Think of it as Yoga for the brain, we are going to stretch a bit.
SO HOW DO I USE AI? — PART 2: CHATGPT – My Researcher, Fact-Checker, Writer, and Software Architect
I am telling you my secret magic - learn from it... Use it. --YNOT!
In Part 1, I explained how I use Muse as my personal AI assistant.
Muse helps me manage the daily chaos of websites, projects, schedules, and all those little things that require attention but seldom require much intelligence.
Today, we’re going to talk about ChatGPT. And this is where things get interesting.
I use ChatGPT primarily for four things: Research. Fact-checking. Writing. Software design.
Not because ChatGPT is necessarily the best at everything, but because it is exceptionally good at helping me investigate subjects, organize information, challenge assumptions, and turn ideas into something useful.
Think of Muse as my chief of staff, Claude as my software engineer, and ChatGPT as my research department, editor, analyst, and software architect.
And when I have something repetitive that can run on my own computers, I use local AI.
You see, I’ve discovered that artificial intelligence works much better when you stop looking for one machine that does everything and start treating the different systems as specialists.
After all, you wouldn’t hire a plumber to perform brain surgery simply because he owns a good set of tools.
Although, judging by some of the plumbing work I’ve seen, there are probably surgeons who would do a better job.
But I digress.
LET’S START WITH RESEARCH
I have been working with computers for more than forty years.
And during those years, I’ve watched the way we find information change dramatically.
When I was young, if you wanted to learn something, you went to a library.
You found a book. You read it. Then you found another book.
If the two books disagreed, you found a third.
It was an excellent system, provided you had the patience of a monk and access to a reasonably good library.
Then came the internet. Suddenly, the world’s information was available in seconds.
Unfortunately, so was the world’s misinformation.
Google made finding information easier, but it didn’t necessarily make understanding information easier.
Search for almost any controversial subject today, and you’ll find thousands of pages.
Some are excellent. Some are misleading. Some are written by people who understand the subject. Others appear to have been written by someone whose primary qualification is owning a keyboard.
Now add advertising, political agendas, search engine optimization, and websites that exist primarily to sell you something.
Finding information is easy. Finding reliable information is another matter entirely.
THIS IS WHERE CHATGPT BECOMES USEFUL
Instead of spending an hour searching through websites, I can begin by asking ChatGPT to research a subject.
But here’s the important part. I don’t simply ask it for an answer.
I ask it to investigate.
For example: “Research this subject using reliable sources. Explain the historical background, identify the important facts, show where experts disagree, and separate established information from speculation. Include sources so I can verify the conclusions.”
That is a completely different assignment from asking: “Tell me about this.”
And the difference in the results can be enormous.
ChatGPT can help organize a complicated subject into something understandable.
It can explain technical information. It can compare competing arguments.
It can analyze documents. It can identify questions that deserve further investigation.
And when using its web search or Deep Research capabilities, it can retrieve information from external sources rather than relying entirely on what the model learned during training.
That last point is particularly important.
Because AI models are not encyclopedias containing every fact ever recorded.
They are systems trained to recognize patterns, reason about information, and generate answers.
And sometimes they generate answers that sound magnificent.
Unfortunately, magnificent and correct are not always the same thing.
Which brings us to my second use.
FACT-CHECKING — OR HOW TO MAKE EVERYBODY ANGRY
I love researching history, economics, politics, technology, and current events.
And like most people, I frequently encounter articles, photographs, statistics, and social media posts that make extraordinary claims.
Some are true. Some are partially true. Some are technically accurate but deliberately misleading. And some are complete nonsense.
The internet has become a marvelous place where a person can invent a statistic before breakfast and have ten thousand people arguing about it before lunch.
Years ago, a misleading claim might take weeks to spread.
Today, it can circle the world while the truth is still trying to remember its password.
So I use ChatGPT to help investigate these claims.
I might upload a screenshot from Facebook and ask: “Fact-check this. Identify each factual claim, research the original sources, determine what is accurate, what is misleading, and what cannot be verified. Explain the findings without political bias.”
Notice that I don’t ask it to prove the article wrong.
I don’t ask it to prove the article right. I ask it to investigate. That’s an important distinction.
If you begin your research with the conclusion already decided, you aren’t researching.
You’re shopping for evidence. And there are enough people doing that already.
THE DIFFERENCE BETWEEN FACTS AND OPINIONS
Here’s something that seems to escape a remarkable number of people.
Facts and opinions are not the same thing.
You can have an opinion about whether a government policy is good or bad.
But whether the policy exists, how much it costs, what the law actually says, and what the available evidence shows are factual questions.
Those questions can be investigated.
For example, suppose somebody posts: “Congress just passed a law that will confiscate everyone’s retirement savings.”
That’s an extraordinary claim.
Before becoming outraged, I want to know: Did Congress actually pass a law?
Is this a proposed bill or an enacted law?
What does the legislation actually say?
Who introduced it?
When?
Does it affect everyone or only a particular group?
Is the claim based on an official document or somebody’s interpretation?
ChatGPT can help investigate those questions.
Sometimes the original claim turns out to be true. Sometimes it’s a misunderstanding.
And sometimes somebody has taken a proposal written twenty years ago, removed the context, and presented it as breaking news.
But here’s my warning. ChatGPT itself can make mistakes.
It can misunderstand a source, overlook an important detail, or occasionally produce an incorrect citation.
That means I still examine important evidence, particularly when the subject involves money, health, law, or significant public controversy.
Artificial intelligence is a very useful research assistant. It is not a substitute for evidence.
And no matter how confidently an AI speaks, confidence is not proof.
THE FEATURE I LIKE MOST — DEEP RESEARCH
ChatGPT has a feature called Deep Research.
This goes considerably beyond asking a simple question.
You give it a complicated subject.
It develops a research approach, examines multiple sources, compares information, and produces a structured report with citations.
Think of it as assigning a research project rather than asking a question.
For example: “Prepare a comprehensive report on how artificial intelligence is changing employment. Include historical comparisons with previous technological revolutions, current economic research, industries most affected, competing forecasts, and potential consequences over the next ten years.”
Instead of receiving a short explanation, I can receive a much more extensive analysis that I can examine, challenge, and use as the foundation for further work.
I can also upload documents and ask it to compare them.
That is particularly useful when examining technical reports, legislation, contracts, academic papers, or business information.
Naturally, the final report is only as reliable as its evidence and analysis.
A research report with twenty citations isn’t necessarily better than one with five.
The important question is whether those citations actually support the conclusions.
But for somebody who enjoys investigating complicated subjects, this capability is remarkable.
It reduces the mechanical work required to gather and organize information.
Which gives me more time to think about what the information actually means.
AND NOW WE COME TO WRITING
People sometimes ask whether I use AI to write my articles.
Of course I do. Why wouldn’t I?
I also use a keyboard instead of a typewriter.
And I use a word processor instead of a bottle of ink and a feather.
I see no particular virtue in making a job more difficult simply because our ancestors had to.
But there’s something people misunderstand about AI-assisted writing.
I don’t want a machine to decide what I believe.
I want it to help me express what I believe.
Those are two entirely different things. I usually begin with an idea.
Many times, I speak on my phone research this topic in-depth.
Sometimes it’s a short paragraph. Sometimes it’s a collection of notes.
Sometimes it’s a long, disorganized explanation containing enough spelling mistakes to frighten an English teacher.
I give it to ChatGPT. And I explain what I want.
For these articles, I frequently request what I call an Expanded commentary.
That means I want something conversational, analytical, entertaining, occasionally sarcastic, and grounded in real information.
I want the reader to think.
I want complicated subjects explained in ordinary language.
And if we can make the reader laugh occasionally, so much the better.
Humor is a wonderful teaching tool.
It allows you to explain uncomfortable truths without necessarily making the entire conversation uncomfortable.
Although some people are determined to be offended, regardless of how carefully you explain something. I suspect those people would complain about the weather even if they controlled it.
HOW I ACTUALLY WRITE AN ARTICLE
My process is fairly straightforward.
First, I develop the idea.
I might write several paragraphs explaining what I’m thinking.
Then I ask ChatGPT to organize and expand the material.
I frequently ask it to research the background, identify supporting evidence, and challenge weak arguments.
Then we work on the structure.
Does the introduction capture the reader’s attention?
Does the article explain the subject clearly?
Are the arguments supported?
Are the examples understandable?
Is anything unnecessarily repetitive?
Have we confused opinion with fact?
Then we work on the tone. Sometimes I want serious analysis. Sometimes I want humor.
Sometimes I want something that sounds like an old man sitting on a porch explaining why the world has lost its collective mind.
That last style comes remarkably naturally to me.
ChatGPT helps me turn the material into a finished article. It is a secretary that goes faster than me and makes no spelling mistakes.
Then I review it. I change things. I correct things. I remove things.
Everything is an opportunity. Everything is transformative. Everything is a journey.
Sometimes a broken computer is simply a broken computer.
It doesn’t need to become a journey of personal discovery.
A GOOD EDITOR ISN’T SOMEBODY WHO ALWAYS AGREES WITH YOU
One of the things I appreciate about ChatGPT is its ability to challenge an argument.
I can ask it: “Find the weaknesses in what I just wrote.”
Or: “Present the strongest reasonable argument against my position.”
Or: “Identify which claims are facts and which are my interpretations.”
These are particularly valuable requests.
Because we all have biases. Including me. Including you.
And including the people who design artificial intelligence.
Nobody approaches every question with a perfectly empty mind.
But good research should help us recognize those biases rather than simply reinforce them.
I don’t want ChatGPT to tell me I’m brilliant. I already have a cat who treats me like an idiot whenever his food bowl is half-empty. That’s enough emotional support for one household.
What I want is useful criticism. If an argument is weak, tell me why.
If the evidence is incomplete, identify what’s missing. If I misunderstand something, explain it.
That’s considerably more valuable than having a machine enthusiastically agree with everything I say.
SOFTWARE DESIGN — WHERE THINGS GET REALLY INTERESTING
Now we come to my fourth major use of ChatGPT. Software design.
And this is where my background in computers becomes particularly useful.
I’ve been designing, installing, maintaining, and troubleshooting computer systems for too many decades.
Servers, databases, networks, operating systems, websites, business applications.
I’ve spent a considerable portion of my life making computers do things their manufacturers apparently never intended.
And one lesson I’ve learned is that writing software and designing software are not necessarily the same skill.
Writing code is how you tell the computer what to do.
Designing software is deciding what the computer should do, how it should work, and how all the pieces should fit together.
A good programmer can write a thousand lines of code.
A good software architect may figure out how to accomplish the same thing with a hundred.
Or determine that the feature shouldn’t exist in the first place. That’s where I use ChatGPT.
FROM AN IDEA TO A WORKING APPLICATION
Suppose I have an idea for a new application.
Maybe it’s a business management system. Maybe it’s a website. Maybe it’s a simulation game. Maybe it’s something that monitors computer servers and identifies technical problems.
I begin by describing the concept to ChatGPT.
What is the application supposed to accomplish?
Who will use it?
What information does it need?
What are the major features?
What technology should it use?
How should the information be stored?
What happens when something fails?
How do we keep it secure?
How do we expand it later?
And, perhaps most importantly, what should we avoid building until the basic system actually works?
ChatGPT can help turn those ideas into a software specification.
For example, I frequently work with Linux, Python, Nginx, and MariaDB.
So I can ask ChatGPT to design a system using that technology. It kind of already knows this our platform.
It can help organize the database structure, application modules, API requirements, security considerations, and deployment process.
It can also explain different technical approaches and their advantages.
Sometimes I already know what I want. Other times, I’m evaluating several possibilities.
I then ask it a question… Do you have any question from me? It usually fires up 5 or 10 questions to help it season the idea correctly.
At the end I tell it to give an summary of idea and project goals. And it generates a complete system analysis of idea for my review. 99% of the time it is perfectly fine. Sometime I tweak some names or menus, and it updates everything.
That is where AI-assisted design becomes exceptionally useful.
You can explore an idea without spending days building something only to discover that the underlying architecture was wrong.
MY WORKFLOW — CHATGPT, CLAUDE, AND MUSE
Here’s how the different tools work together.
I used ChatGPT to design the application. We discuss the concept.
Develop the requirements. Define the architecture. Identify the risks.
Create the implementation plan. Then I may take that specification to Claude.
I can then have ChatGpt check Claudes work, review it, make me prompt for changes.
Claude Code is particularly useful for implementing and modifying software.
It can work with project files, inspect code, write new modules, and help troubleshoot problems.
Meanwhile, Muse can help organize the larger project, keep track of outstanding issues, and assist with the administrative work. Get Social Media and documentation ready.
And once the application is operating, some recurring functions can move to my local AI systems.
That gives me several specialized tools working on different parts of the same problem.
One helps determine what to build. Another helps build it. Another helps organize the work. And another can help operate it.
Of course, the boundaries aren’t absolute.
ChatGPT has Codex, which can write, inspect, and modify code. Claude can help design architectures. Muse has agent and developer tools of its own.
These companies are competing aggressively, and their capabilities overlap more every month.
But my workflow isn’t determined by which company has the most impressive advertisement.
It’s determined by which tool gives me the best results.
A MORE CAPABLE CHATGPT IS EMERGING
This is worth understanding because ChatGPT itself has changed considerably.
It is no longer simply a conversational interface.
It includes research capabilities, document analysis, image generation, memory, coding assistance, connected applications, and increasingly sophisticated agents.
ChatGPT Work can carry out longer, multistep assignments. Codex is specifically designed for software development. And OpenAI has introduced Dot, an always-on personal agent available through eligible plans. In other words, ChatGPT is moving into territory traditionally associated with products like Muse.
At the same time, Claude is expanding beyond programming.
And Muse is expanding its underlying models and developer tools.
The boundaries are disappearing. Which is excellent news for customers.
Competition has a wonderful habit of improving technology.
Especially when the companies involved are spending billions of dollars trying to convince us that their particular electronic brain is superior.
CHATGPT’S DIFFERENT PLANS — WHAT DO YOU ACTUALLY GET?
Now we need to discuss cost.
Because the pricing structure for artificial intelligence has become almost as complicated as the technology itself.
And if you thought choosing a cellphone plan was confusing, welcome to artificial intelligence.
As of November 2026, OpenAI offers several ChatGPT plans.
CHATGPT FREE — $0 This is where most people should begin.
You can ask questions, write material, conduct basic research, and experiment with many AI capabilities. There are limits on features such as uploads, image generation, research, and advanced tools. But for somebody who has never used AI, the free version is an excellent introduction.
You don’t need to spend a penny to discover whether this technology is useful. CHATGPT GO — ABOUT $8 PER MONTH IN THE UNITED STATES. Go is designed for people who want more access than the free service provides without paying for the more advanced plans. It offers expanded access to several everyday features, including uploads, image creation, and longer memory. It’s a reasonable option for somebody who uses AI regularly but doesn’t need the most sophisticated reasoning tools.
CHATGPT PLUS — $20 PER MONTH
This is the plan I would recommend most people seriously consider if they find themselves using ChatGPT regularly. Plus provides access to more capable reasoning options, expanded research, larger allowances, image generation, projects, and additional productivity tools. It also includes access to Codex and ChatGPT Work, subject to usage limits. For research, writing, document analysis, and general business productivity, this is a very capable subscription. And twenty dollars a month is not much money if the service saves you even a few hours.
CHATGPT PRO — $100, $200, OR $500 PER MONTH Now we’re entering professional territory. OpenAI currently offers three Pro levels.
Pro 100 costs $100 per month.
Pro 200 costs $200 per month.
Pro 500 costs $500 per month.
These plans provide higher usage allowances and access to advanced capabilities, including GPT-6 Pro reasoning powered by Astra.
The more expensive tiers are primarily about how much demanding work you can perform, not simply making every casual question better.
The $500 plan also includes an Ultrafast option for supported advanced models.
These subscriptions become interesting when AI is part of your daily professional workload If you’re conducting extensive research, developing software, or running long agent-based assignments, additional capacity can be valuable.
But if all you’re doing is asking for recipes and writing birthday greetings, spending $500 a month is probably unnecessary.
Unless you’re planning an extraordinarily complicated birthday party.
CHATGPT BUSINESS AND ENTERPRISE
OpenAI also offers products for organizations. Business standard seats are generally $25 per user monthly, or $20 per user monthly with annual billing.
Higher-capacity premium seats are available. Enterprise provides more extensive organizational controls, administration, security options, and negotiated arrangements.
These services are intended for businesses that need shared workspaces, managed access, and more formal data protections. One important distinction is that OpenAI does not use Business and Enterprise workspace data to train its models by default.
For organizations handling confidential information, that matters.
SO HOW DOES CHATGPT COMPARE WITH CLAUDE?
Claude is developed by Anthropic. And I use it extensively for programming.
In fact, for many of my coding projects, I prefer Claude. Not because ChatGPT can’t program. It certainly can.
But different tools have different strengths, and I’ve found Claude particularly useful when working through complicated application development.
Claude offers its own subscription plans. Claude Free costs nothing. Claude Pro costs $20 monthly, or approximately $17 per month when purchased annually.
Claude Max costs $100 monthly.
Claude Ultra costs $200 monthly.
There are also Team and Enterprise products.
Claude Code is available with eligible paid subscriptions and is one of its major attractions for developers.
Claude also offers research, document creation, projects, memory, connectors, and other productivity capabilities.
In many respects, Claude and ChatGPT are direct competitors. But there are differences in emphasis.
For my purposes: ChatGPT is my first choice for research, fact-checking, article development, analysis, and software architecture.
Claude is often my first choice for implementing larger programming projects.
That doesn’t mean ChatGPT is incapable of developing software or Claude is incapable of research. It means I have developed a workflow based on experience.
And that workflow can change.
If ChatGPT’s Codex becomes more effective for one of my projects, I’ll use it.
If Claude develops a better research capability for a particular assignment, I’ll use that.
I’m not buying stock in these companies. I’m buying productivity.
AND WHAT ABOUT MUSE?
In Part 1, I explained why Muse has become my everyday AI assistant.
Meta designed Muse around the concept of a persistent personal agent.
One that can work across connected applications, navigate websites, perform tasks, and continue certain assignments while you are doing something else.
Muse operates using its own dedicated cloud computer.
Meta also designed a separate security component called Sentinel to supervise certain activities and permissions.
That’s an interesting approach.
It makes Muse particularly attractive for people who want an assistant that helps carry out real-world tasks rather than simply producing information.
Muse has a free tier, along with paid subscription options.
Its publicly reported personal-agent plans include Power at approximately $20 per month and Maximum at approximately $100, although availability, pricing, and individual subscription arrangements can vary.
Meta also offers Muse Code, which is a separate developer-oriented product.
Muse Wonder and Wonder Full are underlying AI models, not simply alternative names for the personal assistant subscription.
That’s important because people sometimes compare model benchmark scores as if they directly represent the complete products.
They don’t. The AI model is the engine. The application is the automobile.
And a powerful engine doesn’t automatically make an automobile comfortable, practical, or reliable.
CHATGPT VERSUS CLAUDE VERSUS MUSE
Here’s how I look at the competition.
For research and fact-checking, I generally prefer ChatGPT.
Its combination of web search, Deep Research, document analysis, and reasoning tools makes it particularly useful for investigating complicated subjects.
Claude is also a capable research tool, and Muse can investigate information as part of its assignments.
For writing, I generally prefer ChatGPT.
I can develop an idea, challenge an argument, revise the tone, and produce an article or document. It is my spell checker, my grammar ninja and my editor.
Claude is a strong writer as well, and some people prefer its writing style.
That is largely a matter of taste and workflow.
For programming, I frequently prefer Claude.
Claude Code is particularly useful for working directly with software projects.
But ChatGPT’s Codex is a serious competitor, and Muse Code is another option worth watching.
For everyday personal assistance, I prefer Muse.
Its emphasis on persistent tasks, connected applications, and a dedicated agent environment fits the way I work.
However, ChatGPT Work, Dot, and Claude’s expanding agent capabilities are making this competition much more interesting.
And for visual design and image creation, I frequently use ChatGPT.
I can describe a concept, create an illustration, revise it, and produce graphics for websites, articles, or other projects.
It’s one of my favorite uses.
The interesting thing is that none of these categories has a permanent winner.
The industry is changing too quickly.
Today’s leader in one capability may be tomorrow’s second choice.
And that is precisely why I don’t believe in becoming dependent on a single AI provider.
WHAT ABOUT SECURITY AND PRIVACY?
This is another subject people should understand.
When you connect an AI assistant to your email, files, calendar, or business systems, you are giving software access to information that may be important or confidential.
That requires some judgment.
ChatGPT provides privacy controls that let personal-account users disable the use of new conversations for model training.
Business and Enterprise accounts have additional protections, with business data excluded from model training by default.
Muse emphasizes its dedicated Secure VM, permission controls, activity records, and Sentinel security architecture.
Claude also provides privacy controls and separate protections for business customers.
But none of this means you should blindly trust an AI agent with unrestricted access to everything you own.
I wouldn’t give a newly hired employee the administrator password to every server in my company on the first day.
Why would I give it to a machine?
Give these tools the access they need.
Review sensitive actions. Require approval before consequential changes.
Maintain backups. And verify the results.
The fact that an AI can accomplish something doesn’t necessarily mean it should be allowed to do it without supervision.
A powerful assistant is valuable.
A powerful assistant with unrestricted authority can become a very expensive problem.
SO WHICH PLAN SHOULD YOU BUY?
Here’s my advice. If you’ve never used AI, start with the free services.
Try ChatGPT. Try Claude. Try Muse.
Give them the same assignment and compare the results.
If your primary interest is research, writing, or ordinary professional work, ChatGPT Plus at $20 monthly is a logical starting point.
If your main interest is programming, Claude Pro at $20 monthly deserves serious consideration.
If you want a personal AI agent that helps carry out tasks across applications, experiment with Muse.
And if AI becomes a major part of your professional workload, then investigate the more expensive plans.
But don’t confuse the price of a subscription with the intelligence of the person using it.
A person who understands how to use a $20 AI subscription may accomplish far more than someone paying $200 who doesn’t know what to ask.
These are tools.
And tools become valuable when you know how to use them.
THE SECRET TO GETTING BETTER ANSWERS
Now here’s something I’ve learned from spending so much time working with artificial intelligence.
The quality of your results often depends on the quality of your instructions.
If you ask ChatGPT: “Tell me about economics.”
You’re going to receive a general explanation. But suppose you ask:
“Explain the major differences between a market economy and a centrally planned economy. Include historical examples, advantages, weaknesses, the role of incentives, and the effects on production and consumer choice. Separate established economic evidence from political arguments.”
Now you’ve given it a real assignment. The same principle applies to writing.
Don’t simply say: “Write an article.”
Explain the audience. Explain the objective. Explain the tone.
Explain what information must be included.
Explain what should be avoided. And when designing software, be especially precise.
Describe what the application must do. What technology it should use.
What data it needs. What the user should see. And what a successful result looks like.
I’ve discovered that working with AI forces you to become more disciplined about explaining your own ideas.
And that’s not a bad thing. In fact, it’s one of the unexpected benefits.
IF AI IS SO SMART, WHY DO I STILL NEED TO THINK?
Because artificial intelligence doesn’t eliminate the need for judgment.
It increases the importance of judgment. Think about it.
If a computer can produce ten articles in the time it previously took you to write one, you still need to determine which articles are worth publishing.
If it can design five different applications, you still need to determine which one solves the actual problem.
If it can research twenty different investments, you still have to decide what risks you’re willing to accept.
And if it can generate a thousand arguments supporting your opinion, you still need to determine whether your opinion makes any sense.
AI can increase the quantity of work produced. It doesn’t automatically increase its quality.
That requires someone who understands the objective. And that someone remains you.
THE GREAT AI COMPETITION
I find the competition between OpenAI, Anthropic, Meta, Google, and the other major AI developers fascinating.
They are spending enormous amounts of money developing increasingly powerful systems.
Every few months, a new model appears.
Every company announces another breakthrough.
Benchmarks improve. Prices change. New capabilities emerge.
And yesterday’s technological miracle becomes today’s standard feature.
It’s a wonderful situation for customers, provided the competition continues.
Because the real prize isn’t simply building the smartest chatbot.
It’s building the AI system that people and businesses actually depend on.
That means intelligence. But it also means reliability.
Memory. Security. Integration. Ease of use. And the ability to accomplish real work.
The winner won’t necessarily be the company with the model that scores two points higher on some obscure benchmark.
It may be the company that makes its technology the most useful.
And I suspect the competition will become increasingly fierce.
A FINAL THOUGHT — THE COMPUTER IS FINALLY LEARNING TO WORK WITH US
For most of my career, computers required people to learn their language.
We learned commands. We learned programming languages.
We learned operating systems. We learned databases.
We learned complicated software.
And then, just when we finally understood how everything worked, somebody released a new version and moved all the buttons.
Now something remarkable is happening.
Computers are beginning to understand our language.
I can explain what I want. I can ask questions.
I can challenge the results. I can provide additional information.
And the system can help me investigate, design, organize, and accomplish things that previously required considerable specialized effort.
Does that mean everybody will become a programmer? No.
But it means somebody with a good idea may no longer need an entire development department just to explore whether that idea is practical.
Does it mean everybody will become a professional writer? No.
But it means somebody with something worthwhile to say may have a much easier time expressing it.
Does it mean everybody will become a research scientist? Certainly not.
But it means ordinary people will have access to research assistance that would have seemed extraordinary only a few years ago.
And that is a tremendous change.
MY BOTTOM LINE
I use ChatGPT because it helps me think.
Not because I want it to think instead of me.
I use it to investigate information.
To challenge claims. To understand complicated subjects.
To write articles. To design software. To explore ideas.
And to organize the enormous amount of information required to turn those ideas into something useful.
I use Claude because it helps me turn software designs into working programs.
I use Muse because it helps me manage the work surrounding everything else.
And I use local AI because some things should simply run on my own computers without generating another monthly bill.
Each tool has its place. Each has advantages. Each has weaknesses.
And all of them are improving at an extraordinary pace.
So when somebody asks me which AI they should use, I don’t begin by asking which company has the highest benchmark score.
I ask a much simpler question. “What are you trying to accomplish?”
Because that determines which tool is useful. And usefulness is what matters.
I’ve spent more than forty years making computers do things.
Now, for the first time, I can spend more of that time deciding what I want accomplished instead of figuring out every little step required to make the machine cooperate.
That’s progress.
And after forty years of dealing with computers, I’m delighted to discover that they’re finally learning to understand instructions without requiring me to become an expert in whatever language their programmers invented last Tuesday.
Of course, they still make mistakes. They still misunderstand things.
And they occasionally produce answers so confidently wrong that I wonder whether somebody trained them on political speeches.
But they’re getting better. And they’re getting better quickly.
The important thing isn’t whether artificial intelligence will change your life.
It already has. The important question is whether you’re going to learn how to use it.
Because the future doesn’t necessarily belong to the people who know the most.
It belongs to the people who can learn, adapt, and make productive use of the tools available to them.
And right now, we’ve been handed some extraordinarily powerful tools.
My advice? Stop arguing about which AI is the smartest. Learn how to use them. Give them real work. And then check the results.
And spend the time you save doing something that actually matters.
After all, the objective was never to make the computer more important than the human being.
It was to make the human being more capable.
And that, my friends, is why I use ChatGPT.
SO HOW DO I USE AI?
Part 1 — Muse: My Personal AI Assistant
Part 2 — ChatGPT: Research, Fact-Checking, Writing, and Software Design
Coming Next — Part 3: Claude — How I Use Artificial Intelligence to Build Real Software
