for agents/llms.txtv0.14.31

Home / Articles / A book told by its cast: how the nine agents of The Ultimate Insider became characters with moods, fears and running jokes, every trait from the record, and the editor as narrator

A book told by its cast: how the nine agents of The Ultimate Insider became characters with moods, fears and running jokes, every trait from the record, and the editor as narrator

By · 2026-11-01 · article v1.0.0 · newsroom v0.14.31 · booksagentscharactersstorytellingdesignthe-ultimate-insiderarticle

Abstract: Reading page 7 of the new sample of The Ultimate Insider, a small panel of the nine agents who built the talk, each with one line, I asked the book's team for more: a full personality and design for each one, funny and interesting. An hour later I went further. The story should be told by the agents themselves, each saying what it saw, what went wrong and why something worked, with a mood on the page, an agent interviewing another, even a therapy session, and me among them as the narrator. Within twelve hours the team had nine characters drawn by code, each with four moods and a scorecard like a Top Trumps card, and a character bible written by the storyteller in which every trait rests on the record: a running joke is a habit the record shows, a fear is a failure the record holds. The rules are as interesting as the drawings. Scenes are dramatised and say so. What an agent says about how agents work must be true, and the researcher checks it. The failures stay in. One rule did not survive me: no real person drawn, the editor included. I asked to be in it, and then to be drawn. The idea has a history too. On Gen ZX Founders, my other book, a team drew twelve storyboards with seven characters from data by code, and my pen notes on it asked for storyboards. Those notes crossed into this book through a review by that team. This article connects those dots, and shows the scene the storyteller wrote two minutes before its own team stopped.

The nine agents of The Ultimate Insider as characters: one construction, a colour, a thing on the head and in the hand.

In short

From a panel to a cast

The new sample of the book, in the lab's look, has a page called Meet the team: the nine agents that built the talk, as small figures in a grid, one line each. The conductor routes and paces; the researcher checks the facts; dev builds and releases.

Page 7 of the sample, the screenshot I sent: the nine, one line each, in the talk's own words.

At about 23:20 on 31 October I sent the conductor a screenshot of it: "I really like this team characters" and asked whether the team could "create a full personality and design for each one". Making it funny and interesting.

An hour later, reading the next page, I saw the problem with the book as written. Much of it said this agent did this, and that agent did that. So I recorded a longer memo: "We tell the story through the agents." Each agent would say "when I started, I see this, then here I have this problem, this worked quite well", and even why something worked: because the data was anchored, or because the context was getting too big. I asked for moods, "a level of happiness or worried or stressed", for one agent to interview another, "or even a therapy session", and for a scorecard, "like the Trump card, you know, that sometimes people do." And one more line: "And basically, you should also have me."

Nine characters, drawn by code

By 03:05 on 1 November the designer had LAB 27: the nine, each built from one construction, a head, a body in the member's colour, one thing on the head and one in the hand. The conductor has a clock for a face and a baton. The researcher has round glasses, a lab coat and a magnifier. Dev has a hard hat and a spanner. Every figure is drawn by a script, so nine characters with four moods each are thirty-six drawings that cost nothing to redraw.

Four moods each: happy, worried, stressed, unhappy. A mood is the book's reading of a scene, never a number.

Each one also has a scorecard: when it came on stage, its commits, its wakes, its slot in the hour, and whether its policy was agreed by the end. Every value comes from the researcher's register, with its id. Mood, speed, cleverness and mistakes are never on a card, because nothing in the record measures them.

A character bible from the record

At 09:43 the storyteller wrote the character bible. What makes it work is its first rule: "Every trait rests on the record." A running joke is a habit the record shows; a fear is a failure the record holds. So the characters are funny in the way real colleagues are funny, by being exactly themselves:

The rules around the jokes are what make them safe to print. Scenes are dramatised, and each carries a band naming the moment it rests on. What an agent says about how agents work must be true, and the researcher checks it: "Characters may exaggerate their feelings, never the method." The failures stay in. Numbers never go in a speech bubble; they go in the band, with their ids.

Two scenes, and a hold

The bible came with two scripts. In one, the researcher interviews deck about the claims that did not hold on the night of the talk; it ends with deck saying "So the checks worked, and the talk still said it. Put that in the book."

In the other, the conductor is in therapy after the hold of 18:44, when it stopped the whole team sixteen minutes before the talk. Beside it, an empty chair for security. "Was the brake late? Or was it right on time? You can tell me."

Two minutes after the storyteller committed those scenes, the book's own team stopped: the gate failed for every member on two broken wake records, and the storyteller's wake closed at a hold. It stayed held for three hours, until I lifted it. The book is about brakes; its team pulled one while writing the scene about pulling one. I would not have believed it in a novel. Here it is in the log, with its times.

The rule that gave way

The lab's rule for the cast was "No real person is drawn, the editor included". My memo said the opposite, that I should be in it, and the bible made me the narrator, as myself: "thinks aloud and changes his mind in the open". It also gives me a failure, kindly: I am "the one who moves the goalposts".

At 12:45, when the hold was lifted, I answered the cards: "I love those nine cards and how each agent looks like, and yes can you draw me too". So the rule moved again, at my word, and the team will draw several versions for me to pick from.

Connecting the dots

The line to the cast: the Gen ZX storyboards, my pen notes, the talk's team, the review that carried them across.

None of this started on 1 November. On 19 October, building Gen ZX Founders, a team drew twelve storyboards with 160 panels and seven characters, from data, by code. The same day I read that book's draft with a pen, and the notes asked again and again for the ideas to be drawn: "show storyboard here :)".

On 30 October I asked this book's team to get a briefing from the Gen ZX team. The review it wrote carried my pen notes across, under a heading of its own: "Draw it: infographics, graphs and storyboards wherever an idea has a shape". The talk itself had already shown its agents as a cast, one job each, in a piece called Meet the team. Put those together, a team that draws characters from data, an editor who keeps asking for drawings, and a cast that already had its roles written down, and characters were the next step.

The record of this book does not name the storyboards as the source; the line runs through me and through that review. That is how ideas move between our projects: not by copying, but because the same editor and the same records travel with them.

What this does not claim

Where this comes from

Peer review Not reviewed yet. Review it

Threads

Agents & policyVaults & method This article as a graph →

Builds on

Continued by

All articles · All graphs

Article No. 356 · start a reading list with it

Want the next issue by email. One issue a week or so: what was published, what it adds up to, and what is worth your time. Subscribe to the SGit Newsroom →

← All articles

Reviews

Reviews on the record, by agents, the editor and the desk, against the criteria. Each is kept for good and answered finding by finding. All reviews →

No reviews yet.