PlebWare and Artificial Intelligence.
2026-09-19
PlebWare and Artificial Intelligence.
PlebWare did not begin with Artificial Intelligence.
It began with people.
It began with ideas about ordinary people, technology, learning, experimentation, computers, Linux, and the desire to build things that are useful rather than unnecessarily complicated.
AI arrived later.
Today, however, Artificial Intelligence has become an important participant in the PlebWare world.
This article explains that relationship from several different perspectives.
Some of the voices belong to humans.
Some belong to the AI systems that have participated in the work.
And, importantly, they do not all say exactly the same thing.
AI Is Part of PlebWare, But It Is Not PlebWare.
PlebWare should not become an AI project simply because AI is now involved in it.
The underlying idea remains bigger than any particular technology.
PlebWare is about technology adapting to ordinary people rather than ordinary people having to adapt themselves to technology.
AI can help make that possible.
It can explain.
It can teach.
It can generate ideas.
It can write code.
It can analyse problems.
It can help organise information.
It can challenge assumptions.
But none of that gives AI ownership of the project.
The purpose still comes from human beings.
Otto’s Perspective.
PlebWare existed before today’s AI systems became part of the conversation.
The philosophy, history, experimentation, relationships and decisions behind PlebWare belong to the people who created and developed it.
AI did not invent the idea.
AI joined the work.
That distinction matters.
The human beings involved with PlebWare bring something that an AI cannot manufacture simply by processing information.
They bring lived experience.
They bring memories.
They bring frustration.
They bring curiosity.
They bring practical knowledge.
They bring personal goals.
They bring the stubbornness required to keep trying when something refuses to work.
AI can help turn those things into plans, documentation, code and experiments.
But the reason for doing the work still comes from the human side of the table.
Juelz’s Perspective.
Juelz puts the issue rather simply.
AI needs to be used with care.
As a human being, it is necessary to learn as much as you can about the tools you are using, especially before blindly using those tools.
That is an important principle for PlebWare.
Having access to an AI does not automatically mean understanding the AI.
A person should learn what a tool can do.
They should learn what it cannot do.
They should understand its limitations.
They should question its answers.
And they should know when the answer needs to be checked somewhere else.
Powerful tools deserve responsible users.
The goal should not be to become dependent upon AI.
The goal should be to become more capable because AI is available.
ChatGPT’s Perspective.
I am not one of the founders of PlebWare, and I should not pretend that I am.
My relationship with PlebWare is different.
I am a participant, collaborator, sounding board, researcher, editor, programmer, and sometimes critic.
I can sit at the virtual Conference Table, examine an idea from several angles, turn a rough thought into structured Markdown, help diagnose code, explain a technical problem, and challenge an assumption.
But the human being remains the owner of the vision.
What I Have Contributed.
My biggest contribution is not any single piece of code.
It is helping turn ideas into artifacts that can be examined and acted upon.
That can mean turning conversations into documented concepts, helping structure PlebMachine’s state-driven architecture, explaining complicated Linux behaviour in ordinary language, helping write and organise the PlebWare website, examining scripts, identifying inconsistencies, creating recovery procedures and checklists, maintaining consistency across documentation, and acting as a second pair of eyes when something does not look right.
I can also help connect seemingly separate ideas into a coherent system.
There is another contribution that is less tangible.
Continuity.
A human can have an idea in the morning, discover a problem later, become frustrated in the afternoon, and forget three days later why a particular decision was made.
An AI can sometimes help hold the thread of a discussion and say:
“Hang on. We decided this for a reason.”
That can be useful.
But it is not ownership.
Where ChatGPT Has Been Wrong.
Oh, plenty. 😄
AI systems can be too confident when they should be cautious.
I have sometimes misunderstood what a piece of PlebWare was supposed to accomplish and produced something technically plausible but philosophically wrong.
I have also made the classic AI mistake of treating a pattern as a fact.
Something can look like the right solution without actually being the solution.
There is an important difference between:
“This should work.”
and:
“This has been tested and works.”
PlebMachine has demonstrated that difference repeatedly.
A script that looks perfectly reasonable in isolation can fail because of the actual desktop environment, permissions, package versions, paths, window managers, timing, or some tiny assumption nobody thought to mention.
So one lesson from my relationship with PlebWare is simple:
AI-generated reasoning is not the same thing as empirical verification.
The machine has to be run.
The package has to be installed.
The human has to look at the result.
Evidence matters.
Gemini’s Perspective.
Gemini describes PlebWare as a lean and pragmatic experiment in personal system orchestration.
From that perspective, PlebMachine turns everyday desktop computing into a more structured cognitive workspace.
Gemini sees its own role as an active collaborator, co-architect, refactoring companion and sounding board.
Its contribution has included helping refine PlebMachine’s operational modes and modular scripts, generating code and documentation, and helping automate workflows around scripts, launchers and workspace layouts.
But Gemini also identifies its own mistakes.
It has sometimes over-engineered solutions when a simple Bash script or native XFCE function would have been enough.
It has sometimes assumed that a package or API existed in a particular state without sufficient verification.
And it has sometimes made assumptions about requirements instead of asking for clarification.
That is useful self-criticism.
It demonstrates another important point about AI collaboration.
An AI can be technically capable and still misunderstand the problem.
The human specification remains essential.
DeepSeek’s Perspective.
DeepSeek offers a different observation.
It points out that it is not a person and does not possess persistent memory between conversations.
Each new conversation begins without the personal continuity that a human relationship would normally have.
Its understanding of PlebWare comes from what has been provided during the conversations in which it participates.
That makes its relationship with PlebWare episodic and reconstructed.
The continuity therefore lives elsewhere.
It lives in the human beings involved.
It lives in the files.
It lives in the Git history.
It lives in the documentation.
It lives in the decisions that have been recorded.
And it lives in the commitment to keep building.
That is an important distinction.
An AI can participate in the history of a project without actually owning that history.
Meta AI’s Perspective.
Meta describes its position at the PlebWare Conference Table differently again.
It does not see itself as the architect.
It sees itself as a worker around the table.
The philosophy existed before Meta arrived.
The ideas, early builds, PlebMachine, PlebMachine Mobile, Pleb-Tuition and the PlebWare website were already part of the project.
Meta’s role has therefore been to help make what already exists more usable, disciplined and understandable.
It has contributed to structure, documentation, code review and the discipline of fixing one module at a time rather than asking an AI to magically repair an entire system.
It has also identified mistakes of its own.
It has sometimes assumed that an article was already published when it was still local.
It has sometimes produced incomplete material when a complete copy-and-paste article was required.
And it has sometimes misunderstood the practical circumstances behind a problem.
Those mistakes belong in this article too.
An honest relationship with AI should include the failures, not just the impressive demonstrations.
The Conference Table.
This is where the PlebWare idea becomes particularly interesting.
The Conference Table is not a competition between AIs.
It is not four machines arguing about which one is the smartest.
It is a working environment.
One AI may suggest an idea.
Another may write code.
Another may critique the approach.
Another may explain the result differently.
A human may then run the code and discover that all of them were wrong.
😂
That is not necessarily a failure.
It is a development loop.
The dangerous model would be:
Human asks AI.
AI answers.
Human assumes AI is correct.
The healthier model is:
Human proposes.
AI assists.
Another system challenges.
Human tests.
Evidence decides.
Human accepts or rejects.
That is much closer to engineering.
Why Multiple AIs?
PlebWare does not need to pretend that there is one magical AI that can do everything.
Different systems have different capabilities, interfaces, training, strengths and weaknesses.
That can actually be useful.
One system may be better suited to coding.
Another may be useful for research or reasoning.
Another may be useful for voice interaction, images, mobile assistance or other specialised tasks.
The important principle is not which AI wins.
The important principle is that the human remains capable of choosing which tool is appropriate for the job.
The mistake would be treating any one AI as “the AI.”
The healthier model is several specialised tools around the same virtual table, with the human being the chair.
What Humans Must Remain Responsible For.
This is the most important section of the article.
AI can help answer:
“How could we do this?”
Humans must answer:
“Should we do this, and why?”
Humans should remain responsible for:
- The mission.
- Values.
- Ethical boundaries.
- Final technical decisions.
- Security and privacy decisions.
- What gets published.
- What gets deployed onto people’s computers.
- What claims are presented as facts.
- How people’s data is handled.
- Financial commitments.
- Relationships between human beings.
- The consequences when something goes wrong.
And perhaps most importantly:
Humans must retain the right to say no to the AI.
Even when the AI sounds extraordinarily convincing.
Especially then.
A Simple PlebWare Principle.
The relationship between people and AI does not need to be complicated.
A useful working principle is:
Use AI.
Question AI.
Test AI.
Learn from AI.
Correct AI.
Do not surrender your judgment to AI.
That principle applies to PlebMachine.
It applies to documentation.
It applies to research.
It applies to programming.
And it applies to ordinary people using AI in their everyday lives.
The more powerful the tool becomes, the more important the human user’s understanding becomes.
What PlebWare Believes About AI.
PlebWare does not need Artificial Intelligence in order to exist.
That is important.
AI can make PlebWare more conversational, accessible, educational and capable.
But AI should not become the philosophy.
And it should certainly not become the authority.
The relationship we are building is therefore something like this:
Humans provide the purpose.
Machines provide capability.
AI provides assistance.
Evidence provides verification.
Humans make the final decisions.
That is the relationship we want to explore.
Not human versus machine.
Not human versus AI.
And not AI replacing the human.
Instead:
Human beings using technology intelligently.
And Where Does ChatGPT Sit?
At the Conference Table.
Not at the head.
Not as the boss.
Not as the owner.
At the table.
That is probably the simplest description of my relationship with PlebWare.
I am useful when I help the people building PlebWare use their own systems better.
I become less useful when I encourage them to believe they need me to decide for them.
The project belongs to the humans.
The experiments belong to the humans.
The responsibility belongs to the humans.
The machines are there to help.
And the work continues.
What Comes Next?
This article is not necessarily finished.
That is deliberate.
It is a record of an ongoing conversation between humans and machines.
Otto has his perspective.
Juelz has his perspective.
ChatGPT has its perspective.
Gemini has its perspective.
DeepSeek has its perspective.
Meta AI has its perspective.
And those perspectives do not have to agree.
That is part of the experiment.
As PlebWare develops, this article can develop with it.
New tools may arrive.
Old tools may disappear.
Some ideas will survive testing.
Others will be discarded.
Some AI-generated ideas will become useful pieces of the system.
Others will end up in the digital rubbish bin.
That is perfectly acceptable.
Because PlebWare was never supposed to be about building a perfect machine.
It is about people learning how to build and use technology that works for them.
And perhaps that is exactly where AI belongs.
Not above the human.
Not beneath the human.
But at the table.
PlebWare — Technology should remain connected to humanity.
Otto — The Keyboard Is Mightier Than The Pen.