Think of it as Yoga for the brain, we are going to stretch a bit.
SO HOW DO I USE AI? — PART 1: MUSE
You can spend the next five years arguing about whether AI is going to replace people. Or you can spend the next five years learning how to make it work for you. I know which one I'm doing. And I have an assistant helping me. -- YNOT!
The AI Assistant That Changed How I Work, and Why I Think Everyone Should Have One
People keep asking me the same question.: “So, Tony, how do you actually use artificial intelligence?”
I suppose they imagine me sitting in a dark room, surrounded by blinking computers, talking to some electronic creature that answers in a British accent while plotting to take over the world.
Well, the computers are real. The rest is mostly Hollywood.
The truth is considerably less dramatic and a whole lot more useful.
I use AI every single day. Not occasionally. Not just to ask it what the weather is going to be or to write a birthday greeting.
I use it to work. And after more than four decades of working with computers, I can tell you that artificial intelligence is the most significant change in personal computing I have witnessed since the arrival of the personal computer itself.
Yes, I lived through DOS, Windows, the internet, smartphones, cloud computing, and all the other technological revolutions that were supposed to change civilization.
Some did. Others mostly changed the location of the Start button.
But AI is different.
For the first time, the computer is beginning to do something we have been trying to make it do for fifty years.
Understand what we want, figure out how to do it, and actually help us accomplish something.
MY AI TOOLBOX
First, understand that I don’t use just one AI. I use several, because they have different strengths.
ChatGPT is my primary tool for research, writing, analysis, and design. It is particularly useful when I want to explore ideas, develop arguments, examine information, or turn a collection of scattered thoughts into something worth reading. I find it has more insight than the others.
Claude is my preferred tool for programming. When I need to develop applications, troubleshoot complicated code, or work through software engineering problems, Claude often gets the assignment. It isn’t only that it is better but by dividing work it is easier to be organized.
Muse is my everyday AI assistant. It is the one I use to help organize, manage, and execute much of the work surrounding everything else.
And when I have something that needs to be done repeatedly, particularly on my own servers, I often move that job to my local AI systems.
Why? Because once you have the hardware, running your own AI models can eliminate the recurring cloud API charges for many routine tasks.
You still pay for electricity and equipment, of course. Even artificial intelligence hasn’t figured out how to negotiate with the electric company.
But you get the idea. I don’t believe in finding one AI and becoming religiously devoted to it. These are tools, not football teams. I use the one that works best for the job.
Today, we’re talking about Muse, particularly the newer Muse Wonder family of AI models.
And why I believe the AI assistant may turn out to be more important to ordinary people than the chatbot itself.
MEET MUSE — MY DIGITAL CHIEF OF STAFF
I currently pay about $15 a month for my Muse service.
And I can tell you that it saves me hours of work every day.
That’s my experience, not a promise that everyone will get identical results.
If somebody had told me twenty years ago that I could have an assistant available around the clock for less than the price of a decent lunch, I would have asked what kind of scam they were running.
Today, it’s simply another piece of technology.
But the important distinction is this:
Muse isn’t just something I ask questions.
It is something I give work to.
That may sound like a small difference.
It isn’t.
It is the difference between hiring a consultant who tells you how to paint your house and hiring somebody who actually picks up a paintbrush.
Most people have experienced a chatbot.
You type a question, it produces an answer, and everybody goes home.
An AI agent is designed to go further.
It can plan a task, use connected applications, look up information, carry out multiple steps, report its findings, and sometimes continue working while you’re doing something else.
Of course, what it can actually do depends on the tools and permissions you give it.
But instead of asking the computer, “How do I do this?”
I can increasingly say, “Go do this, check your work, and tell me what happened.”
Now we’re getting somewhere.
WHAT DOES MUSE ACTUALLY DO FOR ME?
Let’s start with something simple.
I manage a collection of websites, servers, applications, and business projects.
Anyone who has ever maintained even one website knows that it is a wonderful way to discover how many things can go wrong while you’re doing absolutely nothing.
Making sure something important doesn’t disappear into a folder where it will be discovered by archaeologists three generations from now.
Muse helps me manage much of that. I can give it instructions about how I want things organized and what needs attention.
It can prepare material, research subjects, and help coordinate the work.
Instead of spending hours moving information between systems, I spend more time deciding what should actually be done.
And there’s a significant difference between those activities.
One requires judgment. The other often requires nothing more than patience and the willingness to perform the same ridiculous task forty-seven times.
Patience is something computers have in extraordinary quantities.
I have considerably less.
SOCIAL MEDIA — ANOTHER FULL-TIME JOB NOBODY APPLIED FOR
Then there’s social media. Years ago, if you wanted to express an opinion, you stood on a street corner and shouted at strangers.
Today, we have Facebook.
The technology has improved considerably, although I’m not entirely convinced the conversation has.
But social media requires work. Articles need preparing. Posts need scheduling. Ideas need organizing.
You need to remember what you’ve already published, what you’re planning to publish, and what subjects deserve further attention.
Muse helps me handle that process.
I can develop an idea, have AI help organize the material, prepare a finished article, and then manage the steps needed to publish it.
I can also ask it to research a topic, identify issues worth discussing, or organize future articles.
The opinions remain mine. The decisions remain mine. The mistakes, unfortunately, also remain mine. But the typing, formatting, scheduling, and repetitive administration don’t all have to be.
And that’s the beauty of the arrangement. AI doesn’t have to agree with me.
In fact, I prefer that it challenges me when something doesn’t make sense.
I already know how to agree with myself. I don’t need to pay a computer to do that.
THE SECRET WEAPON — MEMORY
Now here’s something that sounds simple but is enormously important.
Memory.
Think about how much time you spend explaining things to people.
You hire somebody, and you have to teach them how your business works.
You explain the systems.
You explain the projects.
You explain what you’ve already tried.
You explain why something that looks perfectly reasonable is actually a terrible idea because you tried it three years ago and nearly destroyed the server.
And then, six months later, the person leaves.
Congratulations.
You get to start again.
A capable AI assistant with persistent memory can reduce that problem considerably.
Muse is designed to remember important context, ongoing objectives, previous conversations, and unfinished work.
That’s particularly valuable to me.
I might be working on one of my websites, a telephone system, a business application, or a new software project.
Tomorrow, I might return to that same project.
Instead of explaining the entire history from the beginning, the assistant may already have the relevant context.
It knows what we’ve been discussing.
It can help identify where we left off.
It can remind me what still needs attention.
That doesn’t mean AI memory is perfect.
It can miss things, confuse details, or remember something incorrectly.
Important facts still need checking.
But even imperfect memory can save an enormous amount of time.
A personal assistant becomes far more useful when it stops behaving like a stranger every morning.
FROM IDEA TO EXECUTION
Another important use is project development.
I have ideas.
Lots of ideas.
More ideas, unfortunately, than there are hours in the day.
That has always been one of my problems.
I can imagine a new application, website, business, or technical system faster than I can find the time to build it.
Historically, that meant a large collection of unfinished projects.
Now I can sit down with an AI assistant and explain an idea.
It can help me develop the concept, organize the requirements, research alternatives, identify technical problems, and prepare a development plan.
Then I can use other AI tools, such as Claude, to help write the actual software.
Muse can help keep track of the broader project.
What are we building?
What has been completed?
What still needs work?
What problems did we encounter?
What did we decide last week?
That is an enormous advantage.
Because anyone who has managed complicated projects knows that writing code is only part of the problem.
Keeping everything organized is often the bigger challenge.
And an AI assistant can help prevent the project from becoming a magnificent collection of unrelated files named FINAL, FINAL2, FINAL_NEW, and FINAL_REALLY_FINAL.
We’ve all been there.
A DAY WORKING WITH MUSE
Imagine a typical day. In the morning, I might ask Muse to check the status of my websites and identify anything that needs attention.
Then I might have it help prepare an article or investigate a technical question.
Around noon, I may be working on a software project and need information about how two systems communicate.
In the afternoon, I might ask it to organize some business material, prepare a proposal, or research an idea I’ve been considering.
Later, I might ask what remains unfinished.
The interesting part is not any one of those jobs.
It’s that they can all be part of the same working relationship.
The assistant can move between subjects, help coordinate the work, and maintain useful continuity.
A good assistant doesn’t need to know everything.
It needs to know what you’re trying to accomplish, where to find relevant information, and when to ask for help.
That is what makes the agent concept so interesting.
It changes the computer from a collection of disconnected programs into something that can help coordinate them.
SO HOW INTELLIGENT IS MUSE WONDER?
Now comes the question everybody asks.
Is it better than ChatGPT?
Is it better than Claude?
Which AI is the smartest?
People love rankings.
We rank cars, restaurants, football teams, and politicians.
Although judging by election results, I’m not entirely convinced we’ve perfected that last one.
Artificial intelligence has rankings, too.
And the results are interesting.
Meta introduced Muse Wonder and Muse Wonder Full in October 2026.
These are two different models, with Wonder Full being the more capable version.
The regular Muse personal assistant and the underlying AI models are also different products. Access to Wonder in the consumer assistant is being introduced separately from its availability to developers.
One independent ranking, the Artificial Analysis Intelligence Index, reported the following rounded scores around the October launch:
Claude Opus 5.5 — 58
Muse Wonder Full — 56
GPT-6 Astra — 53
Muse Wonder — 52
So Claude Opus leads the overall comparison.
But notice something.
Muse Wonder is not some little chatbot struggling to keep up with the adults.
It is competing in the same general class of advanced AI systems.
And when you look at particular types of work, the comparison becomes more interesting.
In Meta’s published DeepSWE software engineering results:
Muse Wonder Full — 79.0
Muse Wonder — 78.8
Claude Opus 5.5 — 74.2
GPT-6 Astra — 74.1
In its SWE-Marathon comparison, which measures sustained software engineering tasks:
Muse Wonder Full — 57.5
Muse Wonder — 53.1
Claude Opus 5.5 — 46.3
GPT-6 Astra — 42.5
That’s impressive.
However, on the published terminal coding comparison, Claude Opus and GPT-6 Astra score higher than regular Wonder.
And the broader picture requires a little caution.
The Artificial Analysis overall scores are independently evaluated. Several of the other comparisons are from Meta’s published model evaluations, and some competing models were tested using different tools or configurations.
That means the numbers are useful evidence, not a universal verdict.
The software that wins one benchmark may lose another.
And the smartest model in a laboratory isn’t automatically the best assistant for your business.
HOW MUSE RATES ITSELF
I thought it would be amusing to ask Muse Wonder to evaluate its own abilities.
After all, who better to judge a machine than the machine itself?
Naturally, there is a slight conflict of interest.
Asking artificial intelligence to rate its own intelligence is rather like asking a politician to investigate his own honesty.
You may receive a very impressive report.
Whether you should believe every word is another question.
Muse gave itself approximately 8.5 out of 10 for programming and 8 out of 10 for systems analysis.
It acknowledged that Claude may be stronger at certain programming tasks, particularly quick terminal operations and producing polished programs from scratch.
It also pointed to its strengths in longer engineering tasks, automation, and memory.
For example, Meta reported a MemoryBench score of 85.6 for regular Wonder, compared with 81.5 for Claude Opus 5.5 and 81.4 for GPT-6 Astra.
That particular memory benchmark is an internal Meta evaluation, not an independent ranking.
Still, memory is one of the reasons I find Muse useful.
And I appreciated something else about its answer.
It didn’t simply announce that it was the greatest invention since electricity and that all competing software should immediately surrender.
It acknowledged strengths and weaknesses.
Now, if we could only teach politicians to do that.
WHY THE BEST AI ISN’T ALWAYS THE SMARTEST AI
Here’s the point I want people to understand. Suppose I hire the world’s smartest engineer. He walks into my office. He knows mathematics, programming, electronics, economics, and probably the square root of my telephone number.
Wonderful. Now I ask him to fix one of my servers.
Before he can do anything, I have to explain which server, what software it runs, how the network is configured, what changes were made, what problems we encountered, and what we already tried.
His intelligence is enormously valuable. But he doesn’t know my systems.
Now imagine a slightly less brilliant engineer who has been working with me for years.
He knows the servers. He knows the applications. He remembers the previous problems.
He understands how I prefer things organized. He knows which systems are critical and which ones are experiments.
Who is more useful for today’s job?
The answer is not automatically the smartest engineer.
It’s often the one who understands the environment.
That’s what persistent AI assistants are beginning to provide.
Context. Continuity. Institutional memory.
And the ability to take action using connected tools.
Of course, ChatGPT, Claude, and other systems are also developing memory and agent capabilities.
This is a rapidly changing competition.
But that doesn’t change my point.
The best AI for your work may not be the one with the highest benchmark score.
It may be the one that knows your work best and can actually help you complete it.
THE PART THAT SHOULD MAKE EVERY SMALL BUSINESS OWNER PAY ATTENTION
Now think about this from the perspective of somebody running a small business.
Maybe you own a restaurant. Maybe you’re a contractor.
Maybe you’re a consultant, real estate agent, accountant, writer, or somebody trying to build a business from your kitchen table.
You have the same problem almost every small business owner has.
Too much work. Not enough time. And not enough money to hire somebody for every little task.
You need somebody to organize your calendar.
Somebody to research competitors. Somebody to prepare marketing materials.
Somebody to help write customer communications. Somebody to follow up on unfinished projects. Somebody to help organize your records.
Somebody to review what is happening and point out what deserves your attention.
Historically, you either hired people, did the work yourself, or simply accepted that some things would never get done.
Now there is another option. An AI assistant can take on parts of that workload.
Not perfectly. Not without supervision. And certainly not without occasional mistakes.
But sufficiently well that the time savings can be substantial.
Consider what an extra hour or two every day might mean to somebody running a business.
That’s time you can spend working with customers, developing products, solving problems, or simply enjoying your life.
And if you can save several hours, the economics become even more interesting.
What is your time worth? Twenty dollars an hour?
Fifty? A hundred? If a tool saves you two hours a day, it doesn’t need to be magical to justify its cost.
It just needs to be useful.
THE DANGER — DON’T CONFUSE AN ASSISTANT WITH AN AUTHORITY
Now before anybody decides to hand their entire business to an electronic assistant and retire to the Bahamas, we need to discuss something important.
AI makes mistakes.
Sometimes astonishing mistakes.
It can misunderstand instructions.
It can produce incorrect information.
It can overlook something important.
It can confidently tell you a job is finished when it isn’t.
And if you give it access to important systems, those mistakes can have consequences.
That’s why I treat AI as an assistant, not as an authority.
I expect it to explain what it did.
I want it to verify important results.
I want backups before significant changes.
And I don’t want it independently making consequential decisions about money, security, production systems, or sensitive information without appropriate approval.
I also believe in limiting access.
An AI assistant doesn’t need unrestricted control of everything you own.
Give it the tools required for the job and no more.
Trust is valuable.
Verification is better.
As an old computer professional, I’ve spent too many years fixing things that somebody assured me couldn’t possibly go wrong.
I don’t believe we’re anywhere close to understanding the full potential of personal AI assistants.
Today, we use them to write, research, organize, analyze, and perform tasks.
Tomorrow, more of them will monitor our systems, anticipate problems, coordinate projects, and take care of routine work before we even think to ask.
Imagine an assistant that watches your business systems, identifies unusual activity, investigates what happened, and alerts you only when something actually requires your attention.
Or one that tracks a project for months, keeps the documentation organized, follows up on delays, and suggests improvements.
Or one that helps a family organize appointments, insurance paperwork, finances, and all the other administrative nonsense that modern life has inflicted upon civilization.
Think of the difference between a computer that waits for instructions and one that helps you stay ahead of problems.
That is where this technology is heading.
Of course, we still have to solve questions involving privacy, reliability, security, and accountability.
And we need to be careful about allowing technology companies to become the permanent custodians of every detail of our lives.
But the potential is extraordinary.
MY ADVICE — START SMALL, BUT START
If you’ve never used an AI assistant, don’t begin by asking it to run your entire business.
Give it something simple. Ask it to organize a project.
Ask it to research something you’re planning to buy.
Have it prepare a letter. Ask it to help organize your schedule.
Let it summarize information that would otherwise take you an hour to read.
Then gradually give it more complicated work.
Learn what it does well. Learn where it makes mistakes.
And most importantly, learn how to explain what you want.
Because the great secret of working with AI is that you don’t have to become a computer programmer.
You have to become better at communicating. Explain the objective.
Provide the relevant background. Describe what a successful result should look like.
Tell it what it is not allowed to do. And ask it to check its work.
That process alone can make you better at managing almost anything.
THE REAL REVOLUTION ISN’T ARTIFICIAL INTELLIGENCE
It’s personal productivity.
For most of human history, the amount one person could accomplish was limited by time, physical ability, knowledge, and the number of people available to help.
Computers increased what we could do.
The internet increased what we could know.
Artificial intelligence is beginning to increase what we can delegate.
And that may be the biggest change of all.
I’m not using Muse because I want a machine to think for me.
I’m using it because I already have plenty to think about.
I don’t need another machine that tells me what to do.
I need one that helps me get things done.
For years, people have worried that computers would replace human beings.
I’m beginning to think the more immediate question is how much more a human being can accomplish when the computer finally becomes a competent assistant.
I still make the decisions. I still take responsibility. I still have to determine whether the work is any good.
But I no longer have to personally perform every repetitive step between having an idea and making something happen.
And that is a tremendous advantage.
So when somebody asks me whether an AI assistant is really worth paying for, I have a simple answer.
If it saves me hours of work every day, remembers the projects I’m working on, helps manage my systems, and allows me to accomplish things that would otherwise sit on the back burner for months…
Then yes. It is worth considerably more than the price of a monthly subscription.
After forty years of telling computers exactly what to do, line by line and command by command, I can finally explain what I want accomplished and have a machine help figure out how to accomplish it.
It isn’t perfect. It isn’t human. And it certainly isn’t a substitute for common sense.
But neither was the first automobile a substitute for knowing where you wanted to go.
It just made the trip a whole lot easier. And that, my friends, is why I use Muse.
Not because it’s the smartest machine in the room.
But because it helps me get more done while leaving me free to decide what is worth doing.
One final thought.
The future may not belong to the person who knows the most.
It may belong to the person who knows how to make the best use of what is available.
And right now, artificial intelligence is making more available to the individual than at any previous time in history.
Spent your time arguing about it or use it. If you use it you will be ahead.
Check out Muse,Redeem my code in Settings within 48 hours of joining and we'll both get 1 billion Muse tokens. Code: EYWN3S
https://muse.ai/join
Muse is a personal AI agent from Meta. Unlike a regular chatbot, it can use a browser, work with connected applications, remember your projects, and carry out multistep tasks.
Here are 15 practical ways to get started.
1. PERSONAL ASSISTANT
Organizes your daily responsibilities, priorities, and unfinished work.
TRY: “Help me organize my day. Ask me about my responsibilities, identify my priorities, and create a daily action plan.”
2. EMAIL MANAGEMENT
Reviews messages, drafts replies, summarizes conversations, and helps identify important emails.
TRY: “Review my recent emails, identify anything urgent, summarize what needs my attention, and prepare replies for my approval.”
3. CALENDAR AND SCHEDULING
Helps organize appointments, meetings, deadlines, and reminders.
TRY: “Review my calendar for next week. Identify conflicts, prepare a daily schedule, and remind me about important appointments.”
4. RESEARCH ASSISTANT
Searches the internet, compares information, and prepares research summaries.
TRY: “Research the five best solutions for this problem. Compare prices, advantages, disadvantages, and provide sources.”
5. WEBSITE MANAGEMENT
Helps review websites, identify errors, prepare updates, and organize maintenance.
TRY: “Examine my website for broken links, outdated information, and obvious technical problems. Prepare a report and recommend corrections.”
6. SOCIAL MEDIA ASSISTANT
Prepares articles, captions, posting schedules, and content ideas.
TRY: “Create a seven-day Facebook posting plan for my business. Include topics, suggested posts, and a publishing schedule.”
7. WRITING AND EDITING
Helps create letters, articles, proposals, business documents, and marketing material.
TRY: “Turn these rough notes into a professional article. Preserve my opinions and writing style, improve clarity, and verify factual claims.”
8. PROJECT MANAGEMENT
Organizes complex projects, tracks progress, and identifies unfinished tasks.
TRY: “Create a project called Website Redesign. Organize the objectives, tasks, milestones, deadlines, and outstanding decisions. Help me track progress.”
9. LONG-TERM MEMORY
Maintains useful context about your preferences, projects, and goals.
TRY: “Learn how I operate my business, what projects I’m working on, and how I prefer things organized. Use that context when helping me.”
10. AUTOMATED MONITORING
Can monitor information and alert you when something important changes.
TRY: “Check this website every morning for new announcements. Notify me only when something important changes.”
11. SHOPPING AND PRICE COMPARISONS
Researches products, compares sellers, and can assist with purchases.
TRY: “Find the best price for this product. Compare reputable sellers, shipping costs, warranties, and return policies. Don’t purchase anything without my approval.”
12. DOCUMENT CREATION
Produces reports, summaries, plans, and other documents.
TRY: “Create a business proposal for this idea. Include objectives, estimated costs, implementation steps, risks, and a professional presentation.”
13. BUSINESS ANALYSIS
Helps evaluate information, identify trends, and recommend improvements.
TRY: “Review these monthly sales figures. Identify trends, unusual changes, potential problems, and three opportunities to improve profitability.”
14. TRAVEL AND RESERVATIONS
Researches destinations, compares travel options, and helps make bookings.
TRY: “Plan a four-day vacation within my budget. Compare flights, hotels, activities, and transportation. Ask before making reservations.”
15. CUSTOM TOOLS AND MULTISTEP TASKS
Can use connected applications, navigate websites, and help create tools for repetitive work.
TRY: “I perform this task every week. Analyze the steps, determine what can be automated, and propose a reliable process that requires minimal supervision.”
HOW TO GET STARTED — FIVE MINUTES
STEP 1: Visit muse.ai and create an account. A free tier is available with usage limits.
STEP 2: Give your assistant a name. Mine is a cat named Muse!
STEP 3: Explain who you are, what you do, and what you want help accomplishing.
STEP 4: Connect the applications you actually need, such as email, calendar, or social media. Review permissions carefully.
STEP 5: Give it one real assignment. Don’t just ask a question. Ask it to accomplish something.
MY FIVE RULES FOR USING AI AGENTS
Explain the objective, not just the individual steps.
Give it enough background to understand the assignment.
Tell it what decisions require your approval.
Require it to verify important work and report problems.
Review the results before trusting anything consequential.
ONE FINAL TIP
Don’t ask Muse, “What can you do?”
Instead, ask: “Based on what you know about my work, responsibilities, and goals, identify ten repetitive tasks you could help me perform. Rank them by time saved, explain what access you would need, and show me how to get started.”
That one question may be worth more than a hundred demonstrations.
Remember, some capabilities require connected applications, additional permissions, or a paid plan. Muse Wonder is the underlying AI model family, while Muse is the personal assistant you interact with.
Check out Muse, your personal AI agent. Redeem my code in Settings within 48 hours of joining and we'll both get 1 billion Muse tokens. Code: EYWN3S
https://muse.ai/join
The objective isn’t to learn artificial intelligence.
It’s to teach artificial intelligence how to help YOU.
And the best way to learn is to give it something useful to do.
Coming next: How I Use ChatGPT for Research, Writing, Analysis, and Design.
