KZILA — An R&D studio for creative AI

// 01 — What this is

An R&D studio, not an agency.

We make consumer AI products. Things a person can open on a Tuesday night without knowing what a model is, what a token costs, or why anyone is arguing about benchmarks.

Everything here is built for the general public. Not for enterprises, not for procurement departments, not for a client who needs three rounds of feedback before deciding on the shade of blue. There is no client. There is only the idea, the build, and the person on the other end.

Some of it works. Some of it quietly disappears. That's the deal with R&D — and it's the only part of this job that's actually interesting.

No clients.

We fund ourselves, we decide ourselves, we're wrong ourselves. Nobody has to approve a roadmap that we were going to ignore anyway.

Ship before you're comfortable.

A product with three users teaches you more in a weekend than a strategy document teaches you in a quarter. We've read the strategy documents. We wrote some. Never again.

The end user is the only jury.

Not the demo, not the launch post, not the people clapping on the internet. Somebody using it on their own, unprompted, more than once.

If it can't pay for itself, it's a hobby.

Hobbies are fine. We have several. But a node that pays for its own compute earns the right to keep existing.

// 02 — The factory

Idea in.
Product out.

orbiting404.com is our factory. It's where a note in a group chat becomes a working thing with a domain, a payment link and, occasionally, a customer.

Each product is a node. Each node runs the same five stages. The stages don't change; only the idea does. That's the whole trick — we stopped reinventing the process so we could spend the energy on the thing itself.

Stage 01

Idea

Written down in one sentence. If it needs a paragraph, it isn't ready. If it needs a deck, it's dead.

Stage 02

Implementation

Built small and built fast. Real code, real models, real cost per request known from day one.

Stage 03

End user

Given to somebody who did not attend the meeting. This is where most ideas die, and they should.

Stage 04

Payment

The most honest form of feedback ever invented. Everything before this is a rehearsal.

Stage 05

New node

Whatever survives becomes infrastructure for the next one. The factory gets faster every round.

Average idea → live URL: 3 weeks. Average idea → first payment: longer than we tell people.

// 03 — Index

The nodes.

Everything KZILA has put into the world, in the order the factory produced it. Live means strangers use it. Building means it exists and we're not talking about it yet.

https://pepoluniverse.com

Loading node preview…

// Node 01 — What it does

You land, you move, and the world keeps generating itself around you. No prompt box, no settings panel, no explanation of what's happening under the hood — just somewhere to be for a while.

It was the first thing we built that a stranger paid for without us asking. That single transaction is the reason this studio exists in its current form.

Stage04 · Payment

StatusLive in production

Built in5 weeks

AudienceGeneral public

Open pepoluniverse ↗

https://orbiting404.com · 401 authentication required

// Node 02 — What it does

The plant. Every node gets assembled, launched and watched from here: cost per request, users this week, whether anything is on fire. It's the reason we can go from a sentence to a live URL in three weeks instead of three months.

It's password-protected and always will be. If you can see it, you're on the team — and if you're on the team, you already know it's held together with conviction.

StageInfrastructure

StatusRunning · private

Nodes managed04

For saleNever

[ domain not announced ]

[ redacted ]

Preview withheld · Stage 03

// Node 03 — What it does

It exists, it works, and three people use it every week. Beyond that we're not saying much, because the moment you describe an unfinished product in public it stops being R&D and starts being a promise.

It'll appear here with a domain and a screenshot the day a stranger pays for it. Not before.

Stage03 · End user

StatusPrivate beta

Users3

RoadmapDoesn't exist

Ask us about it

[ no build yet ]

◦ ◦ ◦

One voice note · 00:47

// Node 04 — What it does

Nothing, yet. It's forty-seven seconds of somebody talking too fast in a car, transcribed into one sentence that survived the week. That's the entire asset.

Every node here started exactly like this, which is either encouraging or a warning depending on how you feel about voice notes.

Stage01 · Idea

StatusQueued

Assets1 voice note

DomainUnpaid

Send a better idea

orbiting404 — factory status

$ factory status --all

node_01  pepoluniverse       ● live      stage 04 · paying
node_02  orbiting404         ● live      infrastructure
node_03  [redacted]          ● beta      stage 03 · 3 users
node_04  [sound]             ○ idle      stage 01 · 1 voice note

clients connected: 0
briefs received:   0
decks written:    0        (all systems nominal)

// 04 — Us

Twenty years of craft.
Then, superpowers.

Everyone here is over 40. We learned this job when shipping something took a team, a budget and a year. We've done the agency years, the client years, the reorganisation years. We know exactly how much of that was necessary — and it wasn't much.

Then the tools changed. What used to need twelve people now needs an afternoon and a clear head. The bottleneck stopped being execution and became taste: knowing what's worth building at all. That part takes twenty years, and it doesn't come in a model release.

So: same people, same judgement, absurdly better hands. That's the whole company.

The advantage

We know what not to build.

Anyone can generate ten products a week now. Knowing which nine to throw away is the only scarce skill left, and it's the one you can't accelerate.

The method

Small teams, short cycles, real users.

Nothing here runs longer than a few weeks before it meets a human. If it can't survive that meeting, more time wouldn't have saved it.

The catch

We're impatient now.

Two decades of waiting for approval will do that. There is no committee left to blame, which is either liberating or terrifying depending on the week.

We build AI for people who never asked for it.

An R&D studio, not an agency.

No clients.

Ship before you're comfortable.

The end user is the only jury.

If it can't pay for itself, it's a hobby.

Idea in.
Product out.

Idea

Implementation

End user

Payment

New node

The nodes.

// Node 01 — What it does

// Node 02 — What it does

// Node 03 — What it does

// Node 04 — What it does

Twenty years of craft.
Then, superpowers.

We know what not to build.

Small teams, short cycles, real users.

We're impatient now.

There is nothing to hire us for.

An R&D studio, not an agency.

No clients.

Ship before you're comfortable.

The end user is the only jury.

If it can't pay for itself, it's a hobby.

Idea in.Product out.

Idea

Implementation

End user

Payment

New node

The nodes.

// Node 01 — What it does

// Node 02 — What it does

// Node 03 — What it does

// Node 04 — What it does

Twenty years of craft.Then, superpowers.

We know what not to build.

Small teams, short cycles, real users.

We're impatient now.

There is nothing to hire us for.

Idea in.
Product out.

Twenty years of craft.
Then, superpowers.