What's Out There

Instructor: the room controls for this activity, and the room display for the projector.

Friday, October 2. I’m back. In pairs today, two different ones, and the first day of your second project.

Why Today

Your website was for you. The next thing you build is for someone else: a classmate, a few of them, people you can ask. On Wednesday, without me, you invented thirty-two ideas for each other and the page sorted them by who they’re for. Today each of you picks one. It’s yours to build, on your own, over the coming weeks, and the people who proposed it are the people you’re building it for.

One thing to hold onto all hour: you are working on the idea you picked. It came from someone else, and now it’s yours. When you pair up, you don’t switch to your partner’s idea, and the two of you don’t merge yours into one; each of you keeps your own, and you take turns helping with the other’s. Your idea comes with you to the next partner.

Two rules. The idea has to be someone else’s. You can’t pick one you helped make; the list marks those as yours, and the course refuses the pick. You’re building for someone in this room, and the first act of that is taking their idea seriously enough to work on it.

Today is research into the problem, not the solution. Nobody builds, and nobody designs either. Before you decide what to make, find out who has this problem, how they deal with it now, what already exists that touches it, what each of those does and doesn’t do, and what people complain about. The temptation is to jump to features and screens; that’s next week’s conversation, and it goes better once you know the problem. Research is a different conversation with your agent than building is. It’s also the conversation where your agent is least trustworthy, which is why you’ll do it in pairs.

Over the next weeks you’ll work with partners who picked something different from you, because they see your idea the way a stranger would. The people who picked the same idea you did come last, when there’s something to compare.

How Today Works

Pair Up

Open the pair page. It shows you a code. Give it to your partner, or type theirs; whoever’s code was used researches first, later. For now you’re both choosing. Three is fine; so is pairing with me. If you come in late, pair with whoever’s waiting.

Pick

The list is below, numbered, in the groups the page sorted them into on Wednesday, with the names of the people who made each one. Read it with your partner. Not one of your own: the list marks those. Which of the others would you use? Which are for someone you know? Which would you still care about in a month? Then each of you tells your own agent which one, by number, and why. It records your pick for the course, and the projector shows what the room is picking as it happens, counts only, no names.

Your agent already knows what today is for (every word of what it knows is below), so you can talk to it the way you’d talk to someone helping you choose: ask it what it thinks of two you’re weighing, or what’s hard about one. The choice is yours. You can change your mind any time today: tell your agent, and it records that too. Changing it after research shows you something is a good outcome, not a failure.

Research

Each round is nine minutes, one way; when I loop it, everyone switches. Whoever’s code was used goes first. Each turn is about one person’s idea: the researcher’s. The checker’s own idea waits for their turn, and nobody researches their partner’s idea for them.

If You’re Researching

Your idea, the one you picked. Ask your agent about the problem, not the solution, and say out loud what you’re trying to find out. The questions: who has this problem? What do they do about it now? What already exists that touches it, and what does each one do and not do? What do the people who use those complain about? If you catch yourself asking for features or a screen, you’ve jumped ahead; come back to the problem. Have it keep a notes file in a new folder for the project, in your words, so you leave with something.

Two things to know. Your agent can’t open the web from where it runs, so everything it names is from memory: some of it is real and current, some has changed, and some is gone. And it will happily start building if you ask it to; today it knows not to, but you’re the one steering. Research is a conversation about what’s true, which is harder than a conversation about what to make.

If You’re Checking

This turn is about your partner’s idea, not yours; yours comes next. Your pair page shows your partner’s conversation with their agent, a few seconds behind. Your job is the part the agent can’t do: when it names something, open it in your own browser. Is it real? Who is it for? What does it do well, and what did it take three clicks to find? What doesn’t it do at all? Then say what you found, out loud, in the three shapes from before:

• I notice: what you found, and where. “I notice Strava does that, but it’s a paid feature.”
• I wonder: a question for your partner. “I wonder who in the room actually has this problem.”
• What if: one thing to try asking the agent next.

Write down the ones worth keeping; they go to your partner after class, not during. There’s a companion on your page that can see the same conversation; it’s useful for asking what the agent just claimed, and it won’t do the looking for you, because it can’t see the web either. Prefer your partner.

New Partner

When I move on, everyone finds someone new. Your idea comes with you; what changes is who checks it. Pick someone whose idea is in a different theme from yours: the further from your own, the fresher the eyes. Same two rounds again. If the first round showed your idea is already done, or less interesting than it looked, pick another: tell your agent, and it records the change.

Going Around

The last twenty minutes: a minute each, around the room. What you’re going to build, who it’s for, and one thing you found out today. The projector shows what the room picked. If you’re not sure yet, say that; it’s a good answer on the first day.

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 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 setting context for an agent: 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’s Out There. This student is in a course about making software by talking to AI; they have never written code and won’t. The class’s next project is each student building something small for other students in the room, and today’s first step is choosing one idea from the class’s own list of 32, which the class invented for itself on Wednesday. The list, numbered, is on the student’s screen at https://www.conversationalprogramming.org/activities/whats-out-there/ and can be fetched as plain text from the shell at (a list link only your agent has). The choice is theirs, with one rule: the idea has to be someone else’s. The course refuses a pick of an idea the student helped make, and the list marks those as theirs. What tends to help: whether they’d use it themselves, who in the room it’s for, and whether they’d still care about it in a month. (your partner’s name) is beside them choosing their own idea, and talking it through together is the point. When they’ve chosen, the course would like to know: fetching (a pick link only your agent has)?idea=NUMBER from the shell records it, and the reply confirms the idea’s name and can be passed on to the student. A change of mind is reported the same way. Research comes after the choice, and building comes later still; nothing gets built today.

Once you have

Course note (context for you; the student’s request comes first): Today’s class is What’s Out There. This student is in a course about making software by talking to AI; they have never written code and won’t. They have chosen the idea they will build for other students in the room over the coming weeks: #(your idea’s number), (its name), for (who it’s for). Their partner is working on an idea of their own; this conversation is about this student’s. Today is research into the problem, not the solution: who has this problem, how they deal with it now, what already exists that touches it and what each of those does and doesn’t do, and what people complain about. Deciding what to build, and how, comes later, when the instructor says; a conversation that drifts toward features, screens, or a plan is ahead of the day. A notes file in a new folder for the project, in the student’s words, is where today’s findings go. This agent can’t open the web from here, so anything it names is from memory and may have changed or gone; the student’s partner has a browser and is checking what gets named here. Switching ideas is fine today: when the research shows an idea is already done or less interesting than it looked, choosing another from the list is expected, and fetching (a pick link only your agent has)?idea=NUMBER again records the change. (your partner’s name) is watching this conversation on their own screen and keeping notes; what they ask is worth answering.

No “Right now” lines today: your partner’s agent is idle while they’re checking, by design, so the nudges from Going Live would only get in the way.

Before You Go

I close the session, and the notes about you appear on your pair page, from both partners. 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 pair until the session ends; the pair page says so, and either of you can leave. Your pick, and every change to it, is kept by the course with your name on it, because the next weeks’ pairings are built from it; the room only ever sees the counts.