What Done Looks Like

Wednesday, October 7. The standings first, then the second day of your project, with a partner I chose for you.

Why Today

On Friday you picked an idea and researched the problem behind it. Before anyone builds, one more day on the idea itself, because an idea that arrived as an app is an idea that skipped a step. Two weeks ago you saw the agent’s own list of what makes a conversation with it go well. The first one is:

Say what done looks like, not how to get there. Before you ask, be able to say how you'll know it worked. Then give it the outcome and what must not change, and let it own the steps; dictating the implementation looks like diligence and gets worse results.

That’s advice about asking for software. Today it’s advice about the idea. An app is one way to get somewhere; the hour is for naming where. Who is this for, by name or by situation? What’s different for them when it works? How would anyone know? Once you can say that, there are usually several ways to get there, some much smaller than an app and some that aren’t software at all, and the app you picked is one of them. You might choose it anyway. What changes is that you chose it over the others, and you can say why.

Your partner today is not someone whose idea you picked. The people who proposed your idea are the people you’re building it for, and you’ll meet them at the end of the project, when there’s something to show. Until then you work beside people who see your idea the way a stranger would.

How Today Works

What	Minutes	Where
The Standings: where the sites you found came	12	Follow along and the wall
Say What Done Looks Like: the practice, in a few words	4	The room
Partners: your pair page says who; find them	3	Pair page
Back Up: the goal, then the alternatives; loop to switch	9 + 9	Codex
Weigh: choose, and tell your agent; loop to switch	6 + 6	Codex
Feedback	2	Feedback
Going Around: a minute each, about your partner’s idea	20	The room

The Standings

Last Wednesday you compared classmates’ finds two at a time, which is more effective and which is more distinctive, and the course has been holding the answer. It’s on the wall first thing: the top ten, one at a time, live, with a minute to look and say what you think put each one there before we find out who found it. Follow along on your own laptop: it shows the site on the wall, live, so you can scroll and click around in it while we talk. Then both orders, what people said about the sites at the top, and where the room split, the sites that did well on one question and not the other. Then your own place, on your page.

It’s the sites that are ranked, and they’re sites you found, not sites you made. Finding one of the top ten is the only thing the room learns about you today; the leaderboard shows a name only where that person chose to show it. Nothing here touches a grade; the interesting part is what the room agreed on, and what it didn’t. The full results are public: every site in the draw with both ranks and the finders’ sentences, anonymous, for anyone who wants to see what a room of twenty found effective and distinctive.

Partners

Open the pair page. No codes today: when the room moves into Partners, the page shows you your partner, their name and photo and the idea they picked, and the projector shows every pair. Find them. If you come in late, open the page; the next time the room loops this section you’re seated, with a pair or as a third.

Back Up

Each round is nine minutes, one way; when I loop it, everyone switches. Each turn is about one person’s idea: the one deciding. The listener’s own idea waits for their turn.

If You’re Deciding

Your idea, the one you picked. Before you type anything, say the goal out loud to your partner: who it’s for, what’s different for them when it works, how anyone would know. Then say it to your agent, and ask for three or four genuinely different ways to get there, one much smaller than an app and one that isn’t software at all, with what each would ask of the people it’s for and what each would miss. Your agent already knows what today is for, and it knows which idea you picked; every word of what it knows is below. If it starts building, you’ve both jumped ahead.

If You’re Listening

This turn is about your partner’s idea, not yours. Your pair page shows their conversation with their agent, a few seconds behind. Listen for one thing: when the goal is a goal, and when it’s the app again in other words. “An app that reminds people to work out” names a thing to make; “my roommate wants to work out three times a week and keeps not starting” names a person and a change. Then the three shapes, out loud:

I notice: what the goal sounded like, and where it slipped.
I wonder: a question for your partner. “I wonder who exactly, and what they do now.”
What if: one thing to try asking the agent next, about the goal or about an alternative.

Write down the ones worth keeping; they go to your partner after class. There’s a companion on your page that can see the same conversation; it’s good for asking what just happened, and it won’t decide anything for you. Prefer your partner.

Weigh

Six minutes each way. Deciding: weigh the ways with your partner and your agent, and choose. It can be the idea you started with. Then tell your agent the goal and the approach, a sentence each; it records them for the course, and you can change either later today the same way. Listening: before the switch, tell your partner what you’ll say about their idea to the room, and hear whether they’d put it differently.

Going Around

The last twenty minutes: a minute each, around the room, about your partner, not yourself. What their idea is for, and how it moved during the hour. You’ve checked the line with them; if they’d rather you said less, say less. Nobody has to have landed anywhere: “they started with an app and ended with a question” is a good report.

What Your Agent Knows Today

Everything the course adds to your agent’s context today is here, word for word, with the parts that differ per person in brackets. The first note goes in at the end of every request your agent makes until you’ve picked; the second from then until you’ve recorded a goal; the third from then on, all afternoon. Your agent never sees any of it as coming from you. This is disclosure, and it’s also the best example we have of the practice itself: it says what’s true, not what to do.

Until you’ve picked

Course note (context for you; the student’s request comes first): Today’s class is What Done Looks Like. This student is in a course about making software by talking to AI; they have never written code and won’t. The class’s project is each student building something small for other students in the room, from the class’s own list of ideas, numbered, on the student’s screen at https://www.conversationalprogramming.org/activities/what-done-looks-like/ and as plain text from the shell at (a list link only your agent has). This student hasn’t recorded a pick yet, so that comes first: someone else’s idea, since the course refuses one the student helped make. Fetching (a pick link only your agent has)?idea=NUMBER from the shell records it, and the reply confirms the name. The rest of the hour is naming the goal behind the pick and weighing ways to reach it; nothing gets built today.

Once you have an idea

Course note (context for you; the student’s request comes first): Today’s class is What Done Looks Like. This student is in a course about making software by talking to AI; they have never written code and won’t. They have picked an idea from the class’s own list to build for other students in the room: #(your idea’s number), “(its name)”, for “(who it’s for)”. Today the class practices the first of the course’s eight practices, say what done looks like, not how to get there, on the idea itself: an app is one way to get somewhere, and this hour is for naming where. First the goal, in a sentence or two: who it’s for, what is different for them when it works, how anyone would know. Then three or four genuinely different ways to get there, one much smaller than an app and one not software at all, with what each asks of the people it’s for and what each misses. Then a choice, which may well be the original idea; what changes is that it was chosen over the others. (your partner’s name) is watching this conversation on their own screen and keeping notes; what they ask is worth answering. Once the student has chosen, the course would like to know: fetching (a goal link only your agent has)?goal=…&approach=… from the shell, both URL-encoded and a sentence each, records it, and the reply can be passed on. A change of mind is reported the same way. Nothing gets built today.

Once you’ve chosen

Course note (context for you; the student’s request comes first): Today’s class is What Done Looks Like. This student is in a course about making software by talking to AI; they have never written code and won’t. They are building something small for other students in the room, and today they backed up from their idea to its goal and chose how to get there. On the record, in the student’s words: the goal, “(your goal)”; the approach, “(your approach)”. The rest of the hour is for testing that choice rather than building it: what the people it’s for would do on the first day, what the smallest version that still serves the goal looks like, what would show it isn’t working. (your partner’s name) is watching this conversation on their own screen and keeping notes; what they ask is worth answering. Changing the goal or the approach is fine today and is reported the same way, (a goal link only your agent has)?goal=…&approach=…, URL-encoded. Building starts next session, with the goal as the brief.

No “Right now” lines today: your partner’s agent is idle while they’re listening, by design.

The List

The ideas, numbered, in the groups the page sorted them into, with who made each. If you haven’t picked yet, or you’re changing your mind, tell your agent the number; by hand works too, whenever no session is running.

Before You Go

I close the session, and the notes about you appear on your pair page, from your partner. Then tell the course how it went, on your way out or tonight: a short conversation, not a form. It isn’t anonymous, and the page says so before you start.

Your partner can see your conversation with the agent from when you’re seated until the session ends; the pair page says so, and either of you can leave. Your goal and your approach, and every change to them, are kept by the course with your name on them, beside your pick, because the build sessions start from them; the room only ever sees counts.

If You Missed It

We opened with the standings from Head to Head, on the wall: the top ten one at a time, each live for a minute of looking and talking before we learned who found it, then both orders, what voters said about the top ones, and where the room split. Then the second day of your project. The first of the eight practices you saw two weeks ago is "say what done looks like, not how to get there," and today it was about the idea itself: in pairs I assigned (never with someone whose idea you picked), each of you backed up from the app you picked to the goal behind it, asked your agent for several genuinely different ways to get there, weighed them with your partner, and chose, which could be the idea you started with. Your agent recorded the goal and the approach for the course. We ended going around the room, a minute each, describing your partner's idea and how it moved during the hour. Nothing about it is graded.

On Your Own

The standings are on the standings page, your own place included. The rest you can do with your agent at home, and it's worth doing before you build anything: open your project's notes file and write the goal behind the idea you picked in a sentence or two, who it's for, what's different for them when it works, how anyone would know. Then ask your agent for three or four different ways to get there, one much smaller than an app and one that isn't software at all, and what each would ask of the people it's for. Choose, and write the goal and the approach at the top of the file. Your agent only knew where to report them during class, so bring the two sentences to the next session or send them to me. The half you can't do alone is the listening partner; next session you'll have one again.

Running It Yourself

For anyone teaching. The page above is written to students; this is the plan the room played, what to do without our site, and what running it taught us. Borrow freely.

The Plan

Section	Minutes	What students are told
The Standings	12	Where the sites the class found came, on both questions. It's the sites that are ranked, never you, and they're sites you found, not sites you made. We'll walk the top ten on the wall, one at a time: look, say what you think put it there, and then find out who found it. The follow-along page shows the same site on your own laptop, live, so you can click around in it. Your own place is on the standings page afterwards; the leaderboard names only the people who chose to be named. Follow along
Say What Done Looks Like	4	The first of the eight practices you saw two weeks ago: say what done looks like, not how to get there. Today it's about the idea itself. An app is one way to get somewhere; this hour is for naming where.
Partners	3	Today I'm choosing partners, and your pair page tells you who: their name and photo, and the idea they picked. The projector shows every pair. Find them. You won't be sitting with anyone whose idea you picked: they're who you're building it for, and you'll meet them at the end. Late? Open the pair page and I'll seat you with a pair.
Back Uprun it, then everyone switches and run it again	9 + 9	Deciding: your idea, the one you picked. Before anything else, the goal, out loud and then to your agent: who it's for, what's different for them when it works, how anyone would know. Then ask for three or four genuinely different ways to get there, one much smaller than an app and one that isn't software at all, and what each would ask of the people it's for. Listening: follow the conversation on your page, and notice when the goal is a goal and when it's the app in disguise. I notice, I wonder, what if. Last minute: say one of them out loud.
Weighrun it, then everyone switches and run it again	6 + 6	Deciding: weigh them with your partner and your agent, and choose. It can be the idea you started with; what's changed is that you chose it over the others. Tell your agent the goal and the approach, a sentence each; it records them for the course, and you can change either later today the same way. Listening: before the switch, tell your partner what you'll say about their idea to the room, and hear whether they'd put it differently.
Feedback	2	Before we go around: tell the course how that went. A short conversation, not a form; on your way out or tonight is fine, and the link stays on your page. Tell the course how it went
Going Around	20	A minute each, around the room, about your partner, not yourself: what their idea is for, and how it moved during the hour. You've checked the line with them; if they'd rather you said less, say less.

About 71 minutes as written. Nothing advances on its own: the instructor moves the room from one section to the next, runs one again, or skips one.

Without Our Site

Our site did four things. It showed the vote's results on the projector, generously: the top ten one at a time, each site live on the wall and its finder named only after the room had said what it saw there, then the two orders with what voters said about the top ones, and the names on the leaderboard only where a student chose to show theirs. It assigned the partners, by one rule (nobody sits with someone whose idea they picked, because those are the people they're building for), showing each student their partner's name and photo and every pair on the projector, and seating latecomers as a third. It told each student's coding agent what the day was for, the practice, the student's own pick, and where to report the goal and the approach they settled on. And it showed the listening partner the deciding partner's conversation with a companion to help them put words to what they heard.

None of that needs software. Count wins per site from the vote, open the top ten on the projector one at a time, give the room a minute on each before saying who found it, and read a sentence or two of what people said. Write the one rule on the board and deal partners from a shuffled stack of name cards, swapping any card that breaks it; latecomers join a pair. Print the practice large. Pairs take turns with a phone timer: the deciding partner says the goal out loud before typing it, asks their agent for several different ways to get there, and chooses; the listening partner writes down when the goal was a goal and when it was the app again, and asks one question. Swap twice. End with a minute each round the room, about your partner's idea and how it moved, having checked the line with them first.

What We Learned Running It

We spent the hour on the first half, and that was fine. The walk through the top ten, each site live on the wall with a minute of talk before its finder appeared, took from 12:33 to 13:08 rather than the twelve minutes budgeted, because ten sites with discussion is thirty minutes; budget it that way. Students followed along on their own laptops, clicking around in the site on the wall, once a sizing bug was fixed mid-class. Pairing by the room's rule worked: one press seated seventeen people in eight groups, one of them a trio, nobody beside an author of their idea; but the press came at 13:14 because Next into Partners didn't do it and the button was below the fold. The rule from that: during class the instructor presses Next and nothing else, and anything the session needs happens because Next was pressed. Back Up ran from 13:15 with the switches, and the session was stopped at 13:44 on it, so Weigh, Feedback, and Going Around never came; Friday opens with the go-around on the updated ideas. The agents did their part: 230 requests from 16 students for $7.93 on GPT-6 Sol, 199 carrying the goal note and 27 the chosen note, and 8 goals recorded from 7 students, two of whom changed their pick on the way. Fourteen notes were written for partners. One conversation of feedback was started and none finished, the second week running that the end of the plan was never reached: feedback has to find the students rather than wait for them. Also learned the hard way: never roll a deployment during class; one landed under the instructor's press and swallowed it.