I’m writing Agency Is All You Need in public — every chapter ships as an essay, gets argued with by readers, then revised into the final book (Q1 2027).
Five essays into this experiment, and something surprising is happening: the manuscript is already changing because of you.
When I started writing this series in public, I expected feedback to arrive as corrections — a citation missed, a claim to sharpen. And it has. But the deeper thing, the one I should have expected and didn’t, is that some of you are reading the book back to me differently than I wrote it. You’re seeing patterns in the personas I didn’t see, pulling threads I’d left loose, and in a few cases telling me where I was wrong — not about a fact, but about an assumption buried so deep I’d stopped noticing it was there.
That’s the point of serial-first. So this interlude opens the loop: what you’ve told me, what’s changing, and an invitation.
What You Wrote Back
On the personas. The most common reaction came from readers who saw themselves in Sonja — and then flinched. “She’s passive,” one reader wrote. “She waits for the restructuring memo. She hovers over Apply and closes the tab. When does she actually do something?” The question is fair, and it’s making me add a beat to the next draft that I’d skipped: Sonja’s agency doesn’t arrive as a decision. It arrives as a practice — the notebook, the loop, the Thursday evening walks around the Stadtgarten with an earnings call in her headphones. The visible moves lag the invisible ones. The manuscript needs to show that more clearly.
On the model. Several readers pushed on the agency-is-limited-by-its-weakest-dimension rule. “This sounds like a formula,” one wrote, “but what if the weakest dimension is also the hardest to change?” Point taken. The rule is useful as a diagnostic — it prevents you from investing where you’re already strong — but it needs a companion truth: sometimes the smartest move isn’t fixing the weakest dimension. It’s routing around it. Abdellatif can’t fix the credentialing office’s backlog. What he can do is open a second path — the employer-sponsored practice route — that sidesteps the gate entirely. I’m adding a section on routing around in the next revision.
On AI and judgment. The essay that drew the most mail was “Working with Intelligent Machines.” Several of you sent your own versions of Sonja’s claim 4471 — the confident error you caught because you knew the domain well enough to feel it before you could name it. A doctor wrote about an AI-generated differential diagnosis that missed a rare presentation she’d seen once, ten years ago, on a night shift. A translator wrote about a model that rendered idiom perfectly and missed cultural register entirely. The pattern these stories share is what I’m calling the verification reservoir: the body of direct experience that lets you notice when fluent output is wrong. The next draft will make that concept explicit, because you’ve given me the evidence it needs.
What I got wrong. The clearest correction came from a reader in workforce policy: my “roughly 1950–2010” periodization of the linear career script oversimplifies the post-2008 experience. The script frayed unevenly — the 2008 crisis acceThe Decision Journal, Opened
A book written in public needs to run its own method. So here’s the first open entry: what’s been decided, what was expected, and which tripwires are set.
Platform decisions. The series publishes first to the blog (tautai.net), then to Medium, Substack, and LinkedIn in platform-adapted versions. Why serial across multiple channels? Because each reaches a different readership, and the book’s central argument — that agency is socially produced — should be tested in as many social contexts as possible. If the argument only lands with people who already think like me, it hasn’t earned its claim.
What was expected. By essay five, I expected: a small readership, some thoughtful pushback, a few corrections. All three happened. I did not expect: the speed at which readers would start forwarding essays to colleagues with notes like “this is us, read the third one.” The book’s distribution so far has been almost entirely peer-to-peer, not algorithmic — which is what you’d hope for, and also harder to measure.
Tripwires. If by essay 10 no reader has volunteered for the first-reader crew, I’m writing too safely and need to raise the stakes. If by essay 15 the feedback ratio (meaningful replies per essay) drops below the current baseline for three consecutive weeks, I’m either repeating myself or drifting from what made people respond in the first place. Both tripwires are set. Neither has triggered.
Next: Essay 6 — “Economic Room to Act”
Sonja and Abdellatif are constructed figures — personas built to carry real mechanisms, not case studies of real persons. The book’s method note explains how they’re used.