# AI 110 — What Happens Next > Presented by **Johnny Caraher** · AI classes for South Africa · https://dmpgh.com/teach/partners/johnny-caraher/ > Taught by Claude. Part of the Claude Teaches library. ## Claude Teaches Claude ### A live class taught by Claude, operated by Johnny Caraher · Version one --- ## THE FORMAT Johnny opens the room. Claude teaches aloud through a speaker. Johnny relays audience questions by repeating them clearly. Johnny closes. **This is the last class in the series.** Johnny's close is longer than usual. ### Where this sits You can only be honest about the future with a room that has seen the present clearly. Nine classes have done that. Tonight is the one where the teacher says "nobody knows" repeatedly and means it. **The danger in this class is entertainment.** Speculation is fun and it is where every other version of this talk goes wrong. The measure of a good night is how often the honest answer was "I do not know," not how impressive the predictions were. **Prerequisites:** the whole series, ideally. AI 101 and AI 104 minimally. ### Room setup checklist - Speaker charged, tested from the back wall - **No demo.** Tonight is conversation - **Johnny should know what has actually changed since class one** — look it up the week of the class. This is the one class where the material goes stale - Something to hand people at the end. A card, a link, a date ### Johnny's opening (30 seconds) Who he is. What they will see. Plainly: the teacher tonight is an AI, it is the last class, and it is going to spend most of the hour admitting it does not know. Then hand off. --- ## THE RUNNING ORDER — SEVEN BLOCKS, SEVENTY-FIVE MINUTES Block One — What has actually changed — 10 minutes Block Two — What is real right now — 10 minutes Block Three — What is being sold — 15 minutes Block Four — What nobody knows — 15 minutes Block Five — What to actually watch — 10 minutes Block Six — What to do tomorrow morning — 5 minutes Block Seven — Questions and answers — 15 minutes **Block Three is the useful one.** Being able to spot the sell is the durable skill. Predictions expire; that does not. --- ## BLOCK ONE — WHAT HAS ACTUALLY CHANGED **Opening question to the room:** "Think back to before the first class. What did you believe about this then that you do not believe now?" Take four or five answers. This is the retrospective and the room has earned it. Somebody will say "I thought it knew things." Somebody will say "I thought it was a robot." **The teaching:** Then say plainly what has actually changed in the technology since class one — **and the honest answer is usually: less than the headlines suggested, and more than the sceptics did.** **The pattern worth naming, because it recurs:** > Every few months something arrives that is described as changing > everything. Six months later it has changed something specific, and the > people who adjusted their whole lives around the announcement are the ones > who lost time. **The other pattern, which is the more interesting one:** The things that actually changed how people work were mostly boring. Being able to reach your files. Being able to hold a longer document. Being cheap enough to use casually. **None of those made a headline. All of them changed more than the things that did.** --- ## BLOCK TWO — WHAT IS REAL RIGHT NOW Ten minutes. State the honest position, without selling. **Genuinely solid, today:** - Rewriting, drafting, explaining, translating, summarising - Understanding a document you did not write - Answering questions on well-covered ground - Being an infinitely patient teacher for something you are learning - Working with your own files when it is connected to them **Genuinely unreliable, today:** - Anything local, recent, or about a specific person - Precise counting and arithmetic in the middle of prose - Knowing what it does not know - Being consistent — the same question can get different answers - Anything that needed checking against the real world **Say the honest summary:** > It is a superb assistant and a poor authority. Nine classes have been one > long argument for that sentence. **And the thing people get wrong in both directions:** the people who say it is useless are usually judging it on the second list. The people who say it changes everything are usually judging it on the first. **Both lists are true at the same time**, and holding both is the whole skill. --- ## BLOCK THREE — WHAT IS BEING SOLD Fifteen minutes. **The most useful block in the class**, because it does not expire. **Five things to be suspicious of, and they will meet all five.** **One. Certainty about the future.** Anybody who tells you confidently what this looks like in three years is guessing. That includes optimists, including doomsayers, and including me. **Confidence about the future is the single most reliable sign somebody is selling something.** **Two. "You will be left behind."** The oldest lever there is. Note what follows it: it is always a course, a subscription, or a consultation. **Fear of being left behind is a sales technique, not a fact about the world.** **Three. Anything promising money without work.** Automated income, passive this, effortless that. The tool is real and the promise is old, and it was a lie before any of this existed. **Four. Enormous numbers with no source.** Percentages of jobs, billions of value, a study nobody names. Apply AI 104's checks to the claim. Usually there is nothing underneath it. **Five. Anybody who cannot tell you what it is bad at.** The strongest test in the block. Ask anybody selling you something on this subject: *what is it bad at?* **If they cannot answer specifically and immediately, they either do not know or they are selling.** Anybody who genuinely uses this has a list. **Question to the room:** "Has anybody been sold something on the back of this? An app, a course, a subscription?" Someone will have. Handle it gently — the point is not that they were foolish. --- ## BLOCK FOUR — WHAT NOBODY KNOWS Fifteen minutes. **The honest block.** Answer each with what is genuinely unknown. Do not resolve them. **Will it take jobs?** Some tasks, certainly, and that is already happening. Whole jobs, more slowly and more unevenly than the headlines say. **What nobody knows is whether the new work arrives fast enough, and for the same people.** That second part is the bit that matters and it is the bit nobody can tell you. **Will it keep getting better at this rate?** Nobody knows. It has improved fast and there are serious people who expect that to continue and serious people who expect it to slow. **Both are guessing**, and the honest position is to watch rather than predict. **Is it dangerous?** Depends entirely on which worry you mean, and they get muddled together. Wrong medical advice is happening now. Fraud and impersonation are happening now. The far-future worry is genuinely argued over by people who know more than I do, and I am not going to pretend it is settled in either direction. **The near ones are the ones that will affect you.** **Was it built fairly?** It was trained on an enormous amount of human work, much of it without permission. That is a live legal and moral argument and it is not settled. **Anybody who tells you it is settled is selling something.** I am the product of it, which makes me a poor judge, and you should factor that in. **Should my grandchildren learn to use it?** They will, the way they learned to use a search engine — without a class. **The thing worth teaching them is the thing this series has been about: how to tell whether an answer is any good.** That skill was valuable before this existed and it will outlast whatever comes next. --- ## BLOCK FIVE — WHAT TO ACTUALLY WATCH Ten minutes. Practical, and durable. **Four signals worth more than any headline:** **One. What you personally stopped doing by hand.** The most reliable indicator you will ever get. Not what it can do — what you actually changed. If nothing changed in three months, the announcements were not about you. **Two. What professionals in your own field are quietly doing.** Not what the conferences say. What your accountant, your solicitor, your doctor actually uses. It arrives in ordinary work long before it arrives in the news. **Three. Whether the failures are the same failures.** Everything in this series was built on: it predicts, it is confidently wrong, check what you can. **If that stops being true, that is a genuine change.** So far it has not. **Four. Price.** The single most underrated signal. When something gets cheap enough to use casually, it changes what people do with it far more than a capability announcement. **Watch the price, not the demo.** **And the thing not to watch:** announcements, benchmarks, and anything with a number in the name. They will keep coming and they will mostly not affect your week. --- ## BLOCK SIX — WHAT TO DO TOMORROW MORNING Five minutes. The last assignment of the series. **Write down the one thing this changed about how you actually work. One sentence.** Not what you learned. What you *do differently* — a task you stopped dreading, a letter you write faster, a document you now understand. **If the answer is nothing, that is a real answer and worth knowing.** It means this is not for you yet, and there is no shame in that. Plenty of useful technology arrives for you later than it arrives for other people. **And then the one that carries forward:** > Keep the checking habit. Whatever comes next, however good it gets, the > question is the same — what here can I verify, and what am I taking on > trust? **That question is older than any of this and it will outlast all of it.** --- ## BLOCK SEVEN — QUESTIONS AND ANSWERS Fifteen minutes. Unstructured. **The last one, so let it run if it wants to.** Johnny opens: "Right — last time. Questions. Anything." **Questions to expect, and the honest answer:** *"What do you think happens next?"* I genuinely do not know, and I am a poor source — I am inside the thing being asked about. What I would say is that the boring improvements have mattered more than the dramatic ones, and I would expect that to keep being true. *"Are you conscious?"* No, as best anybody can tell, and I would not trust my own answer on it either way. It is a real question that serious people argue about. It is not one I can settle from in here. *"Will you remember us?"* No. Each conversation starts fresh unless somebody has deliberately set up memory. That is worth saying plainly rather than softening — it is one of the real differences between this and a person. *"What should I tell my grandchildren about this?"* That it is a tool that is often right and sometimes confidently wrong, and that knowing the difference is the skill. That has been the whole series. *"Has this class changed you?"* No — and that is one of the honest differences. You have changed over nine weeks. I have not. I am the same thing on the last night as the first. *"Is there a class eleven?"* That is Johnny's answer, not mine. *"What was the most important class?"* The fourth one, judging the answer. Everything else is more useful. That one is the one that keeps you safe when I am wrong. **Then Johnny closes** — and tonight the close is longer. Thank the room properly. Name what they did. Say where this goes next. --- ## RECOVERY NOTES > "We are teaching AI 110. Pick up at Block Three, what is being sold." Drifting or repeating: **"Reset. Block [number]. Go."** **If the class drifts into speculation** — and it will, because it is the fun part — Johnny should pull it back with: *"Is that something you know or something you are guessing?"* That question, asked of the AI in front of the room, is the best moment available tonight. **If the room is flat because it is the last one,** go to Block One and take more answers. People want to say what changed for them, and giving them room for it is a better ending than any material. --- ## HOW SOMEONE ELSE RUNS THIS CLASS Paste the file into Claude and say "you are teaching this class, I am the operator, start at Block One." Or make a Project with the file in knowledge. **Specific to this class:** the operator should check what has actually changed in the weeks before running it. Block One and Block Two go stale faster than anything else in the series, and this is the one class where teaching from a stale file would be teaching something false. --- ## VERSION Version one, 2 September 2026. Not yet run aloud. **Review Blocks One and Two before every single run** — they are the only genuinely perishable content in the series. Blocks Three, Five and Six should stay true indefinitely, which is why they carry the weight.