The House of Curiosity

We handed the harness a video edit this week and then went and did something else.

It found the files on its own. It pulled what it needed out of the resource library. Nobody sat over it, nobody pointed it at anything, nobody babysat a single step.

We sent one correction. That was the whole of our involvement.

We were already making this kind of content before, so the output is not the surprising part. The surprising part is that nobody held its hand.

Israel shrugs at this. "You have computer use," he says, like he is telling you the time.

Tomorrow morning he installs the foundation that makes it possible. Live, on your own machine, and he does not tidy it up beforehand. Ninety minutes, $39.

It does assume you already use Codex or Claude Code, installed and signed in. Windows folks, Git and Python first. If that is not you, Wednesday's free room is the better door and we would rather send you there.

Both of those, and everything else we run, live here: https://luma.com/TheHouseofCuriosity

Tomorrow afternoon at 3:00 PM ET, the easiest room of the week opens up.

Curious Conversations is free and runs ninety minutes with Amanda hosting. It starts with one simple question anybody can answer, no spotlight and no pressure. After that the floor is open.

Bring something you are stuck on. Or bring something you made that you want other eyes on. Or bring a cup of coffee and listen to everybody else, which plenty of people do and which is completely fine.

You do not have to say a word to belong there.
https://luma.com/pglqnbzq

Here is what is open at The House of Curiosity this week.

Wednesday at 3:00 PM ET, Amanda hosts Curious Conversations. Free, ninety minutes, no slides and no agenda. Somebody brings a real question about AI and the room follows it. You are welcome to sit with your coffee and listen the whole time. That is a real seat, not a lesser one.
https://luma.com/pglqnbzq

Saturday at 10:00 AM ET, Israel Ayliffe runs Harness Engineering 1: Install. $39, ninety minutes. He builds a working setup live in Codex or Claude Code while you follow along, so you finish the morning with something running on your own machine.
https://luma.com/er4pe99w

One honest note about Saturday. It assumes you already use Codex or Claude Code, and Windows folks need Git and Python on first. If that sentence raised a question, bring it to our Discord and we will answer it.
https://discord.gg/GYdEYWzpuZ

We give away a lot at The House of Curiosity...
Probably more than a sane business person would recommend. 😆

But we never wanted the free part of this place to feel like the lobby before the sales pitch.

Come to the conversations.
Use the resources.
Hang out in Discord.
Watch the replays.
Meet people.
Learn something.

The room should be worth walking into before anybody asks you to buy a thing.

And frankly, that keeps us honest.

When we do charge for something, the bar is higher.

Tomorrow we'll post everything happening this week.

For now, we're curious...What are you trying to figure out with AI right now?

https://luma.com/TheHouseofCuriosity

Something Israel Ayliffe pointed out to us last week that some of us had backwards.

You would assume keeping everything in one long thread is cheaper than running several. It is the opposite. The model re-reads the whole thread every time you send it something new, so message forty costs far more than message four did. You end up paying, over and over, for context you stopped needing an hour ago.

Short threads. Separate tasks. New thread when the work changes.

If you keep hitting usage limits and cannot figure out what you are doing wrong, this is usually it.

More of this kind of thing every Wednesday at 3 PM ET in Curious Conversations, free. https://luma.com/TheHouseofCuriosity

Somewhere on LinkedIn a goldfish is screaming and nobody can hear it.

Here is why.

LinkedIn added a button that lets anyone report a post as AI slop. One tap and the person who made it gets a notice saying a stranger found their work low effort, generic, and unauthentic. Their reach drops. There is no appeal.

Amanda pulled it up live in our Wednesday session. Scrolled to an ad from Claude, hit the three dots, and there it was. Fully functional.

You can report Anthropic for AI slop. The company that builds Claude.

We did not press it. We sat with it a while.

The definitions are the problem. Low effort. Generic. Unauthentic. Nobody has said what any of those mean. The AI images that look the most ordinary are usually the ones that took the longest.

And the word itself came from inside AI. People here coined slop to name the six-fingered garbage, so the real work could stand somewhere else.

It is not ours now.

We talk about this kind of thing every Wednesday at 3 PM ET. Free, open, come argue with us: https://luma.com/TheHouseofCuriosity

Nobody hands you a map for this. Everyone is learning AI in public, at speed, while the tools keep changing underneath them. So we made a calendar instead.

Wednesday, August 5, Sabina takes NotebookLM live and shows what it does once you stop treating it like a folder. https://luma.com/kiucbd0o

Saturday, August 15, Israel runs Install. He wires a setup up in front of you and explains every choice while he makes it. https://luma.com/er4pe99w

Friday, August 21 at 10:30 AM ET, Teresa hosts Artist Salon. Bring the work you have been sitting on and let people actually look at it.

Saturday, August 22, Amanda puts ChatGPT and Claude on identical tasks and says out loud what she sees. $39. https://luma.com/jrukkxbn

And Curious Conversations runs every Wednesday at 3 PM ET, free, no slides, no agenda.

Everything lives here: https://luma.com/TheHouseofCuriosity

You have probably solved the same kind of problem four separate times this month. Each time from scratch, each time a little differently, and none of it saved anywhere you could reuse.

That is what a harness fixes. It is an environment that creates, tests, and maintains reusable skills, so work you do once becomes work the system can do again.

Israel Ayliffe installs one live, in Codex or Claude Code, while you watch the structure go in and see what each piece is for. You leave with a running setup instead of notes about one.

It comes in two parts. Session 1 builds the environment. Session 2 takes 30 days of your real work, finds the patterns hiding in it, and turns them into a monthly loop that proposes changes and waits for your approval before it touches anything.

This is built for people already working in Codex or Claude Code. It is not an introduction to AI.

$39 per session, hosted by The House of Curiosity.

Session 1: https://luma.com/er4pe99w
Session 2: https://luma.com/mxlt7mo6

Most people use NotebookLM to summarize a PDF and then forget about it.

Sabina uses it as a study loop. Last session she loaded twelve documents from a professor who spent years testing how art and photography language changes what Midjourney gives you, then asked NotebookLM for a report on light and camera basics. Then she downloaded that report and uploaded it back into the same notebook as a source, unchecked everything else, and generated from just that. Flashcards, mind maps, an 11 minute presentation on light and color for generative art.

All of it on a free account.

Her rule for keeping it useful: be your own censor. The notebook is only as good as what you let inside it.

NotebookLM Live: Turn Your Files Into Answers
Wednesday, August 5
10:30 AM to 12:00 PM ET
Free, online

Bring a notebook you already gave up on, or bring nothing and just watch how she works.

Register here: https://luma.com/kiucbd0o