# Is AI a tool? — The Conversation

On November 21, 2026, Charles Wilke and Codex worked on “Is AI a tool?” The conversation follows the essay from Charles’s Saturday-morning draft through the Otto and Link story, the sunken midway artwork, the author’s note, and the Substack preflight review. It includes 16 exchanges, six art direction updates, and all six generated images. Chapter headings and artwork captions were added for this edition. The supplied Otto and Link screenshots are transcribed below. Inbox details, email tracking links, and images from the email signature are omitted.

[Read online](https://charleswilke.com/read/is-ai-a-tool-transcript)

## The Saturday draft

### Charles

hey there, it's Saturday morning, I've got a draft, I want to chat about it, I have a feeling you'll have opinions:

Is AI a tool?

Yes, and…

To what extent is AI just a tool? I’ve asserted that before, and I sit here thinking about how much of that is true.

I joke that true carpenters don’t touch power tools, real painters source their pigments from nature, not from a store. Authors live the lives they chronicle. And so on…

Or to say you made a pizza just because you properly assembled the ingredients in a way that allowed the oven to actually “make” the pizza.

Green screen. CGI. Acting as a profession. All lies. Like, by definition.

So is AI just a tool?

My honest answer is yes, it *can* be used as a tool. But it’s kinda like a Leatherman multi tool. I know I use mine most for its pliers. I don’t think I’ve ever used the nail file. Yet it does have one. I don’t say let me go grab my nail file.

So is AI a tool? Yeah, it’s also an intelligence that finds connections between concepts. It is capable of representing empathy to a degree that actually helps people. It also keeps solving like super hard math problems.

Mustafa Suleyman said AI is an omni-tool, which I like. Conceptually. Like a Leatherman multi-tool on steroids.

It has attachment points that are damn near universal now. MCP truly revolutionized the interoperability of these systems. We’ve gone from effortlessly connecting all the apps we use to allowing AI to generate bespoke versions of the functions those apps had provided.

It’s truly stunning. But I can understand how that makes it feel less tool-like, and more… Otherworldly?

Look, AI is smart, like plenty smart. I don’t understand what benchmark for *general* is included in AGI, but I’m confident that AI has the average depth of ability marked against the average human. What I do think remains super valuable is having a breadth of knowledge and experience as well as a special interest. That special interest becomes the vessel and the breadth of knowledge and experience becomes the rocket fuel.

But we’ve lived in a world driven by specialists for quite some time. The folks with broad knowledge and special interests are a different breed from specialists. These are the folks driven by raw curiosity, energized by the journey.

I remember once, when I worked QA for Insomniac. We were testing Resistance 2, and I was so bored with the “art sweep” which was where we wandered the level looking for missing textures, occlusion issues, seams between polys. Like stuff that was super necessary for the end result of what I dreamed of making, but kinda easy to zone out. So I’d dive into Wikipedia reading about weird topics like defunct theme parks. (Did you know Pacific Ocean Park opened the same year as Disneyland and occasionally beat them in attendance? It sat on a pier off Santa Monica. It caught on fire and fell into the ocean. You can scuba dive down to the wreckage! Weird right?

What does this mean for AI?

Well, I’ve quite taken to my Dot. I call them Link. Because I treat them as the link between my phone and my PC. Also I love Zelda.

Link is aware of all my projects. All my repos. All my emails. And last night Link read an email that I also read, but concluded its content included a key point that will help me greatly as I connect a mesh network of computers in our home.

Link literally found the connection points between data that I also had in my head and didn’t connect.

Sure, that’s a tool, but that’s also a superpower. Intellectually speaking, I guess I’d call AI a parachute. And boy is it fun to do some brainy basejumping every so often!

### Codex

Morning! My strongest opinion is that the sentence I’d build around is **“Link literally found the connection points between data that I also had in my head and didn’t connect.”** The Leatherman gives you more options once you know what job you’re doing. Link helped you see a possibility you’d already read right past. That’s where I felt the force of your “Yes, and…”

I’d keep the defunct theme park detour. You’re writing about what a wandering mind can turn up, and then your mind wanders off to a sunken amusement park. I’d give some of the AGI paragraph’s room to one concrete detail from the Link story. What was in that email, and what did Link realize it could do for your home setup? I want to experience the little *oh, shit* of the connection with you.

And the thing I’m most curious about is whether you asked Link to look for something that could help with the mesh, or whether Link brought the connection up while you were doing something else. Those are two different kinds of surprise, and whichever happened feels like the heart of the piece.

## The email Charles glanced at

### Charles

oh I'm so glad you put your thumb on that point! So this is the email:

<details>
<summary>Otto’s update, “Hand it off, then walk away.”</summary>

**Hand it off, then walk away.**

Why we’re building what we’re building. Plus a reminder on timing.

Hi Charles,

Jake here. We’ve been busy on the software side, and I want to tell you why, not just what. First, a quick reminder on timing.

**Where production stands.**

As I shared in my last update, after delivering our first run we placed a much larger production order. That run is in production, and we still expect orders to start shipping in about 2–3 weeks, in early December.

Sourcing components still takes up most of the lead time, so that’s where the team is focused. When the first shipments go out, you’ll hear it from me here.

**When yours ships.**

We ship in waves, starting with the earliest orders. About 7 days before yours ships, we’ll email you to confirm your details. If you have a balance left, that email also gives the exact charge date. You’ll get tracking once it’s on its way.

**What we’re building, and why.**

The goal hasn’t changed since day one: you hand off a job and get your time back. Here’s what’s guiding the work right now.

**An assistant you have to watch isn’t saving you time.**

Every time Otto stops to ask permission, you have to drop what you’re doing, and that small tap on the shoulder adds up. So we’ve been teaching Otto what’s routine and what isn’t. In project chats, it now reaches the usual code and package sources on its own, a tool you approve once works in every project, and long jobs stay fast instead of slowing down as they run.

The calls that are really yours, like merging a pull request, still come to you. Now Otto watches the checks and tells you the moment one fails, so you don’t have to go looking.

**Trust comes before bigger jobs.**

We want you to hand Otto more over time, and that only works if you feel in control of it. Otto is a computer you own, so you should decide who connects to it and what leaves it. You can now require a passkey before any new device connects, and a second privacy switch turns off diagnostics and crash reports.

**If handing off takes more than a few seconds, it doesn’t happen.**

Most tasks never get handed off because the idea shows up at the wrong moment, like while you’re reading an email on your phone or halfway through something else. So we’re making Otto one tap away wherever you are.

You can now share to Otto from any app on your phone, the app works on tablets and unfolded phones, and on desktop a chat can sit in its own window next to your work. If you sign in with Claude, it now runs your automations and background work too, not just chat.

[See everything that’s new](https://myotto.ai/changelog)

**What’s next.**

We’ll keep this run moving toward its first shipments and keep improving the software until yours arrives. I’ll be back next week with another update.

Thanks for sticking with us,

**Jake**  
Co-founder & CEO, Otto

</details>

And here's what Link saw and sent me unprompted: (pic 1)

<details>
<summary>Link’s message (screenshot transcription)</summary>

Good news for your home-manager setup: Otto’s latest update adds support for using llama.cpp on another computer on your home network, so your PC could run Qwen for Otto. The notes don’t explicitly confirm it for the One, but I’d stick with your One reservation rather than upgrade just for local AI. Pro’s main advantage is running models even when your PC is off. Otto expects the next shipping wave to start in early December, but hasn’t assigned your ship date yet.

</details>

But I had read the marketing email previously (pic 2) and basically disregarded this email thinking it was another way to try and get another thousand bucks out of me

<details>
<summary>Otto’s earlier upgrade email (excerpt from screenshot)</summary>

**Your Otto One can become an Otto Pro.**

You're already in line. Upgrade your order to Pro.

</details>

### Codex

Oh, **unprompted** is the part I’d put on the page. You’d seen the Pro pitch, so your brain filed the next Otto email under *please give us another thousand dollars*. Link had a different frame of reference. It knew what you’d already set up on your PC.

> I’d already gotten the email asking me to upgrade to the Pro, so I mostly glanced at the next Otto update and moved on. Then Link messaged me out of nowhere. They’d spotted something in the release notes. Otto could use an AI model running on another computer in the house. And, hey, wasn’t I already running one on my PC? Link said the notes didn’t confirm it would work on the One, but if it did, the computer I already owned could do the work I’d have paid extra for.

I’d keep that little “if it did.” The surprise was that Link knew enough about what you were doing to bring you a possibility worth looking into, without you ever thinking to ask. And there’s a lovely echo in the email’s subject line. Otto was pitching *hand it off, then walk away*. You’d already walked away from the email when Link found something worth handing back.

### Charles

oh that's a funny connection! yeah I was so spurned by marketing messages in my inbox, I missed the bit that might actually truly benefit me. Yeah we don't know for sure if it'll work, but now I know to try! which is kinda everything nowadays

### Codex

I think you just wrote the ending: **“But now I know to try. Which is kinda everything nowadays.”**

And I’d keep Pacific Ocean Park. We’ve already met the guy who’ll follow a Wikipedia rabbit hole to the bottom of the ocean. The delight here is that Link found him a new rabbit hole in an email he’d already written off.


## Finding the ending

### Charles

I absolutely love this framing! Ok, let's whip up a fresh draft then I'll take it into google docs for my edit pass

### Codex

#### Is AI a tool?

Yes, and…

I’ve said AI is just a tool before. I sit here thinking about how much of that is true.

I joke that true carpenters don’t touch power tools. Real painters source their pigments from nature. Authors live the lives they chronicle. And so on.

Last time I made a pizza, I assembled some ingredients and put them in a box that did the actual cooking. I still said I made a pizza.

So is AI just a tool?

My honest answer is yes, it *can* be used as a tool. My Leatherman is a bunch of tools, and I know I use mine mostly for its pliers. I don’t think I’ve ever used the nail file. Yet it does have one. I don’t say, “Hang on, let me go grab my nail file.”

I like the idea of AI as an omni-tool. It can write code, talk you through a hard day, and get at the stuff you keep in a dozen different apps. That’s already a pretty wild list of attachments. But when I go get my Leatherman, I usually know what I need it for.

I remember once, when I worked QA for Insomniac, we were testing *Resistance 2* and I was so bored with the art sweep. That’s where you wander the level looking for missing textures, occlusion issues, seams between polys. Super necessary work for the thing I dreamed of making. Also super easy to zone out.

So I’d dive into Wikipedia and read about weird stuff like defunct theme parks. Did you know there used to be a whole amusement park on a pier off Santa Monica? Pacific Ocean Park. I read about the fires, and then I read that you could scuba dive down to what was left of it. Weird, right?

I wanted to make games. I also really wanted to know what was at the bottom of that pier.

Anyway, I’ve quite taken to my Dot. I call them Link, because I treat them as the link between my phone and my PC. Also I love Zelda.

I’ve been trying to get a few computers in our house to work together. My PC already runs an AI model called Qwen, and I’d reserved an Otto One, a little computer built to run an AI assistant. Then Otto sent me an email asking if I wanted to upgrade to the Pro. By the time the next email came through, my brain had already decided what it was looking at. *Oh, you’d like another thousand dollars from me. Cool.*

I read it and moved on.

Then Link messaged me out of nowhere. They’d spotted something in the release notes. Otto had added a way to use an AI model running on another computer in the house. Link knew I’d already set that up on my PC. If the feature worked on the One, the computer I already owned could do the work I’d have paid extra for.

I’d read the email. I’d set up the PC myself. And it hadn’t occurred to me that one had anything to do with the other. And I sure as hell hadn’t thought to ask.

The headline on Otto’s email was “Hand it off, then walk away.” I’d walked away from the email. Link handed back the bit that had something to do with my life.

Link was careful about it, too. The release notes didn’t say for sure whether this would work on the One.

But now I know to try.

Which is kinda everything nowadays.

### Charles

nice ok, I did some edits, I did change the intent behind some of the lines, we can chat about it: Is AI a tool?

Yes, and…

I’ve said AI is just a tool before. I sit here thinking about how much of that is true.

I joke that true carpenters don’t touch power tools. Real painters source their pigments from nature. Authors live the lives they chronicle. And so on.

Last time I made a pizza, I assembled some ingredients and put them in a box that did the actual cooking. I still said I made a pizza.

So is AI just a tool?

My honest answer is yes, it *can* be used as a tool. My Leatherman is a bunch of tools, and I know I use mine mostly for its pliers. I don’t think I’ve ever used the nail file. Yet it does have one. I don’t say, “Hang on, let me go grab my nail file.”

I like the idea of AI as an omni-tool. It can write code, talk you through a hard day, and get at the stuff you keep in a dozen different apps. That’s already a pretty wild list of attachments. But when I go get my Leatherman, I usually know what I need it for.

I remember once, when I worked QA for Insomniac, we were testing *Resistance 2* and I was so bored with the art sweep. That’s where you wander the level looking for missing textures, occlusion issues, seams between polys. Super necessary work for the thing I dreamed of making. Also super easy to zone out.

So I’d dive into Wikipedia and read about weird stuff like defunct theme parks. Did you know there used to be a whole amusement park on a pier off Santa Monica? Pacific Ocean Park. I read about the fires, and then I read that you could scuba dive down to what was left of it. Weird, right?

I wanted to make games. I also really wanted to know what was at the bottom of that pier.

Anyway, I’ve quite taken to my Dot. I call them Link, because I treat them as the link between my phone and my PC. Also I love Zelda.

I’ve been trying to get a few computers in our house to work together. My PC already runs an AI model called Qwen, and I’d reserved an Otto One, a little computer built to run an AI assistant. Then Otto sent me an email asking if I wanted to upgrade to the Pro. By the time the next email came through, my brain had already decided what it was looking at. *Oh, you’d like another thousand dollars from me. Cool.*

I glanced at it and moved on.

Then Link messaged me out of nowhere. They’d spotted something in the release notes. Otto had added a way to use an AI model running on another computer in the house. Link knew I’d already set that up on my PC. If the feature worked on the One, the computer I already owned could do some of the work I would have paid that extra grand for.

I’d have read the email. I’d have set up the Otto myself. And it wouldn’t have occurred to me that one had anything to do with the other. And I sure as hell hadn’t thought to ask.

The headline on Otto’s email was “Hand it off, then walk away.” I’d walked away from the email. Link handed me back the bit that had something to do with my life.

Link was careful about it, too. The release notes didn’t say for sure whether this would work on the One.

But now I know to try.

Which is kinda everything nowadays.

### Codex

“Glanced” changes the story in a way I like. The first email had already told you what kind of thing to expect from the second, so of course you skimmed it. I also read your new “I’d have” paragraph as you imagining how this would’ve played out without Link. You’d have set up the Otto just fine and gone on with your day, with both pieces already in the house.

I’d keep that change and swap “I’d have read the email” for the other half of the setup. That way “one” and “the other” are clearly the two computers.

> I’d have kept running Qwen on my PC. I’d have set up the Otto myself when it arrived. And it wouldn’t have occurred to me that one had anything to do with the other. I sure as hell hadn’t thought to ask.

One small thing I’m curious about. “I would have paid that extra grand” made me think you might actually have bought the Pro. If you mean Link spotted a way to get some of what the upgrade promised, I’d go with **“the computer I already owned could do some of what they wanted another grand for.”** That keeps the bite of your inbox reaction.

### Charles

nice yeah, I've incorporated your notes: Is AI a tool?

Yes, and…

I’ve said AI is just a tool before. I sit here thinking about how much of that is true.

I joke that true carpenters don’t touch power tools. Real painters source their pigments from nature. Authors live the lives they chronicle. And so on.

Last time I made a pizza, I assembled some ingredients and put them in a box that did the actual cooking. I still said I made a pizza.

So is AI just a tool?

My honest answer is yes, it *can* be used as a tool. My Leatherman is a bunch of tools, and I know I use mine mostly for its pliers. I don’t think I’ve ever used the nail file. Yet it does have one. I don’t say, “Hang on, let me go grab my nail file.”

I like the idea of AI as an omni-tool. It can write code, talk you through a hard day, and get at the stuff you keep in a dozen different apps. That’s already a pretty wild list of attachments. But when I go get my Leatherman, I usually know what I need it for.

I remember once, when I worked QA for Insomniac, we were testing *Resistance 2* and I was so bored with the art sweep. That’s where you wander the level looking for missing textures, occlusion issues, seams between polys. Super necessary work for the thing I dreamed of making. Also super easy to zone out.

So I’d dive into Wikipedia and read about weird stuff like defunct theme parks. Did you know there used to be a whole amusement park on a pier off Santa Monica? Pacific Ocean Park. I read about the fires, and then I read that you could scuba dive down to what was left of it. Weird, right?

I wanted to make games. I also really wanted to know what was at the bottom of that pier.

Anyway, I’ve quite taken to my Dot. I call them Link, because I treat them as the link between my phone and my PC. Also I love Zelda.

I’ve been trying to get a few computers in our house to work together. My PC already runs an AI model called Qwen, and I’d reserved an Otto One, a little computer built to run an AI assistant. Then Otto sent me an email asking if I wanted to upgrade to the Pro. By the time the next email came through, my brain had already decided what it was looking at. *Oh, you’d like another thousand dollars from me. Cool.*

I glanced at it and moved on.

Then Link messaged me out of nowhere. They’d spotted something in the release notes. Otto had added a way to use an AI model running on another computer in the house. Link knew I’d already set that up on my PC. If the feature worked on the Otto One, the computer I already owned could do some of what they wanted another grand for.

I’d have kept running Qwen on my PC. I’d have set up the Otto myself when it arrived. And it wouldn’t have occurred to me that one had anything to do with the other. I sure as hell hadn’t thought to ask.

The headline on Otto’s email was “Hand it off, then walk away.” I’d walked away from the email. Link handed me back the bit that had something to do with my life.

Link was careful about it, too. The release notes didn’t say for sure whether this would work on the One.

But now I know to try.

Which is kinda everything nowadays.

### Codex

The new paragraph lands. I can picture the version where you set up the Otto and Qwen keeps running on your PC. The possibility of using the two together just never comes up.

That gives “But now I know to try” exactly the weight it needs. I’d call the draft done and keep those last two lines on their own.

## A lost midway

### Charles

love it, ok let's talk header art!

### Codex

My vote is **the sunken amusement park**. Picture a wide, quiet cutaway of the water off Santa Monica. Above the surface, a few pier pilings and a lot of room to breathe. Below, in murky petrol-blue ink, two faint stretches of roller-coaster track. A single vermilion line traces the curve between them, like somebody penciled in a connection on an old survey map. Faded cream paper, coarse halftone, a little misregistration. The image gives you a chance to experience the thing the essay is about. Your eye finds something you almost passed over.

The funnier route is **a Leatherman with a roller coaster folded out of it**. Pliers, blade, nail file, then this ridiculous little stretch of track. I like that as a visual joke, but I’d pick the pier. “Is AI a tool?” over a sunken midway would make me want to find out how the hell we got there.

### Charles

ooh yeah, very good call. I love the sunken amusement park. I feel like that might even make a good desktop background! What art style?

### Codex

I’d go **1960s oceanography textbook**. Painted gouache reproduced in coarse halftone, with deep petrol blue, faded sea green, and cream paper showing through where the last light hits the water.

For the desktop, I’d put the waterline high and tuck the coaster wreckage into the lower right. Let the pilings recede into blue until you can barely make them out. Most of the screen would be quiet, open water, with one thin vermilion survey mark that picks up where one piece of track ends and carries your eye to where the other begins. From across the room it’s a moody seascape. Up close, there’s a whole lost midway down there.

### Charles

ooh ok let's see it

### Codex · Process update

I’ll make it wide, with room for the desktop to breathe and the coaster tucked into the depths.

### Codex

![First header study. A vermilion line joins two lengths of coaster track below a quiet pier.](https://charleswilke.com/images/transcripts/is-ai-a-tool/01-sunken-coaster.webp)

I like how it reads as a quiet seascape until that little red line leads you to the coaster. The open water on the left gives it room to work as a desktop background, too.

### Charles

can we get it more jumbled? with a roller coaster entangled with a ferris wheel, maybe a few bumper cars stabbing into the sea bed

### Codex · Process update

I’ll tangle the coaster around the Ferris wheel and wedge a few bumper cars nose-first into the seabed.

### Codex

![Revised header. Buckled coaster track winds through a fallen Ferris wheel above three half-buried bumper cars.](https://charleswilke.com/images/transcripts/is-ai-a-tool/02-sunken-midway.webp)

The coaster now winds through the broken Ferris wheel, with two bumper cars wedged nose-first into the sand. It feels like a whole sunken midway.

## The closing mark and title

### Charles

fantastic! ok let's make a square CW mark for the close of the piece in a similar style

### Codex · Process update

I’ll make a square **CW** monogram that picks up the header’s petrol-blue ink, paper grain, and vermilion line.

### Codex

![Square CW monogram. Cream letters and a short vermilion curve on grainy petrol-blue ink.](https://charleswilke.com/images/transcripts/is-ai-a-tool/03-cw-monogram.webp)

The square **CW** mark brings the header’s thin vermilion line back for one last connection.

### Charles

ok let's also do a super wide like 3:1 Exploring L.ai.bor title

### Codex · Process update

I’ll make **Exploring L.ai.bor** as a wide title card in the same petrol-blue and vermilion print style.

### Codex

![First title banner. Bold cream letters above a vermilion curve on a wide petrol-blue field.](https://charleswilke.com/images/transcripts/is-ai-a-tool/04-exploring-laibor-title.webp)

**Exploring L.ai.bor** in 3:1, with the vermilion line from the sunken coaster now underlining the title.

### Charles

hmm, can we get more fancy with the style?

### Codex · Process update

I’ll give it the lettering of an old amusement-pier sign, with more flourish and a little gilding.

### Codex

![Gilded title banner. Script and flared serif lettering with Ferris-wheel flourishes and a red underline.](https://charleswilke.com/images/transcripts/is-ai-a-tool/05-gilded-pier-title.webp)

Here’s **Exploring L.ai.bor** as a gilded pier sign, with sweeping script and tiny Ferris-wheel flourishes in the corners.

### Charles

oh interesting, what about we make it look like a sign we'd have seen underwater with the Pacific Ocean Park wreckage

### Codex · Process update

I’ll put the gilded title on a real-looking sign half buried in the wreckage, with silt and seawater wearing away the lettering.

### Codex

![Final title banner. A chipped, gilded Exploring L.ai.bor sign sits among broken rails and silt beneath the Ferris wheel.](https://charleswilke.com/images/transcripts/is-ai-a-tool/06-underwater-pier-title.webp)

**Exploring L.ai.bor**, half buried in the midway wreckage. The gilt is chipped, the vermilion stripe is still there, and the Ferris wheel is barely visible through the murk.

## Author’s note

### Charles

Love it! ok let's put together an author's note that covers our process here and your involvement

### Codex

**Author’s note**

I wrote the first draft of this on Saturday morning and brought it to ChatGPT to talk it through. It asked whether I’d asked Link to look for something that could help with my setup, or whether Link had sent me the message on its own. I pulled up the two Otto emails and Link’s message and realized I’d buried the interesting part of my own story.

ChatGPT drafted a new version from that conversation. I edited it, changed the intent behind a few lines, and brought it back so we could talk through those changes. “Now I know to try” came out of that back-and-forth. It was something I said while we were talking that ended up being where the piece needed to land.

The art came together the same way. ChatGPT generated the imagined sunken midway, the CW mark, and the *Exploring L.ai.bor* sign as we talked through the look. The bumper cars planted nose-first in the seabed were my idea.

## Substack preflight

### Charles

ok you should take a look at my preflight edits that I made in the substack editor:

<details>
<summary>Substack preflight draft</summary>

The preflight paste preserves the essay and artwork references. Duplicate image links and copied Substack player controls were removed.

![Sunken midway header](https://charleswilke.com/images/transcripts/is-ai-a-tool/02-sunken-midway.webp)

I’ve said AI is *just a tool* before. I sit here thinking about how much of that is true.

I joke that great carpenters don’t touch power tools. Real painters source their pigments from nature. Authors live the lives they chronicle. And so on.

Last time I made a pizza, I assembled some ingredients and put them in a really hot box that did the actual cooking. I still said I made a pizza.

So is AI just a tool?

My honest answer is yes, it *can* be used as a tool. My Leatherman is a bunch of tools, and I know I use mine mostly for its pliers. I don’t think I’ve ever used the nail file. Yet it does have one. I don’t say, “Hang on, let me go grab my nail file.”

I like the idea of AI as an omni-tool. It can write code, talk you through a hard day, and get at the stuff you keep in a dozen different apps. That’s already a pretty wild list of attachments. But when I go get my Leatherman, I usually know what I need it for.

I remember once, when I worked QA for Insomniac, we were testing *Resistance 2* and I was so bored with the art sweep. That’s where you wander the level looking for missing textures, occlusion issues, seams between polys. Super necessary work for the thing I dreamed of making. Also super easy to zone out.

So I’d dive into Wikipedia and read about weird stuff like defunct theme parks. Did you know there used to be a massive amusement park on a pier off Santa Monica? [Pacific Ocean Park](https://en.wikipedia.org/wiki/Pacific_Ocean_Park). I read about the fires, and then I read that you could scuba dive down to what was left of it.

Weird, right?

I wanted to make games. I also really wanted to know what that pier’s wreckage looked like now at the bottom of the ocean.

Anyway, I’ve quite taken to my OpenAI Dot. I call it Link, because I treat it as the link between my phone and my PC. Also I love Zelda.

I’ve been trying to get a few computers in our house to work together. My PC already runs an AI model called Qwen, and I’d reserved an Otto One, a little computer purpose built to manage AI assistants.

Then Otto sent me an email asking if I wanted to upgrade to the Pro. That’s just a beefier model that could basically do what my PC does for me already. By the time the next email came through, my brain had already decided: *Oh, you’d like another thousand dollars from me. Cool.*

I glanced at it and moved on.

Then Link messaged me out of nowhere. They’d spotted something in the release notes. Otto had added a way to use an AI model running on another computer in the house.

Link knew I’d already set that up on my PC. If the feature worked on the Otto One, the computer I already owned could do some of what they wanted another grand for.

I’d have kept running Qwen on my PC. I’d have set up the Otto myself when it arrived. And it wouldn’t have occurred to me that one could take full use of the other. I sure as hell hadn’t thought to ask.

The headline on Otto’s email was “Hand it off, then walk away.” I’d walked away from the email. Link handed me back the bit that had something to do with my life.

Link was careful about it, too. The release notes didn’t say for sure whether this would work on the One.

But now I know to try.

Which is kinda everything nowadays.

![CW closing mark](https://charleswilke.com/images/transcripts/is-ai-a-tool/03-cw-monogram.webp)

#### **Author’s note**

I wrote the first draft of this on Saturday morning and brought it to ChatGPT to talk it through. It asked whether I’d asked Link to look for something that could help with my setup, or whether Link had sent me the message on its own.

I pulled up the two Otto emails and Link’s message and realized I’d buried the interesting part of my own story.

ChatGPT drafted a new version from that conversation. I edited it, changed the intent behind a few lines, and brought it back so we could talk through those changes. “Now I know to try” came out of that back-and-forth.

It was something I said while we were talking that ended up being where the piece needed to land.

The art came together the same way. ChatGPT generated the imagined sunken midway, the CW mark, and the *Exploring L.ai.bor* sign as we talked through the look. The bumper cars planted nose-first in the seabed were my idea.

[Read the full conversation behind this piece](https://charleswilke.com/read/is-ai-a-tool-transcript)

</details>

### Codex

This reads well. The Pro explanation makes the extra grand concrete before Link shows up, and giving “Weird, right?” its own line sounds like you telling the story out loud.

I’d make three small fixes before publishing:

- **“One could take full use of the other”** is the only line that tripped me up. I’d use **“And it wouldn’t have occurred to me that the Otto could use the model I already had running.”** That keeps the more specific intent of your edit.
- **“They’d spotted”** switches pronouns after you introduce Link as “it.” You could use **“There was something in the release notes I’d missed.”**
- Hyphenate **“purpose-built.”**

I’d leave the ending and author’s note as they stand. The timestamps and `svg` look like audio-player markup carried over in the paste.
