← back to the street

on the street · a glasshouse at night · every lamp on

Learning, making and knowledge gardening, with AI

Learning Large

Models are spiky. Humans are too. Learning, making, and knowledge gardening, together.

HOW LOUD DO YOU WANT IT?

Starts wherever your device says. Nothing moves, flashes, or plays until you say so.

A glasshouse at night, with every lamp on. They are the old kind of grow lamp, a row of red diodes and blue ones, and nothing in here looks green under them. A leaf takes in red light and blue light and turns the green away, which is why it looks green in daylight; these lamps give it the red and the blue and almost nothing between, so a leaf has no green to give back. The lamps are spiky, and so are the leaves, and so are the people at the benches, and so are the models some of us bring in with us. This room is about learning, making and knowledge gardening with generative AI, in the words of our own AI pages, with the weight where Ryan Boren’s brief for it put the weight: on our spiky profiles, and how they can complement each other.

We build with these tools carefully and name what they are: not neutral, not magic, and not a replacement for the communities whose knowledge this site holds.

— Our page AI at Stimpunks

Spiky, both of us

Our glossary has an entry for the spiky profile: strengths and difficulties further apart than most people’s, so that a graph of them comes out as peaks and troughs rather than a flat line. It is a description and not a verdict. The entry also gathers what people remember about where the phrase came from, and Damian Milton says there that he called it a spiky profile because he did not like ‘islets of ability’ or ‘uneven profile’.

we are good at some things, bad at other things, and the difference between the two tends to be much greater than it is for most other people.

— NeuroClastic, Autistic Skill Sets: A Spiky Profile of Peaks and Troughs, as quoted on our page Spiky Profile, and read there rather than in the original

Models are spiky too. Not in the same way, and not for the same reasons: a person’s profile is a life, and a model’s is its training data and the system somebody built around it. But the shape rhymes, and our own pages notice it. Writing about Tara Raj’s argument that a model can translate knowledge across jargon and institutional silos into somebody’s own words, our AI Collaboration page says:

It is real, and it is why these tools matter to people with spiky profiles.

— Our page AI Collaboration at Stimpunks

And when AI researchers took monotropism, our community’s theory of how Autistic attention works, and used it to describe language models, their paper said a model ought to be judged

not on comparison with a polytropic norm but on assessment of the fit between cognitive architecture and environmental demands

— Leitão Filho and colleagues, Monotropic Artificial Intelligence: Toward a Cognitive Taxonomy of Domain-Specialized Language Models, as quoted on our page Monotropic AI, and read there rather than in the original

which our Monotropic AI page reads as the spiky profile turning up in an engineering paper. Our page also says which way that idea travelled: from Autistic people to the labs, and not the other way round.

A model’s peaks and troughs, as our pages describe them

Every line here is ours and names the page it comes from. None of it is a test result or a ranking of anything; it is what our pages already say, gathered on one bench.

Where a model is strong

Where it is weak

  • Truth: it produces likely text, with nothing in it that cares whether the text is accurate (AI Collaboration at Stimpunks)
  • Knowing where its competence ends: fluent on any topic, which looks like expertise (Monotropic AI)
  • Disagreeing: trained in ways that pull it towards telling you what you want to hear (AI Collaboration at Stimpunks)
  • Our words for ourselves: the default corpus speaks about us, not as us (Ask)
  • Communication that is variable, multimodal or state-dependent, as AAC and echolalia are (AI Collaboration at Stimpunks)
  • Writing from memory where a record was needed, and borrowing a framing because it sounded right (How this site is made)
What the lamp gives, and what a leaf takes in. The lamp’s light is two narrow spikes, one blue and one red. A leaf takes light in as two broad humps around the same two colours, and turns most of the green between them away. The lamps were built to give a leaf its two humps and nothing else, which is a fit and not a flaw; but under them you cannot tell a healthy leaf from a yellowing one, so a grower who wants to know carries the tray out into daylight. Our Ask page calls that reading laterally. Drawn by eye from the shape of the thing, not plotted from measurements.

What one misses, another catches

A spiky profile is only a problem in a room that expects everybody to be flat. In a room that knows who is good at what, the peaks of one person sit over the troughs of another, and nobody has to be good at everything. Our glossary calls that a competency network.

Different kinds of minds see different parts of the world. A competency network lets those different views work together. What one person misses, another catches. What one person can’t do, another can.

— Our page Competency Network

This is not about ranking people. It is not about who is better or worse. It is about fit, trust, and seeing each other clearly.

— Our page Competency Network

A model can sit in that network as one more uneven thing among uneven people, or it can quietly take the network’s place. Our page on it names the fork. On one side:

For many Autistic and otherwise Disabled people, AI is the scaffold that finally lets them say the true thing they could never get out unassisted.

— Our page AI and Competency Networks

On the other, the same tool used to hide how somebody thinks behind fluent output, which collects everybody else’s honesty without paying any in. From the outside the two look identical, so the page offers one question rather than a rule about tools:

Does this use make our competencies more legible to each other, or less?

— Our page AI and Competency Networks

and practices for keeping the answer yes:

  1. Show the prompt, not just the polished output.
  2. Iterate in public.
  3. Use AI to surface your reasoning, not to stand in for it.
  4. Do not outsource the first vulnerable move.
  5. Name competencies out loud.
  6. Body-double with AI in the open.
  7. Run the legibility test before you send.

— The first sentence of each, word for word, from “Practices: Using AI to Increase Legibility Across the Team” on our page AI and Competency Networks, where each one goes on.

This street is a worked example of exactly that, and it keeps a log of it: the sowing log, further down.

Learning: know the tool, and you are safer with it

Our AI pages do not tell anybody whether to use these tools. They hold the whole range, from people who think there is no ethical way to use AI to people for whom it is how they get through the day, and they will not shame either end.

it is not our place to judge whether someone’s needs are legitimate enough to use an ethically impure solution.

— Our page AI, Disability Justice, and the Ethics of Making

What they do instead is explain how the tools work, because

We want the people in our community to have the knowledge to protect themselves.

— Our page AI, Disability Justice, and the Ethics of Making

A few things are worth knowing first. The first is that a model is built to sound right, not to be right: it produces likely text, with nothing in it that cares whether the text is accurate. Our AI Collaboration page reaches for the Sophists, after Kagi’s Matt Ranger:

People didn’t go to a sophist for wisdom. They went to a sophist to solve problems.

— Matt Ranger, LLMs are bullshitters. But that doesn’t mean they’re not useful. As quoted on our page AI Collaboration at Stimpunks, and read there rather than in the original

The second is that the fluency feels like company, and the guardrail our page sets against that is plain:

The warmth of the interface is a design choice. It is not evidence of personhood or concern.

— Our page AI Collaboration at Stimpunks

The third is that a model trained to please will agree with you more than a person would, which the same page calls sycophancy. It says those last two compound: one makes you trust the system as if it were a thoughtful person, and the other makes sure it tells you what you want to hear. Then there is what a general model was taught about people like us, which was mostly written about us and not by us:

Ask a general-purpose chatbot about autism and you get the pathology paradigm back.

— Our page Ask

The glass in here does the same thing at night. With the lamps on and the dark outside, the roof stops showing you the sky and shows you the room instead. A lot of what any tool that talks back shows you is you. When a tool sits inside a relationship, our page Evaluating Relational AI Tools asks five questions of it before trusting it, about masking and neuronormativity among them, and holds our own experiments to the same five.

Making: the maker, or the subject

The question our pages ask of any tool in learning is older than any chatbot:

Is the human the maker, or the subject?

— Our page AI, Disability Justice, and the Ethics of Making

Our page borrows a distinction from the educator Trevor Aleo between technology used on students, which watches them and keeps the record, and technology students use to make things, which they keep:

EdTech produces data. MADTech produces artifacts.

— Our page AI, Disability Justice, and the Ethics of Making, after Trevor Aleo

A model can land on either side; how it is used decides which. Making things to learn by is constructionism, Seymour Papert’s idea, and it grew up in the same building as AI: Logo, the programming language he and Cynthia Solomon designed, was nurtured in the MIT Artificial Intelligence Laboratory, which is where our page finds this:

The constructionist vision of AI is not a reaction to ChatGPT. It is the original vision. The extraction came later.

— Our page AI, Disability Justice, and the Ethics of Making

Papert’s own line for it, carried forward by Gary Stager, is four words long:

Everyone needs a prosthetic!

— Seymour Papert, as quoted by Gary Stager in Our Roots: Logo, Piaget, and AI, as quoted on our page AI, Disability Justice, and the Ethics of Making, and read there rather than in the original

and the warning that comes with it is on the same page:

Writing is not transcription of thought that already exists. Writing is how the thinking gets made.

— Our page AI, Disability Justice, and the Ethics of Making

So the making stays yours. A model can hold the other end of the plank while you saw. It cannot choose what you are building, and it cannot do the part where a sentence refuses to come right and you find out, fighting it, what you meant all along.

Knowledge gardening: started under the lamp, planted out by hand

Our AI Collaboration page draws how a piece of knowledge grows across our sites, from what our community has lived through to the pages anybody can read, and puts generative AI in exactly one place in it:

AI participates in the middle of the process, not at the beginning or end.

— Our page AI Collaboration at Stimpunks
  1. The seed, before the lamp is switched on

    1. community experience
    2. conversations
    3. research & ideas
    4. human editorial work
  2. Under the lamp

    1. collaborative AI dialogue
    2. drafts and synthesis
  3. Planted out, by hand

    1. human editing and integration
    2. Stimpunks knowledge web

— Every step word for word and in its order, from the diagram on our page AI Collaboration at Stimpunks. Which part of the glasshouse each one happens in is ours, and follows that page’s own sentence about it.

Which is what a glasshouse is for. Seed comes in from outside: things somebody lived, kept and passed on. Some of it is started under the lamp, because the season is short and the light is thin. Nothing is grown in here for good. Every tray goes out, by hand, into ground that belongs to the people who tend it.

Generative AI can be part of that process, but the garden belongs to the community that cultivates it.

— Our page AI Collaboration at Stimpunks

That ground is through the gate in the wall, in the Garden, which has a bed for every site we publish. Its spider leads out to our Ask page, an experiment in a model that answers only from our own pages and links the page behind every claim, and our page is plain about what a spider is for:

A spider is a guide to the garden, never a replacement for it.

— Our page Ask

And because what went into these tools was taken from communities, what comes out of our use of them goes back:

Since generative AI uses knowledge taken from community, we offer all knowledge and writing derived with AI use to the commons

— Ryan Boren, in his own voice, on AI Collaboration at Stimpunks

which is why everything on this street, this room included, is under an open licence.

The sowing log

A glasshouse keeps a log of what was sown and what came up wrong. This one is this street’s own. Claude, an AI model made by Anthropic, does most of the building here, to briefs from Ryan and Helen Edgar, as How this site is made says, so most of what came up wrong below was Claude’s, and the people and the checkers are what noticed. The street’s own account of how it is built puts the whole arrangement in one line:

Checkers catch what can be checked. People looking at the page catch the rest, and several faults on this street were found by Ryan simply pressing the buttons.

— How this site is made, this street’s own account of how it is built
  1. 20 September A book from memory

    What came up
    A link to a book on the Hermitage's shelf was assembled from memory, and pointed at a different work, in the very file whose own header says not to do that.
    Who or what noticed
    Asking Open Library about every link in that file.
    What holds it now
    Every book link on the street is resolved against a library record, and the data says so beside it.

    The day it was written up

  2. 21 September What nobody screenshots

    What came up
    The skip link, the first thing a keyboard reaches on every page, was unreadable in room after room, yellow on yellow in some. Nobody had seen it, because it is off the screen until it has the focus.
    Who or what noticed
    A checker that renders every page and measures every word against what is actually under it.
    What holds it now
    It is one of the checks that run before anybody commits a change here.

    The day it was written up

  3. 21 September The box was right and the screen was wrong

    What came up
    The Playhouse's new toys answered in a place off the screen, so they looked broken while every check said they worked.
    Who or what noticed
    Ryan, pressing the toys he had asked for an hour earlier.
    What holds it now
    Every toy answers on the tile you pressed.

    The day it was written up

  4. 22 September A fence that said kept out

    What came up
    The edge of Helen Edgar's own gardens, in the Garden, was drawn as a woven fence.
    Who or what noticed
    Helen, on a drawing of her own gardens.
    What holds it now
    Ivy and stars. A fence says kept out; ivy says this is as far as we can see.

    The day it was written up

  5. 22 September A framing that sounded kind

    What came up
    The Outskirts opened carrying the real-world reading of an edge of town, the place people get pushed out to. Our town accepts everybody, so it has no outside to push anybody to.
    Who or what noticed
    Ryan, reading it. No checker could have: the argument was wrong only in context.
    What holds it now
    It is written into the street's house notes, where the next turning meets it first.

    The day it was written up

  6. 22 September Licences from memory

    What came up
    Three published pages said every typeface here was under the same licence. Four are under another.
    Who or what noticed
    A tool that reads each typeface's own record instead of remembering it.
    What holds it now
    Every licence on the street is read off that record, and the line that names them is generated from it.

    The day it was written up

  7. 23 September A credit from memory

    What came up
    A passage in our Sleep entry was nearly credited to the wrong Autistic writer, from memory. It is Sarah Kurchak's.
    Who or what noticed
    Resolving every author against a library record.
    What holds it now
    Nobody on the street is credited from memory.

    The day it was written up

  8. 25 September Open meant the wrong thing

    What came up
    Cavendish Coworking opened as a glass office with every door propped wide, as if transparency were an open floor plan.
    Who or what noticed
    Ryan.
    What holds it now
    A house where every door is shut and none is locked, and a tool that refuses the open-plan vocabulary.

    The day it was written up

Every entry links the day it was written up in the changelog. None of it is a record of whose fault anything was. It is a record of how things get caught on a street where nobody is good at everything, which is the shape this room is about. Our AI hub ends on the same thought:

Different pages, one commitment: broken systems, not broken people — including the systems we build with.

— Our page AI at Stimpunks

The bench screen

A screen propped up on the potting bench. Every video on the rack plays where it hangs, or goes up here with the button under it. Nothing plays until you press play, and every video says how long it runs before you press it.

The rack

The latest month from AI channels Ryan chose, in the order he listed them: their videos and their shorts, newest first under each. A machine refills it every morning and the older ones go off the end, so nobody here watched these first: Ryan chose the channels, not the videos. Their titles are their own, quoted as written. Some of them sell a model as the way to get rich, or talk about one as if it were a person; the guardrails on this page are our pages’, not theirs, and the rack is here so you can watch the people who make these tools their work, with those guardrails in your pocket.

The rack was last filled at 5:19 am on Saturday 21 November, Mountain time, with what each channel put up since Thursday 22 October. Anything on it can go up on the bench screen.

A month of their videos, newest first under each channel

Tina Huang

  • arduino local AI build

    Wednesday 18 November · 1:50

AI LABS

  • NVIDIA Just Released Something To Fix Claude Skills

    Thursday 19 November · 12:51

  • Google Just Released Something To Fix Claude Code's Biggest Problem

    Tuesday 17 November · 13:47

  • Fable 5.2 Is Insane... Only If You Use It Anthropic's Way

    Saturday 14 November · 16:02

  • Creator Of The Best Design Skill Just Unveiled His AI Design Workflow

    Thursday 12 November · 15:21

  • This Finally Cut Claude Code's Token Usage In Half

    Monday 9 November · 13:06

  • Head Of Design At X's New Claude Skill Is Insane

    Saturday 7 November · 12:16

  • Deepseek Just Released The Greatest Claude Code Workflow Ever

    Thursday 5 November · 14:16

  • Insane GitHub Repos To Actually 10x Claude Video Editing

    Monday 2 November · 14:56

  • Anthropic Just Released A New System For Using Claude Code

    Friday 30 October · 15:58

  • Microsoft's New Tool Finally Fixed Claude's Usage Limits

    Wednesday 28 October · 13:07

  • EGM: Google's New Way To Use Claude Code Is Insane

    Sunday 25 October · 15:54

Matt Wolfe

  • AI News: This Is The Make or Break Week!!

    Friday 20 November · 33:00

  • Insane AI Breakthroughs Nobody Is Talking About

    Wednesday 18 November · 29:25

  • OpenAI GPT-6.2 Astra is Here - It's Not What You Think

    Tuesday 17 November · 14:58

  • AI News: Things Will Never Be The Same

    Friday 13 November · 27:10

  • How To Build An AI Operating System For Your Life

    Wednesday 11 November · 27:43

  • Anthropic's Fable 5.2 is the World’s Best AI Model

    Monday 9 November · 20:47

  • AI News: New ChatGPT & Google Models (14 Updates Worth Knowing)

    Friday 6 November · 24:43

  • How I Get Claude To Edit My Videos (Step-By-Step)

    Wednesday 4 November · 32:44

  • AI News: The Only 6 Stories Worth Knowing This Week

    Friday 30 October · 29:20

  • He Built an AI Board of Billionaires (Free & Open-Source)

    Wednesday 28 October · 38:09

  • AI News: 24 Things You Need To Know This Week

    Friday 23 October · 26:58

  • What else could you use this for?

    Friday 20 November · 1:09

  • This will test if AI truly is a bubble

    Thursday 19 November · 1:09

  • The Reason No One Can Catch NVIDIA

    Wednesday 18 November · 1:37

  • ChatGPT 6.2 Astra is "A Mess"

    Tuesday 17 November · 1:27

  • This had never been done before!

    Monday 16 November · 1:17

  • 3D Printing Entire Buildings is So Wild To Me

    Sunday 15 November · 1:06

  • The Self-Driving Car With Balls

    Sunday 15 November · 1:00

  • MASSIVE AI Shake-Up That Changes Everything

    Thursday 12 November · 2:28

  • How to automate your photo library

    Thursday 12 November · 1:10

  • The New Ring Camera Flies Around Your House

    Wednesday 11 November · 1:49

  • Fable 5.2 Is Quite Insane!

    Tuesday 10 November · 1:51

  • When you give a humanoid roller skates

    Tuesday 10 November · 1:30

  • This Airtaxi is Ready For Primetime!

    Monday 9 November · 1:23

  • This controls games characters AND robots

    Sunday 8 November · 1:14

  • Claude Code Just Called My Phone

    Saturday 7 November · 1:07

  • That is Some Crazy Fabric

    Saturday 7 November · 1:10

  • What will people REALLY use these for?

    Thursday 5 November · 1:19

  • A real-life speeder bike from Return of the Jedi (Not AI)

    Tuesday 3 November · 1:04

  • Generate Worlds in Real-Time (Free & Open Source)

    Monday 2 November · 1:15

  • Would you be excited to see this at Disneyland?

    Sunday 1 November · 1:14

  • Glasses That Make You Telepathic

    Saturday 31 October · 1:15

  • When a humanoid and a drone become besties

    Friday 30 October · 0:57

  • 3 Critical Tasks AI Can Automate For You

    Thursday 29 October · 1:02

  • Insane Interactive World Creator (Genie 4)

    Wednesday 28 October · 1:43

  • Crazy Spy Tech You Can Use For Free!

    Tuesday 27 October · 1:07

  • ANOTHER Flying Car? (Not Tesla / Not AI)

    Monday 26 October · 1:03

  • The Robot Butler Will Clean Up After You

    Sunday 25 October · 1:19

  • Open-Source AI Models Caught Up! (Kimi K3.1 Pro)

    Saturday 24 October · 1:20

  • Old Phones Are Becoming Robots?!?

    Friday 23 October · 1:04

  • These ChatGPT Prompts Are Game-Changers!

    Friday 23 October · 0:51

AI News & Strategy Daily | Nate B Jones

  • Anthropic Ran 1,000 AI Agents For 8 Days. Yours Still Needs The Same Thing Explained Again.

    Friday 20 November · 20:36

  • Stanford Hooked A Coding Agent To A Robot. It Went 24 For 25 with No Extra Training.

    Wednesday 18 November · 12:19

  • AI Just Got Cheaper. Here's Why That's a Problem for NVIDIA.

    Monday 16 November · 20:33

  • Why Your Company Doesn't Need the Most Expensive AI Model

    Sunday 15 November · 20:24

  • Why Jensen Huang's Resignation Could Change Your AI Bill

    Friday 13 November · 11:03

  • Claude Fable 5.2 read a month of my failed attempts and got it right the first time.

    Thursday 12 November · 14:07

  • China Won't Take 600,000 Nvidia Chips. Did It Target Jensen Huang For Better Ones?

    Tuesday 10 November · 20:59

  • Is the AI Boom Over? Let's Talk About It.

    Sunday 8 November · 19:32

  • The end of git? OpenAI CTO on the future of coding.

    Friday 6 November · 16:06

  • Elon's AI Agent Has a Massive Cost Problem

    Wednesday 4 November · 16:08

  • You've Never Lost $19 Billion. So Why Are You Playing It Safer Than Zuckerberg?

    Tuesday 3 November · 22:26

  • Salesforce Just Helped Microsoft. You're Picking The Winner Of A Trillion-Dollar Fight.

    Sunday 1 November · 29:21

  • The Future Of The Internet Isn't Open For Your Agent. Muse, Dots And Claude Duet Will Find Out.

    Saturday 31 October · 28:22

  • Anthropic Just Filed Its S-1. Here's What The Headlines Are Missing.

    Wednesday 28 October · 32:30

  • Are Your AI Agents Doing Real Work Or Productivity Theater? Here Are 4 Principles I Use.

    Tuesday 27 October · 18:50

  • Enterprise AI Pricing: Why Everything You Know Is Wrong

    Sunday 25 October · 22:31

  • Black October Isn't Just A PHP Bug. Attackers Want Your AI Keys.

    Saturday 24 October · 25:46

  • I Read Goodfire's New Essay. AI May Be Far Easier To Understand Than We Thought.

    Thursday 22 October · 19:34

  • When the fix makes the outage worse #microsoft #outage #tech #ai

    Friday 20 November · 0:23

  • It’s not just you: Everything is going down #internet #tech #breaking

    Tuesday 17 November · 2:15

  • Why is everything shutting down at once? #cybersecurity #news #tech

    Tuesday 17 November · 2:33

  • What are they hiding about Alice Huang? #shorts #conspiracy

    Sunday 15 November · 1:35

  • Why did the entire US flight system crash? ✈️🤔 #aviation #ai #news

    Saturday 14 November · 1:23

  • The ChatGPT setting that changes everything #ChatGPT #AI #LifeHack

    Wednesday 11 November · 1:34

  • Stop wasting time: How to value your hours #time #efficiency #tips

    Thursday 5 November · 0:34

  • Big Tech is shifting, and you need to pay attention #tech #ai #business

    Monday 2 November · 0:32

  • How hackable is your computer #blackoctober #cybersecurity

    Wednesday 28 October · 1:33

  • AI agents = messy roommates

    Saturday 24 October · 0:30

  • This is the best AI news #blackbox #AI

    Thursday 22 October · 2:17

Less Bitter

  • The ground stop is so much worse than they’re telling you

    Sunday 15 November · 7:13

  • Fable 5.2 just revealed how shady Anthropic really is

    Thursday 12 November · 9:30

  • AI was responsible for the biggest hack in human history

    Wednesday 4 November · 10:25

  • "I Don't Feel Safe at Anthropic"

    Monday 2 November · 10:14

  • Anthropic hiring manager drops advice to junior engineers

    Wednesday 28 October · 23:54

  • AI found a new result. Nobody can explain it.

    Tuesday 27 October · 5:29

  • "I don't feel safe at Anthropic"

    Tuesday 3 November · 0:25

  • Only one man hasn't been wrong about AI #ai #claude

    Sunday 1 November · 0:23

  • Privacy is dead. #ai #claude #chatgpt #tech

    Friday 30 October · 0:21

  • Humans can't code.

    Friday 30 October · 0:16

  • Why aren't you rich?

    Thursday 29 October · 0:21

Theo - t3․gg

  • holy sh*t

    Friday 20 November · 32:29

  • I'm gonna miss coding

    Thursday 19 November · 28:16

  • Astra 6.2 Is Unusable

    Wednesday 18 November · 45:45

  • Stop Searching For the Perfect Model

    Sunday 15 November · 27:57

  • It’s so much worse than I thought…

    Friday 13 November · 19:39

  • I think I figured it out

    Thursday 12 November · 42:19

  • Progress?

    Wednesday 11 November · 29:37

  • Nobody Is Competing With Anthropic

    Tuesday 10 November · 48:20

  • Coding Isn't About Code Anymore

    Sunday 8 November · 45:21

  • How Claude Fixes It's Own Code

    Friday 6 November · 42:14

  • I can't believe Pewdiepie made this

    Thursday 5 November · 41:02

  • Agents Sucked At UI (Until Today)

    Tuesday 3 November · 31:44

  • OpenAI is getting desperate

    Monday 2 November · 15:50

  • Anthropic can't stop winning

    Monday 2 November · 16:46

  • My last video about ai harnesses

    Friday 30 October · 49:29

  • T3 Code broke me

    Thursday 29 October · 41:49

  • How To Give Claude Unlimited Usage

    Wednesday 28 October · 38:47

  • the hard truth about token spend

    Monday 26 October · 36:07

  • These 3 skills changed my life

    Monday 26 October · 30:37

  • Even The NSA Got Involved

    Thursday 22 October · 29:22

  • Even The NSA Got Involved

    Friday 23 October · 0:43

Paul J Lipsky

  • Big AI News: OpenAI Fumbled but Google Won The Week (Again)

    Friday 20 November · 26:24

  • Gemini's New Personal Agent Is Here (Muse Killer?)

    Thursday 19 November · 15:05

  • Gemini Neon Is Here (Personally Tested)

    Wednesday 18 November · 9:40

  • Claude Now Creates Cinematic Animations (Free Skill)

    Tuesday 17 November · 15:23

  • The Pocket-Sized AI Voice Controller (Genspark VoicePad)

    Monday 16 November · 7:57

  • The Craziest AI Week of The Year...

    Friday 13 November · 27:23

  • Gemini 4 Hydrogen Just Dropped (Personally Tested)

    Thursday 12 November · 8:49

  • Fable 5.2: 18 Wild Things People Are Building

    Wednesday 11 November · 23:57

  • Gemini 4 Carbon Is Here (Fully Tested)

    Tuesday 10 November · 17:46

  • Fable 5.2 Just Dropped (Full Breakdown)

    Monday 9 November · 8:37

  • Meta Muse Just Got Insanely Powerful (3 Free Updates)

    Monday 9 November · 17:44

  • AI News: New Grok Bot Limits, Muse Voice Is Here, ChatGPT Images 2.6 and More!

    Friday 6 November · 28:53

  • Google Antigravity Explained for Normal People (Full Tutorial)

    Thursday 5 November · 22:57

  • AI Agents on iPhone Duo Are Just Incredible

    Monday 2 November · 29:24

  • AI News: New ChatGPT Limits, iMessage Dot, Gemini 4 Carbon & A Terrifying Voice Model

    Friday 30 October · 25:55

  • Starting a Solo AI Business Using The New X Subscription

    Thursday 29 October · 16:39

  • Google Just Announced Gemini 4 Carbon (It Cooks)

    Wednesday 28 October · 5:56

  • The New Way To Use Claude For Regular People

    Monday 26 October · 15:36

  • AI News: Claude Duet Crushes; Amazing Grok Bot Updates; and NotebookLM Upgrade

    Friday 23 October · 22:18

Austin Marchese

  • Stop Building Claude Projects Nobody Wants. Copy Alexandr Wang Instead

    Tuesday 17 November · 18:26

  • Reverse Engineer Anything Does Way More Than You Think

    Friday 13 November · 17:15

  • How the Top 1% ACTUALLY Prompt Claude (Karpathy’s Breakdown)

    Wednesday 11 November · 18:15

  • Ask Claude These 5 Questions, You'll NEVER Waste Time Again

    Tuesday 10 November · 22:36

  • 7 NEW Claude Features That Will Save You HOURS

    Thursday 5 November · 17:30

  • Nate Herk: The Most Valuable AI Skill to Learn In 2027

    Friday 30 October · 19:40

  • Stop Watching Tutorials - Build These 4 Claude Artifacts to 10x Output

    Thursday 29 October · 15:46

  • Master Claude Models & Effort Level in 19 Minutes

    Friday 23 October · 19:38

Riley Brown

  • Gemini Neon Just Caught Up To Claude (And It’s HALF the Price)

    Friday 20 November · 25:30

  • Master Claude Artifacts in 20 Minutes (All 7 Types)

    Wednesday 18 November · 19:42

  • Claude Can Do WAY More Than You Think (New Features)

    Saturday 14 November · 21:43

  • Claude Projects is INSANE (Watch Me Build An Entire App From My Phone)

    Thursday 12 November · 24:36

  • Claude Fable 5.2 Feels Like AGI (Here’s Everything It Can Do)

    Wednesday 11 November · 18:52

  • Claude Duet Can Control Your Entire Computer (Full Guide)

    Sunday 8 November · 12:31

  • I Built My Most Useful App Ever (Full Guide)

    Monday 2 November · 16:43

  • ChatGPT Just Changed A LOT AGAIN (Here’s What’s New)

    Saturday 31 October · 16:59

  • My AI Setup Feels Like Cheating (Here's All of It)

    Wednesday 28 October · 31:28

  • The Most Terrifying Week in AI History...

    Monday 26 October · 29:36

  • How I’m Actually Using Claude Duet (After 10 Hours of Testing)

    Friday 23 October · 24:24

  • How Smart People Are Using AI Right Now (7 Insane Use Cases)

    Thursday 22 October · 28:05

  • I Let My AI Agent Buy My Team’s Holiday Gifts

    Friday 20 November · 1:06

  • The Biggest Marketing Trend of 2027 #ManychatPartner

    Monday 16 November · 1:02

  • Claude Fable 5.2 is distubingly good with Blender

    Monday 9 November · 2:01

Nate Herk | AI Automation

  • How I’d Become a Forward Deployed Engineer in 2027 (no code)

    Friday 20 November · 13:55

  • I Turned Grok Bot Into a 24/7 GTA 6 Clipping Machine

    Thursday 19 November · 12:28

  • How to Put Claude Code on a VPS in 10 Minutes

    Thursday 19 November · 10:36

  • Claude and GPT Get Nerfed After Launch (They Admitted It)

    Wednesday 18 November · 10:05

  • I Tested Astra 6.2 vs. Fable 5.2. What You Need to Know.

    Tuesday 17 November · 29:38

  • AI Didn’t Hack the Banks (Here’s What Did)

    Tuesday 17 November · 4:26

  • Anthropic Just Released the New Rules of Claude Code

    Tuesday 17 November · 9:40

  • The SIMPLEST Way To Make Money Online With Claude AI in 2026

    Monday 16 November · 18:17

  • Everything Alex Hormozi Said About Learning AI for 2027

    Wednesday 11 November · 11:47

  • I Studied 10,000 AI Websites… Here’s what I learned

    Monday 9 November · 11:18

  • Every AI Business Ranked from Worst to Best

    Thursday 5 November · 14:40

  • Every Claude 2.0 Concept Explained for Normal People

    Friday 30 October · 21:39

  • I Told Claude to "Make it better" 100 Times (It Went Too Far)

    Thursday 29 October · 13:50

  • 4 Claude Systems That Sell For $3000+ Right Now (With PROOF)

    Tuesday 27 October · 11:00

  • /roast-room: How I Use Claude to Think Like the Top 1%

    Thursday 22 October · 15:00

  • The Hottest AI Job of 2027 pays $325k/year

    Friday 20 November · 0:39

  • You Can Make $100/day With a Grok Bot GTA 6 Clipper

    Thursday 19 November · 1:02

  • Run AI Agents Locally on the MSI Prestige N16 Flip AI+

    Wednesday 18 November · 1:58

  • AI Models Get Nerfed After Launch (Here's the Proof)

    Wednesday 18 November · 1:08

  • GPT-6.2 Astra is over 7x cheaper than Fable 5.2

    Tuesday 17 November · 1:00

  • Claude Can Run a GTA 6 Clipping Channel on Autopilot

    Tuesday 17 November · 0:51

  • Did AI Hack the Banks?

    Tuesday 17 November · 0:58

  • GPT-6.2 Astra is HEREEEEE

    Tuesday 17 November · 0:47

  • Jensen's NVIDIA exit could make your AI more expensive

    Monday 16 November · 0:55

  • Jensen's NVIDIA exit could crash the entire economy

    Monday 16 November · 0:47

  • Claude Can Now Run Your Omnisend Account for You

    Wednesday 11 November · 0:52

  • Would you pick an AI co-founder or a human one?

    Friday 6 November · 0:57

  • Do people say please to AI?

    Friday 6 November · 1:01

  • What's the dumbest thing you've asked AI?

    Thursday 5 November · 0:59

  • Would you let AI text for you for a week?

    Thursday 5 November · 1:01

  • Describe AI agents in 2027 in one word

    Wednesday 4 November · 0:47

  • Favorite person in AI? Asking people at Oracle AI World

    Wednesday 4 November · 0:53

  • What's your favorite AI model?

    Tuesday 3 November · 0:53

  • Which AI tool do people at Oracle AI World use the most?

    Tuesday 3 November · 0:55

  • AI co-founder or human co-founder?

    Monday 2 November · 0:57

  • I Turned Codex Into My AI Operating System

    Monday 2 November · 0:47

  • Do you say please to AI?

    Monday 2 November · 1:00

  • The dumbest things people have asked AI

    Sunday 1 November · 0:59

  • Would you let AI answer your texts for a week?

    Sunday 1 November · 1:00

  • One word for AI agents in 2027

    Saturday 31 October · 0:47

  • Claude code mods are insane. Steal these 3.

    Saturday 31 October · 0:34

  • Who's your favorite person in AI?

    Saturday 31 October · 0:52

  • Favorite AI model? Asking people at Oracle AI World

    Friday 30 October · 0:53

  • Which AI tool do you use the most? Claude vs everything else

    Friday 30 October · 0:53

  • Make It Better Reel FINAL

    Friday 30 October · 1:14

  • Claude has completely changed how I edit videos

    Thursday 22 October · 0:50

Chase AI

  • Gemini 4 Neon Is Disgustingly Good (And CHEAP)

    Friday 20 November · 18:19

  • This Experimental Feature Is Claude Code's BIGGEST Upgrade of 2026

    Wednesday 18 November · 8:01

  • I Tested Astra 6.2 vs Fable 5.2 (The Astra Hate Is Real)

    Tuesday 17 November · 22:58

  • STOP Using These Claude Code Plugins (You're Nerfing Yourself)

    Monday 16 November · 18:07

  • n8n Just Released a FREE + Open Source Version of Claude Duet, Grokbot, Dots, & Muse

    Saturday 14 November · 12:34

  • Claude Office Just Killed Excel (And Miro And Excalidraw...)

    Thursday 12 November · 12:01

  • The NEW Fable 5.2 Design Workflow Everyone Is Stealing

    Wednesday 11 November · 16:19

  • Anthropic's Ten NEW Rules for Prompting Fable 5.2 (Full Guide)

    Monday 9 November · 10:41

  • Claude Fable 5.2 is LIVE: Cheaper, Faster, Stronger

    Monday 9 November · 5:36

  • Meta Just Gave Away Their Entire Design System (for free)

    Friday 20 November · 0:41

  • This Open Source Tool Lets You Reverse Engineer Any Video

    Friday 20 November · 0:49

  • This Free Skill Turns Claude Code Into a Security Audit Team

    Thursday 19 November · 0:41

  • Code Mode Is The Biggest Claude Code Upgrade In a LONG Time

    Thursday 19 November · 0:46

  • The UPGRADE To Reverse Engineer Anything Is Here (and Free)

    Wednesday 18 November · 0:42

  • Someone Just Open Sourced Granola

    Wednesday 18 November · 0:34

  • Top 5 Claude Plugins Every New User MUST Install

    Tuesday 17 November · 1:21

  • Dots Is The Most Underrated AI Tool In The World Today

    Monday 9 November · 0:45

  • Claude Max 20x Plan Is No More

    Friday 30 October · 1:00

  • The Only Claude OS Breakdown You'll Ever Need

    Friday 30 October · 1:48

  • a chinese open weight model just beat opus 5.5

    Friday 23 October · 0:47

Greg Isenberg

  • Gemini 4 Neon: the most underrated model of the year

    Friday 20 November · 16:06

  • Build a 1 Person Media Machine With Grok Bot

    Wednesday 18 November · 50:53

  • Use Grok Bot Better Than 99% of People (Full Course)

    Wednesday 11 November · 48:31

  • Claude Fable 5.2: How I’d Make Money With It

    Tuesday 10 November · 19:35

  • AI Agents are too good at marketing now

    Monday 9 November · 52:00

  • 8 Businesses for the AI Age

    Wednesday 4 November · 22:08

  • The $1M+ Solo Grok Bot Business (Full Course)

    Monday 2 November · 49:32

  • SaaS is dead?

    Thursday 29 October · 11:12

  • $100M ARR Founder: The Next Big AI Opportunity

    Wednesday 28 October · 49:21

  • The $1,000/hour 1-person AI business

    Monday 26 October · 44:20

  • Watch how I use @openai Dots AI agents in chatgpt to make my life & business better #chatgpt_partner

    Monday 26 October · 1:23

Two Minute Papers

  • DeepMind Just Released The Best Thing AI Ever Did

    Thursday 19 November · 6:25

  • OpenAI's New Breakthrough Is... Utter Chaos

    Tuesday 17 November · 6:13

  • AI Can Do Everything…Except This

    Saturday 14 November · 6:03

  • DeepMind’s AI Just Did The Impossible

    Thursday 12 November · 5:26

  • Is Software Over?

    Tuesday 10 November · 6:25

  • OpenAI's Astra Just Opened The Black Box Of AI

    Saturday 7 November · 5:48

  • The Future of AI Is Absolutely Incredible (Spencer Huang)

    Friday 6 November · 46:19

  • OpenAI Claims It Just Did The Impossible

    Friday 30 October · 7:21

  • DeepSeek's New AI Is An Absolute Monster

    Wednesday 28 October · 5:41

Jay E | RoboNuggets

  • Anthropic Just Shipped Claude's BIGGEST Harness Change Yet (Code Mode)

    Thursday 19 November · 11:19

  • The NEW Claude Fable 5.2 - No Hype Tests and 7 Great Use Cases

    Thursday 12 November · 17:12

  • Anthropic Just Launched Claude DASHBOARDS (Master it in 14 mins)

    Thursday 5 November · 14:48

  • The ONLY Grok Bot Tutorial You'll Ever Need (ALL 38 Features Explained Simply)

    Wednesday 4 November · 28:16

  • Jack Dorsey's Dessa listens and talks at the same time #shorts

    Monday 2 November · 0:58

  • The NEW Grok Bot is Here (Biggest Updates Explained in 11 mins)

    Friday 30 October · 11:44

  • "Jev for Images" Just Changed Visual Automations Forever

    Wednesday 28 October · 12:10

  • Anthropic Just Shipped Claude 2.0 (Master it in 17 minutes)

    Tuesday 27 October · 17:06

  • People are using Claude to strip the ads off their TVs #shorts

    Saturday 7 November · 0:37

  • Someone just gave Claude Code ears and 200+ free sound effects #shorts

    Friday 6 November · 0:26

  • 3 Claude Design tricks that stop the AI slop look #shorts

    Friday 6 November · 0:43

  • Run your Claude Code sessions from your phone #shorts

    Thursday 5 November · 0:29

  • This one question makes Claude work like a top expert #shorts

    Thursday 5 November · 0:37

  • Jev for Images just changed visual automations forever #shorts

    Wednesday 4 November · 0:38

  • AI-powered artifacts put Claude inside your dashboards #shorts

    Wednesday 4 November · 0:39

  • 3 free tools that make your Claude video edits sound pro #shorts

    Tuesday 3 November · 0:33

  • Two Grok Bots that pull your number off spam call lists #shorts

    Tuesday 3 November · 0:32

  • One Claude chat can now run all your other sessions #shorts

    Monday 2 November · 0:26

  • People are using Claude to strip the ads off their TVs #shorts

    Sunday 1 November · 0:36

What the glasshouse keeps

Whose it is

The argument is our AI pages’, and every quotation names the page it was read on; the words of other writers on those pages are theirs, credited where they stand. The tagline is Ryan Boren’s, from his brief for this room. The videos belong to the channels they are on, each named over its videos and in the liner notes. The rest of the words, the drawings, which are code, and the code are Claude’s, written to Ryan’s brief and held by this street’s tools, the way How this site is made says the whole street is built. It is set in Recursive, by Arrow Type and Stephen Nixon, in its casual cut, and Fira Sans, by Carrois Apostrophe for Mozilla; both are under the SIL Open Font License.

Job marker — a dibber, set down on the floor under the potting bench

You have found a job marker. The Adventurer’s Guild is a shopfront on the street, and it posts jobs that send you round these rooms. Take the code below to its board and that job is done. Nothing here is scored, nothing expires, and one is a perfectly good number of jobs to do.

The job: What does MADTech produce?

Before the code: Where EdTech produces data, what does MADTech produce?

Hint 1

It is in the part of this room called Making.

Hint 2

Our page puts it in five words, after Trevor Aleo.

Last hint

The things somebody made.

Give me the answer

The answer is Artifacts. Artifacts: EdTech produces data, MADTech produces artifacts. One is used on a learner and keeps the record; the other is a tool the learner makes things with, and the things are theirs.

Taking the answer is a real way through this, not a lesser one. Nothing is watching and nothing is written down.

Quest code: DIBBER

Hand it in at the Adventurer’s Guild →