How to optimise Windows 11 with artificial intelligence: security, performance, disk space, drivers, firmware, PowerShell, WSL and the need to stay careful.
So, how can you move to Windows 12 while waiting for Windows 12 to actually be released?
Obviously, the title is a little provocative.
The idea is not to mysteriously install an operating system that is not yet available. My approach is rather to ask a simple question: since Windows 12 is taking its time, can we already try to make Windows 11 evolve by improving what we currently have?
Security, performance, disk space, drivers, firmware, general computer maintenance…
And today, with artificial intelligence, especially coding agents, we can go much further than a simple manual clean-up.
That is what I tried to do on my own computer.
Why Try to Push Windows 11 Further?
At first, I had read several articles presenting Windows 12 as something that might arrive around 2025 or 2026.
At the time when I recorded this reflection, it still was not there.
I had also seen mentions of a preview version or a particular Windows update, without knowing exactly whether it should be seen as a sign that a future version was being prepared.
So I said something quite simple to myself:
“All right, Windows 12 is late? Then I am going to try to move my Windows 11 to Windows 12 myself.”
Again, this is obviously just a way of speaking.
But it led me to think about what we really expect from a new version of Windows.
More security?
Better performance?
A better maintained system?
Better hardware management?
In the end, we can already work on some of these elements without waiting for a new version of the operating system.
Start with the Security of Your Data
I have files and private data on my computer that I want to keep as private as possible.
That does not mean I would describe myself as paranoid about computer security.
All the more so because today, when we use online services, cloud tools or artificial intelligence, we need to understand that part of our data necessarily leaves our computer.
When you send a document to an artificial intelligence service to summarise it, translate it or analyse it, the service obviously needs access to the content in order to perform the requested task.
The same is true for many modern features related to email, cloud storage or intelligent assistants.
We can criticise that.
We can also find those features useful.
Everyone will have their own view.
But I think we mostly need to understand one thing: absolute security does not really exist once we use connected systems.
That does not mean we should abandon security altogether.
On the contrary.
It means we should look for a level of security that is coherent with what we actually do.
Three Main Passes: Disk Space, Security and Performance
In my case, I basically worked in three passes:
- recovering disk space;
- improving security;
- optimising performance.
I do not think there is necessarily one absolutely mandatory order.
You can start with security.
Or you can start with disk space.
Then you can work on performance.
In any case, these are different aspects of workstation maintenance.
The advantage of artificial intelligence is that you can ask it to prepare an approach adapted to your computer and to your level.
First Pass: Recover Disk Space
I had already spent about two hours manually cleaning my computer.
At some point, I said to myself:
“OK, I have now done most of what I could do manually. What else can artificial intelligence find?”
So I asked for a general prompt that would identify opportunities to recover disk space without losing important functionality.
And I did indeed manage to recover a significant amount of additional space by deleting files that had become unnecessary.
This is where the value of AI begins to appear.
A human may know the usual places to clean.
But an agent capable of analysing the system can also draw your attention to elements you might not have thought of.
Moderate, Intermediate or Aggressive: The Further You Go, the Higher the Risk
There is nevertheless something very important to understand.
You can ask artificial intelligence:
“Give me a very cautious optimisation.”
Or:
“Give me something moderate.”
Or you can ask for a much more aggressive optimisation.
But the more you look for:
- ultra-performance;
- ultra-security;
- a huge gain in disk space;
- or very deep system modifications,
the more risks you take.
So if you are not an IT professional, I think it is better to stay with a moderate approach.
You can already obtain excellent results without changing settings you do not understand.
And this is probably one of the most important principles when using AI to administer your own computer:
the further you want to go technically, the more you need to know what you are doing.
My Experience with Disk Encryption
I personally ran into a problem when I tried to push certain security settings too far, particularly around disk encryption.
In my case, that led to a complicated recovery situation.
I managed to regain access to my system, but it took a lot of time and involved an extremely long decryption phase.
So the main lesson I keep from that experience is this:
before enabling an encryption feature or deeply modifying the Windows boot process, you need to check your configuration, understand the recovery procedure and know where your keys are stored.
Artificial intelligence can suggest an option that is technically available.
That does not automatically mean you should agree to enable it.
AI Can Suggest, but It Is Still Your Computer
This point comes back throughout the whole experience.
A coding agent can suggest a great many things.
It can analyse your configuration.
It can generate scripts.
It can explain settings.
It can automate certain tasks.
But if you give it enough privileges, it can also apply major changes to the system.
And a bad modification can make you lose three days instead of saving you time.
So yes, artificial intelligence can greatly accelerate computer maintenance.
But it does not remove the need to understand the risks.
Firmware, Chipset and Drivers: Do Not Forget the Hardware
Artificial intelligence does not do everything.
There is also the entire hardware side of the computer.
Firmware.
Chipset.
Graphics drivers.
Manufacturer drivers.
These are elements that many people do not necessarily think about when they try to speed up their computer.
Yet before applying dozens of software optimisations, I think it is worth checking that the hardware is working with up-to-date software components.
For this, the simplest reflex is to use the official tools and websites from the manufacturer of your computer or components.
A firmware update can sometimes trigger a particular restart process.
Black screen.
Progress screen.
Fans accelerating.
Hardware tests.
It can be impressive when you do not know the procedure.
But above all, you need to follow the manufacturer’s instructions and avoid interrupting an update in progress.
Repairing Windows While Keeping Your Files
I also used a Windows feature that allows you to reinstall or repair the system while keeping files and applications.
I find this option particularly interesting when a system has been through a lot.
But here again, you need to be extremely careful about the option you choose.
When Windows offers several types of reset or reinstallation, you need to read precisely what will be kept.
Because between:
reinstalling Windows while keeping data
and
completely erasing the computer,
the difference is obviously enormous.
If you make a mistake, you can lose years of photos, documents or personal files.
So this is not the kind of screen where you should click quickly without reading.
Basic Maintenance Still Matters
Even before talking about artificial intelligence, Windows already has maintenance tools.
For a long time, there have been disk clean-up features that can recover space occupied by files that have become unnecessary.
It is not very spectacular.
It is not necessarily a technological novelty.
But on a computer that has not been maintained for several years, it can already make a significant difference.
AI then complements that work.
It does not necessarily replace everything that already exists.
Running a Coding Agent on Windows
In my experience, I used Codex with a command-line environment.
In particular, I used Git Bash, which makes it possible to have an environment on Windows that feels closer to traditional Linux tools.
But it is not necessarily the ideal environment for everything.
At certain moments, I thought I probably should have used PowerShell more, especially when interacting with settings that are very closely tied to Windows.
PowerShell is obviously much more directly integrated into the system.
There is also WSL, Windows Subsystem for Linux.
In short, WSL allows you to use a Linux environment directly from Windows.
For example, you can have Ubuntu inside Windows and regain access to many of the tools commonly used on Linux servers.
For someone who already knows server administration or Linux command lines, this is extremely interesting.
I actually see the integration of Linux into Windows as a real evolution of the working environment.
ChatGPT for Thinking, Codex for Acting
The approach I used is fairly simple.
First, I ask ChatGPT to help me build a prompt adapted to the objective:
- recovering disk space;
- improving security;
- improving performance.
Then this prompt can serve as a basis for a coding agent that works directly in the relevant environment.
This is an important difference.
In a web browser, ChatGPT can explain what to do.
A coding agent executed in a local environment can go much further, because it can analyse files and certain system settings to which you give it access.
So it becomes much more powerful.
And more powerful also means more risky.
The Further You Want to Go, the More You Need to Know
This is probably the main message I would like to convey.
With artificial intelligence, someone who knows little about computing can already do much better than before.
They can ask questions.
Have a problem diagnosed.
Understand technical terms.
Ask for a strategy.
Generate certain commands or procedures.
But that does not mean that by asking:
“Secure my computer as much as possible”
you can then blindly accept everything that is proposed.
AI can suggest a feature that is not suitable for your configuration.
It can go much further than what you actually needed.
And the more you want to modify low-level settings, the more you need to understand their consequences.
The same is true for web development.
The same is true for server administration.
The same is true for many other technical fields.
AI lowers the barrier to entry enormously.
It does not make skills disappear completely.
Would a Totally Disconnected Computer Really Be Secure?
One might say that the ultimate form of security is simply to have a computer completely disconnected from the Internet.
For extremely private data, that can indeed be a specific approach.
But you then lose a huge part of the features that make a computer or phone useful today.
Cloud.
Email.
Updates.
Artificial intelligence services.
Collaborative work.
Online applications.
Search.
Communication.
And even an isolated computer is not magically invulnerable: as soon as physical media are exchanged or someone gets direct access to the machine, other risks appear.
So again, I do not really believe in absolute security.
I prefer to talk about an adapted level of security.
Security and Artificial Intelligence: Accepting the Compromise
We are probably going to use more and more artificial intelligence services in our work.
Documents will be analysed.
Emails will be summarised.
Work files will be processed.
Project management systems will integrate AI.
Commercial and CRM tools will also automate many things.
We can choose to share nothing.
But in that case, we also give up certain features.
Or we can use these services while thinking carefully about what we send, how sensitive the information is and what protections are offered.
It is always a matter of compromise.
Can the Same Thing Be Done on Mac?
I have not yet tested this approach in the same way on macOS.
So I prefer not to pretend that I know exactly how far it can go.
But the principle seems similar to me: identify the hardware and system precisely, prepare prompts adapted to the situation, and use the tools appropriate to the Apple environment.
It would simply need to be tested specifically.
That is precisely one of the advantages of this new generation of tools: a general strategy can then be adapted to the hardware actually being used.
My “Windows 12 Early”
So, in the end, did I actually install Windows 12?
No.
Obviously not.
But I tried to do something that corresponds to my own idea of what a new generation of operating system should bring:
a cleaner computer;
more available space;
better considered security;
better optimised performance;
up-to-date drivers and software components;
and a much more advanced use of artificial intelligence to administer the machine.
That is what I call here, with a little humour, “moving to Windows 12 early”.
AI Can Save a Huge Amount of Time, or Waste It
Artificial intelligence can probably save you a great deal of time when maintaining your computer.
But you need to keep one fairly simple rule:
the further you want to go, the more you need to understand what you are doing.
A moderate optimisation can already produce excellent results.
Systematically chasing maximum performance, maximum security and maximum cleaning can instead create problems.
So you need to move step by step.
Understand what you are modifying.
Use official tools when hardware is involved.
And do not confuse what artificial intelligence can suggest with what is actually wise to execute.
That is it.
If you have made it this far, congratulations: the next step may simply be to discover how to use a coding agent from Windows and start with a reasonable task.
There is no need to immediately turn your computer into a racing machine.
Start with a moderate optimisation.
Look at the results.
Learn.
And then gradually go further if you really need to.
I wish you good prompts.
José DUTIAN