Why we are doing this

We are testing our own method. On our own garden. With our own money.

Collab365 Spaces is built on a claim: that if you start from the things actually stopping people, rather than from a curriculum, you produce help people finish and use. This project is us testing that claim somewhere we cannot quietly fudge the result.

See what went wrong What Collab365 Spaces is

The claim being tested

Generic content fails because the work is not generic.

That is the sentence the whole of Collab365 Spaces rests on. A Space does not start by asking what we could teach. It starts by asking what is stopping people, names those blockers one at a time, and then packages an answer scaled to the size of the fix rather than to the length of a course.

Briefings

When the fix is a check, a rule or a decision, it should take ten minutes, not a module. Ours includes a ten-second test for whether an AI garden image contains a single real measurement.

Courses

When the fix genuinely needs depth, it gets depth — and ends with something real. Ours ends with a photorealistic view of your own garden, built from your own measurements.

Blueprints

When the honest answer is that a tool should exist, we write the build spec — screens, data model, acceptance tests — so an AI coding tool can build it. We have twelve so far, all from real blockers.

Boards

Individually, a briefing answers one question. A Board turns the briefings, courses and blueprints into an ordered path someone can actually finish — in our case, the route from an empty scruffy garden to a built one, in the order the work has to happen. That ordering is the part you cannot get from searching, because search returns answers and gives you no idea which one you need first.

Why a garden

A garden is not a typical business project. It is a better one to watch.

Every hard thing about running a project at work is present in a domestic garden build, with one difference: you can see it. Nobody wants to follow along with a SharePoint migration. Everybody understands a patio.

In our garden	In your work
A brief nobody wrote down properly	A vague requirement everyone interprets differently
A budget that moves the moment you price it	A budget set before anyone scoped the work
Regulations you find out about too late	Compliance discovered after the design is signed off
A sequence where one mistake gets paved over	Dependencies where one wrong order costs a rebuild
Suppliers quoting from a drawing with no dimensions	Vendors quoting from a spec nobody validated
Decisions you cannot undo once the concrete is down	Decisions you cannot undo once it is in production

The AI fails in the same places too. It loses the context you gave it last week. It answers confidently about things it has no way of knowing. It produces something that looks finished and is not. Those are not gardening problems. They are the problems, and a garden simply makes them visible.

Why it is honest

The garden either gets built or it does not.

This is the part we could not get from a case study or a pilot. A demo can be quietly abandoned when it stops going well. A garden cannot: it is at the back of our house, it is our money, and if the method produces something unbuildable we pay for it in concrete.

It is ours

Our home and our savings. That makes every reality check mean something a hypothetical exercise never could.

Nothing is edited out

Every problem is published whether or not it makes us look competent, including the ones caused by trusting an AI answer we should have checked.

The proof is a photograph

Every render gets re-photographed from the same position once it is built, with what differs written underneath. That is a standard we cannot argue our way around.

Who we are

Mark and Helen, and a scruffy back garden.

We run Collab365, which builds Spaces for people navigating AI at work. We are not landscapers, designers or builders, and we have never run a construction project. That is the point: if the method only works for experts it is not a method, it is expertise.

If you are here for the garden

The method is why the answers here are better than a forum thread. Every problem is researched against UK sources, checked, and honest about where it needs a professional instead of a prompt. Follow along, and take whatever is useful.

If you are here for the method

The garden is the proof it survives contact with reality. Watch the loop run in public — blocker, research, check, package, publish — and take it back to a project of your own that is nothing to do with plants.

Start free See how Spaces works for teams

The name

Plan it. Prompt it. Plant it.

The AI Garden Project borrows the spirit of vibe coding: using AI to make real progress in a domain where you are not the expert. The difference, and the whole reason this is worth watching, is that soil, weather, dimensions, budgets and building regulations cannot be prompted away.