ISSUE NO. 03

Build Projects,
Not Endless Chats.

Give important work a permanent home.

A useful AI project is not just a folder full of conversations. It is a working environment with enough stable context for the AI to begin further ahead.

ADAM JUKES FOUNDER, ONE BLADES

READ THIS IF YOU ONLY HAVE TWO MINUTES

03

A useful AI project
is not just a folder
full of conversations.

It is a working environment.

ONE 2026 works because the AI does not need me to explain the whole company from zero every time.

The project can contain stable context such as:

01

Purpose

02

Current goal

03

Source of truth

04

Priorities

05

Constraints

06

Non-negotiables

07

Terminology

08

Decisions already made

09

Retired ideas

10

Current state

11

Communication rules

Different areas should have different projects.

The ONE business project does not need to be the same environment as a personal training project.

The important part is that you do not need to maintain these project documents manually.

When something changes:

01

Give AI the current file.

Upload the current Markdown source.

02

Tell it what changed.

Explain the new facts, decisions or direction.

03

Find what is outdated.

Ask AI to identify what is now wrong, missing or incomplete.

04

Resolve ambiguity.

Let it ask the questions needed before changing the source.

05

Create the new version.

Generate a complete replacement document.

06

Replace the old file.

Save the updated `.md` as the new source.

The AI should help maintain
the context it relies on.

THE WHOLE IDEA

Do not become the administrator of your AI.

Build a system the AI can help maintain.

END OF QUICK VERSION CONTINUE FOR THE FULL PAPER ↓

Give important work
a permanent home.

As AI use becomes more serious, the problem changes. The challenge is no longer only getting one useful answer. It is maintaining enough good context that the AI can keep helping without you constantly repeating yourself.

01

Eventually the problem
becomes context.

As AI use becomes more serious, the problem changes.

At first, the problem is getting a useful answer.

Later, the problem is maintaining enough good context that the AI can keep helping without you constantly repeating yourself.

Projects solve part of that problem.

Stop making every conversation
begin from zero.

02

A real
ONE example.

ONE Blades now contains many connected areas:

01

EDGE

02

ONE Academy

03

Creators

04

Coaching

05

Content

06

Operations

07

Product development

08

Narrative

09

Current priorities

If every new conversation required me to explain all of that again, AI would create more work than it saved.

Instead, the project has a defined operating context.

It knows that Skater First matters.

It knows which terminology is approved.

It knows which systems have been retired.

It knows the phase model.

It knows what is currently public and what remains private.

It knows that the live store outranks an old document if the two conflict.

That means a conversation can begin much further ahead.

Context compounds.
Repetition does not.

03

What belongs in
a project file?

A strong project context usually contains the following.

01

Purpose

Why does this project exist?

02

Goal

What outcome are we trying to create?

03

Current state

What is true now?

04

Source of truth

If sources conflict, which one wins?

This is extremely important.

05

Priorities

What matters most right now?

06

Constraints

What cannot be ignored?

Budget, time, technical limitations, legal requirements, operational reality and personal capacity.

07

Non-negotiables

Which principles must every recommendation respect?

08

Decisions already made

What has already been settled?

09

Retired ideas

What should stop coming back unless genuinely new evidence appears?

10

Terminology

Which words are correct?

Which words should not be used?

11

Communication rules

How should the AI work inside the project?

At ONE, the live store outranks an older planning document.

Without that rule, the AI might treat an old file as current simply because it contains detailed information.

A project should explain
not only what it knows,
but how that knowledge
should be interpreted.

04

Do not write
the whole thing yourself.

People often make AI systems harder than they need to be by turning documentation into homework.

You can build the project through conversation.

Tell the AI:

I want to create the operating context
for this project.

Interview me until you understand
the purpose, current state, priorities,
constraints, rules, terminology
and decisions already made.

Then create the document.

Speak the answers.

Let the AI structure them.

Save the result.

05

Updating the project
should also be conversational.

This is one of the most important workflow improvements.

Suppose the current project file is Version 4.

Over the next month, ten things change.

Do not open the file and manually hunt for every sentence that needs editing.

Start a conversation.

Upload Version 4.

Then say:

Read the current project file.

Based on what we have decided since it was written, identify everything that needs to change.

Ask me questions where the correct update is unclear.

Do not rewrite yet.

The AI produces the update list.

You answer the questions, preferably by voice if that is faster.

Then say:

Now create Version 5 as a complete replacement Markdown document.

Preserve everything that remains valid.

Remove anything retired.

Resolve contradictions.

Download the file.

Replace Version 4.

Now the source stays current without you becoming a document editor.

CURRENT SOURCE PROJECT_v4.md
CONVERSATION Explain what changed
AI REVIEW Find outdated, missing and conflicting context
QUESTIONS Resolve ambiguity
REPLACEMENT SOURCE PROJECT_v5.md

Do not become
the administrator
of your AI.

06

One source.
Many AI systems.

This workflow has another advantage.

The project file can be used in more than one AI system.

You may prefer one model for deep analysis and another for a particular task.

The project knowledge does not have to be rebuilt from scratch in each environment.

THE INTELLIGENCE LAYER

The context files you control.

Purpose, current state, decisions, terminology, rules and source material.

THE MODEL

The engine temporarily reading them.

Use the model that best fits the work you are doing.

The intelligence layer
is the collection of context files
you control.

The model is the engine
temporarily reading them.

07

The same structure works
outside ONE.

A COACH

Coaching project

Coaching philosophy.

Athlete profiles.

Lesson structure.

Technical terminology.

Current goals.

Injury constraints.

A CREATOR

Creator project

Audience.

Formats.

Tone.

Publishing principles.

Current campaigns.

Things already tested.

YOUR LIFE

Personal project

Annual goals.

Current commitments.

Financial constraints.

Health priorities.

Non-negotiables.

The principle is simple.

Stop making important work
depend on one long chat
remembering everything.

Give the work a home.

Give one important area
a permanent home.

Choose one area where you regularly use AI and stop rebuilding the context from zero.

01

Choose the project.

Business, coaching, content, training, personal planning or another important area.

02

Define its purpose.

Why does this project exist and what outcome are you trying to create?

03

Capture the current reality.

Current state, priorities, constraints and source of truth.

04

Add the rules.

Non-negotiables, terminology, decisions already made, retired ideas and communication rules.

05

Let AI build the file.

Talk through the context and let AI structure it into a clean source document.

06

Maintain it through conversation.

When reality changes, provide the current file, explain the change, resolve ambiguity and create the replacement.

THINK WITH AI · ISSUE NO. 03

Stop making important work
depend on one long chat
remembering everything.
Give the work a home.

Build a system the AI can help maintain.

NEXT ISSUE 04 / 07

ISSUE NO. 04

Give AI Real
Source Material.