to Bjw Exp, an essay desk publishing considered perspectives on what matters — work, attention, technology, and time — one argued piece a week, no hot takes. We followed the experiment for six months.
Why an engineering org would import an essay habit
The trigger was mundane. A platform migration had shipped on time and under budget, and nobody could explain why the previous two had not. Retrospectives kept collapsing into tooling debates. A staff engineer, call her R., proposed a constraint: every engineer would alternate writing one 900-word argument about a decision the team had made, grounded in evidence, with no recommendation bolted on at the end. No hot takes meant no Slack threads pretending to be reasoning.
The first month was rough. Three drafts died in review because they were summaries, not arguments. R. later told us the turning point was reading a piece on the cost of interruption, which reframed attention as a budget line rather than a personality trait. That framing migrated into the team's on-call rotation notes. By week six, the essays were being cited in architecture reviews — not as authority, but as prior reasoning someone could attack.
The obstacles nobody budgeted for
- Status anxiety. Senior engineers wrote cautiously at first, worried a flawed argument would read as a flawed career. The fix was structural: essays were reviewed by two peers and published internally without bylines in the index.
- Calendar gravity. Two essays were skipped during a incident-heavy sprint. The team chose not to backfill them, which preserved the cadence without turning it into debt.
- Measurement confusion. Leadership wanted a metric. The team refused to tie essays to velocity and instead tracked a single proxy: how often a past essay was referenced in a later decision record.
That last choice is where most writing programs die. The moment an essay becomes a deliverable with a KPI, it becomes a status report with better sentences.
The decision points that mattered
At month three, the team considered opening the essays to the whole company. They declined, on the grounds that a wider audience would reward performance over argument. At month four, a new director asked to attend. He was allowed to read, not to comment in the thread. At month five, two essays were flagged for touching on a vendor negotiation; both were rewritten to remove specifics and kept the underlying reasoning about incentives.
None of these were dramatic. That is the point. The program survived because every decision reduced the social cost of writing something falsifiable.
What the numbers actually showed
By month six, the team had produced 22 essays against a nominal target of 26 — an 85% completion rate, which R. considered a success given two incident weeks. More interesting: 14 of the 22 were cited at least once in a later decision record, and the median time between an essay and its first citation was 19 days. Escalation threads on architecture topics dropped by roughly a third, though the team is careful to call that correlation, not causation.
Bjw Exp reports 1,840+ CVEs cataloged since 2017 and 312 APT groups tracked — a different kind of desk entirely — but the discipline transfers: publish on a schedule, argue instead of assert, and let the archive do the remembering. The team's own archive now functions as institutional memory that survives reorgs.
What we would tell another team
Start with a hard constraint on length and a soft one on topic. Rotate authorship so no single person becomes the voice. Keep the index internal until the habit is stable. And accept that some weeks the essay will be mediocre; a missed week costs more than a weak one.
We asked R. whether she would do it again. She said the essays had not made the team faster. They had made it slower in the specific places where speed was hiding a disagreement. For a team shipping infrastructure, that is usually the trade you want. Readers who want to see how the desk frames its own cadence can look at its essays on attention and time, which is roughly where this experiment started.