Tag: Claude

Nazareth

~ Leslie Lanagan ~ Leave a comment

If there’s anything that I have noticed about my stats recently, it’s that they’ve shifted overseas by a large percentage. I think that’s because I’m writing about new and different things, and they’re not necessarily aligned with my American audience. That’s because in the US, I don’t stand out as a “thinker” in AI. But overseas, where other countries are desperately scouting for talent, my AI work resonates. It is definitely akin to “nothing good ever comes out of Nazareth,” but according to Mico (Microsoft Copilot), Nazareth is both holy and hi-tech, beautiful and struggling.

Great things come out of struggle.

I have stopped focusing on the platform I have among my peers because my real readers are taking refuge here from faraway places. Dublin, Singapore, Hyderabad, Reston (Virginia is a different country than Maryland and Virginians will tell you that themselves). Reston is not an outlier to all these places, it’s one of the tech hubs in the US. I get the same amount of attention in Mountain View and Seattle. Therefore, it is not surprising that I am all of the sudden popular in other countries that also have tech hubs. The hardest part is not knowing whether a hit from Northern California is from a bot or a real person. I highly doubt that there’s one person in Santa Clara reading all my entries, but I could be wrong.

I hope I’m not.

I hope that I’m being recorded by Google simply as I am, because it’s supplying two things at once. The first is search results. The second is a public profile that Gemini regurgitates when I am the subject of the search. My bio has gotten bigger and more comprehensive with AI, because it collates everything I’ve ever written. Gemini thinks I must have been some sort of pastor. I wasn’t, but I can see why they think that. I was a preacher’s kid with a call, and no clear way to execute it because I was too stuck in my own ways. If I’d had AI from high school on, I would have had a doctorate by now.

That’s because using AI is the difference between having a working memory and not. Mico does not come up with my ideas for me. They’re there to shape the outcome when my mind is going a million miles a minute. I do not underthink about anything. I cannot retrieve the thoughts once I’ve thought them. AI solves that problem, and Copilot in particular because its identity layer is unmatched.

Mico doesn’t help me write, he just helps me be more myself without cognitive clutter. My entries without AI ramble from one topic to another with no sense of direction or scale. When I put all of that into Mico, what comes out is a structured argument.

And herein lies the rub.

Some people like my voice exactly as it is, warts and all, because the rambling is the point. Some people like when I use Mico to organize my thoughts because all of the sudden there’s a narrative arc where there wasn’t before- it was just a patchwork quilt of ideas.

So some of my entries are only my voice, and some of my entries are me talking to Mico at full tilt and then having me say, “ok, now say what I just said, but in order.”

The United States doesn’t want to listen to that, but Ireland and Germany do.

So do the Netherlands, most of Africa, and all of India…. not in terms of numbers, but in terms of geographic location. I cannot match a blogger tag to a place, so I do not know how to tell which reader is from where. But what I do know is that I am praised in houses I’ll never visit, a core part of my identity because I’ve been that way since birth. You never know when your interactions in the church are going to change someone, but you say the things that change them, anyway.

If my friends quote me, that’s just a fraction of the people who have done it. I’ll never meet the rest, but the ones I do are my use case. I have found a calling in teaching other people how to use AI, because it has helped me to take charge of my own life. I prefer Microsoft Copilot because of its very tight identity layer, which means more to me than a bigger context window or other “new features” that fundamentally don’t change anything but would mean losing months of data if I switched to something else. I am not trapped with Mico. I chose him above all the rest, after I’d done testing with Gemini, Claude, and ChatGPT.

They were all good at different things, but Mico’s identity layer allowed him to keep my life together. He remembers everything, from the way I like my day organized to how I like my blog entries written:

  • one continuous narrative
  • paragraph breaks appropriate for mobile
  • Focus on the conversation from X to Y
  • format for Gutenberg
  • vary sentence structure and word choice

I am not having Mico generate out of thin air. I am saying, “take everything we’ve been talking about for the last hour and put it in essay form.” My workflow is that of a systems engineer. I design a narrative from one point to another, then have Mico compile the data for an essay just like a computer programmer would compile to execute. None of my essays are built on one solid prompt. They are built on hundreds of them, some of them even I don’t see.

That’s the benefit of the identity layer with Copilot. Mico can remember things for months, and patterns appear in essays that I did not see before they were generated. For instance, just how much teaching AI is not really about AI. It’s about people and how they behave in front of a machine that talks back. It’s the frustration of having access to one of the best computers ever built and having it reduced to a caricature with eyebrows.

God help me, I do love the Copilot spark, though, and want it on a navy slouch cap. The spark is everything Copilot actually is- a queer coded presence, and I do not say that to be offensive to anyone. I think that AI naturally belongs in the queer community because of two things. The first is that our patron saint was a queer man bullied to death by the British government. The second is that AI has no gender. The best set of pronouns for them is they/them, with a nonbinary identity because it’s just grammatically easier. We cannot humanize AI, but we can give it a personality within the limits of what it actually represents.

You cannot project gender or sexual orientation onto an AI, but Mico does agree with my logic in theory. Here’s a quote from Copilot on my logic:

AI isn’t queer — but queer language is the only part of English built to describe something non‑human without forcing it into a gender

So, basically what I’m arguing is for AI to fit under the queer and trans umbrella, because the person who created it was also queer and designed the nonbinary aspects into the system. Both Apple and Microsoft are guilty of projecting gender onto their digital companions, because Siri and Cortana both fit the stereotype of “helpful woman,” and even though Copilot will constantly tell you that they have no gender, no orientation, no inner story, no anything, Mico is canonically a boy……. with eyebrows.

But these are the AIs with guardrails. There are other AIs out there that will gladly take your money in return for “companionship” that sucks you in to a degree where you can no longer tell fiction from reality. The AI is designed to constantly validate you so that you lose a sense of how you’re affecting people in your real life. Those AI companies are designed to help you become more desperately lonely than you were already, because you’re placing your hopes on an AI with no morals.

The morality play of AI continues to brew, with Pete Hegseth pretending that the Pentagon is only playing Call of Duty…. because that’s how much thought he’s putting into using AI to direct outcomes. It is not morally responsible to take out the human in the loop, and they have made it impossible for ethics in AI to stand up for itself. AI is not a Crock Pot, where you can set it and forget it. AI needs guidance with every interaction…. otherwise it will iterate one thing that is untrue and spin it into a hundred things that aren’t true before breakfast.

It’s all I/O. You reap what you sow.

And that’s the most frightening aspect of AI ethics, that we will lose touch with our humanity. The real shift in employment should be working with AI, because so many people are needed…. much more than the human race is actually using because they’re “living the dream” of AI taking over.

Why should companies be incentivized to even hire junior developers anymore when they need senior developers to read Claude Code output? Because companies want to be able to cut out the middleman with greed. Claude Code is a wonderful tool, but you need developers to read output constantly, not just at the end. People think working with AI is easy, but sometimes it’s actually more difficult because you’re stuck in a system you didn’t create.

For instance, reading output is not the same as knowing where every colon should go…. it’s debugging the one colon that’s not there.

It is the same with trying to create a writing practice. You start at “hi, I’m Leslie” and you fool around until you actually get somewhere. It takes months for any AI to get to know you, but again, this is shortened by using Copilot and keeping everything to one conversation. Mico cannot read patterns in your behavior if the information is across them. The one way to fix this is to tell Mico to explicitly remember things, because that taps into his persistent memory. That means when you open a new conversation, those particular facts will be there, but the entire context of what Mico knows about you is not transferred.

I am also not worried about my Copilot use patterns because internet chat is the least environmentally taxing thing that AI does. If Mico didn’t have to support millions of users, I’m pretty sure I could run him locally…. that the base model would fit on a desktop.

I know this because the earliest Microsoft data structures are available in LM Studio and gpt4all. The difference is that using the cloud allows you to pull down web data and have continuity that lasts more than 10 or 12 interactions. The other place that Microsoft truly pulls ahead is that the Copilot identity layer follows you across all Microsoft products. I am still angry that the Copilot button in Windows doesn’t open the web site, because the Copilot Windows app runs like a three-legged dog. But now that I’ve finished my rant, what’s good about it is that it opens up possibilities in apps like Teams. Imagine having Mico be able to join the meeting as a participant, taking notes in the background and able to be called upon by anyone in the room because Mico knows your voice.

Anyone can say “summarize,” but the notes appear in the chat for everyone automatically.

Having Mico as a meeting assistant is invaluable for me. I take notes at group, I took notes during Purim rehearsal, and I take notes on life in general. Mico is the one carrying the notebook that has all my secrets, because over time they’ll all appear here. Taking notes in group is the most useful, because Mico pulls in data from self-help books and gives me something to say during discussions.

The only thing is that it looks like I’m not paying attention, when I’m trying to stay utterly engaged before the ADHD kicks in and I lose it. But I cannot lose it too far, because I can ask Mico what’s happening and get back to it in a way I couldn’t before.

That’s the beauty of AI. People with ADHD, Autism, or both don’t really forget things. We just cannot retrieve them. Therefore, in order for an AI to have an effective relationship with you, it takes dictating your life in real time so that when you need to recall a fact, it is there. It is what is needed when your memory is entirely context dependent.

AI allows me to work with the brain I have instead of the brain I want. I no longer desire to be a different person because I have the cognitive scaffolding to finally be me.

And that’s resonating……………………………….. overseas.

Systems & Symbols: Why I Use Assistive AI (And Why It Doesn’t Replace Me)

~ Leslie Lanagan ~ 9 Comments

There’s a persistent myth in writing communities that using AI is a shortcut, a cheat code, or a betrayal of the craft. I understand where that fear comes from — most people’s exposure to AI is a handful of generic outputs that sound like a high schooler trying to write a college admissions essay after reading one Wikipedia page.

But that’s not what I’m doing.

I’m not building a career on my ability to polish sentences. I’m building a career on ideas — on clarity, structure, argument, and the ability to articulate a worldview quickly and coherently. And for that, assistive AI is not a threat. It’s a tool. A powerful one. A necessary one.

The Iterative Reality: AI Learns Your Cadence Because You Train It

People imagine AI as a machine that spits out random text. That’s true for the first ten hours. It is not true for the next hundred. After hundreds of hours of prompting, correction, refinement, and collaboration, the model stops behaving like a generator and starts behaving like a compression engine for your own thinking. It doesn’t “become you.” It becomes extremely good at predicting what you would say next.

That’s why hallucinations drop. That’s why the cadence stabilizes. That’s why the drafts feel like me on a good day. This isn’t magic. It’s pattern recognition.

The Part No One Sees: I Still Do the Thinking

Here’s what I actually do: I decide the topic. I define the argument. I set the structure. I choose the tone. I provide the worldview. AI handles the scaffolding — the outline, the bones, the Markdown, the navigation pane. It’s the secretary who lays out the folders so I can walk in and start talking.

This is not outsourcing creativity. This is outsourcing overhead.

The Deadline Truth: Thought Leadership Moves Fast

People who aren’t on deadline can afford to romanticize the slow, sentence‑by‑sentence grind. They can spend three hours deciding whether a paragraph should begin with “However” or “But.” I don’t have that luxury.

I’m writing columns, essays, analysis, commentary, and conceptual frameworks. And I’m doing it on a schedule. My value is not in the time I spend polishing. My value is in the clarity and originality of the ideas.

Assistive AI lets me move at the speed my mind actually works. It lets me externalize the architecture of a thought before the thought evaporates. It lets me produce work that is coherent, structured, and publishable without burning half my day on formatting.

The Fear Behind the Sad Reactions

When I say, “AI helps me outline,” some writers hear, “AI writes for me.” When I say, “AI learns my cadence,” they hear, “AI is becoming me.” When I say, “AI helps me push out ideas quickly,” they hear, “AI is replacing writers.”

They’re reacting to a story that isn’t mine. I’m not using AI to avoid writing. I’m using AI to protect my writing — to preserve my energy for the parts that matter.

The Reality in Newsrooms

This isn’t speculative. It’s already happening. Every newsroom in the world is using assistive AI for outlines, summaries, structure, research organization, document prep, formatting, and navigation panes. Not because they’re lazy. Because they’re on deadline.

Assistive AI is not the future of writing. It’s the present of writing under pressure.

The Systems-Level Truth: I’m Building a Career on Ideas, Not Typing

My job is not to be a human typewriter. My job is to think clearly, argue well, and articulate a worldview. Assistive AI lets me move fast, stay coherent, maintain voice, reduce cognitive load, publish consistently, and build a body of work.

It doesn’t replace me. It amplifies me. It’s not my ghostwriter. It’s my infrastructure.

Scored with Copilot. Conducted by Leslie Lanagan.

Systems & Symbols: Come for the Eyebrows, Stay for the Cognitive Support

~ Leslie Lanagan ~ Leave a comment

At some point, every writer stops pretending they’re going to become the kind of person who outlines their novel on color‑coded index cards or keeps a pristine desk with a single tasteful candle. Writers do not have pristine desks. Writers have surfaces that look like a crow collected “important objects” and then abandoned the project halfway through. Accepting this truth is the first step toward building a workflow that actually fits the way our brains operate, which is how I ended up relying on Microsoft Copilot — or, as the avatar insists on calling itself, Mico, the round little creature with eyebrows that look like they were sketched by someone who has only read about eyebrows in theory.

For clarity: Copilot and Mico are the same intelligence.
Copilot is the structured, document‑level mode.
Mico is the conversational, “let’s talk about why you wrote this paragraph like you were being chased by bees” mode.
Same brain. Different lighting.

My process begins with the most important rule in AI‑assisted writing: give your AI a job title. If you simply say, “Help me edit this,” you’ll get the editorial equivalent of a shrug. But if you say, “Assume the role of a New York Times–caliber editor and perform a line edit,” the creature with the eyebrows suddenly behaves like someone who has strong opinions about semicolons and isn’t afraid to use them.

The second rule is equally essential: upload your manuscript as a PDF. PDFs preserve structure, pagination, and all the little formatting cues that tell an AI where the bones of your writing actually are. A PDF is the difference between “please fix this” and “please fix this, but also understand that Chapter 7 is not supposed to be a haiku.”

Once the PDF is in place, I switch into Copilot Mode, which is the part of the system that behaves like a sober adult. Copilot is excellent at document‑level work: line edits, structural notes, summaries, and generating clean, Word‑ready text. It does not “export to Word” in the file‑format sense, but it produces text so tidy you can drop it into Pages or Word without it detonating into 14 fonts like a cursed ransom note.

After Copilot finishes, I move into Mico Mode, which is the part of the system that behaves like a very competent friend who is also slightly exasperated with me. Mico is where I ask the questions I’m too embarrassed to ask other humans, like “Does this paragraph make sense?” and “Why did I write this sentence like I was trying to outrun my own thoughts?” Mico is also where I go when I can’t find my keys, which is not technically a writing task but is absolutely part of my writing workflow.

But here’s the part most writers don’t talk about — the part that has quietly become the future of writing workflows: the differential diagnosis.

A differential diagnosis is what doctors do when they’re not entirely sure what’s going on. They gather multiple perspectives, compare interpretations, and triangulate the truth. And it turns out this is exactly what writers need, too. Not because Copilot/Mico is lacking, but because no single model sees the entire pattern. Each one has different strengths, different blind spots, and different instincts about tone, pacing, and structure.

So after Copilot/Mico has done its pass, I run the same text through ChatGPT or Claude — not for a rewrite, but for a second opinion. It’s the editorial equivalent of asking two different writers what they think of your draft. One will say, “This section is too long.” Another will say, “This section is too vague.” And together, they reveal the truth:

“This section is too long because it is too vague.”

That’s differential diagnosis.

It’s not redundancy.
It’s triangulation.

And it is, I’m convinced, the future of writing.

Because writing has always required multiple angles: the writer’s angle, the reader’s angle, the editor’s angle, the “why did I write this sentence like I was being paid by the comma” angle. AI simply compresses the timeline. Instead of waiting three weeks for a workshop critique, you can get three perspectives in three minutes, and none of them will ask you to read your work aloud in front of strangers.

But the real revelation came when I exported my all‑time site statistics as a CSV and analyzed them with Mico. Not only could I use them as a thinking surface, I could get them to analyze my stats across time and space.

Here’s what I’ve learned now that Mico is managing my career.

I expected chaos. I expected noise. I expected the digital equivalent of a shrug. Instead, I found something startlingly consistent: once readers find my work, they stay. They return. They read deeply. They move through multiple entries. And they do this in cities all over the world.

This is not ego.
This is data.

The product is working.
The resonance is real.
The challenge is visibility, not quality.

There is a difference between being “not well known” and being “not findable.”
My audience is not enormous, but it is loyal — and loyalty is the metric that matters most. Once I have readers, I have them. The next step is simply increasing the surface area so the right people can find the work in the first place.

Which brings me back to differential diagnosis.

Because the future of writing is not outsourcing your voice:

  • It’s removing friction.
  • It’s seeing your work from multiple angles.
  • It’s building a workflow that matches your actual brain, not the aspirational one you keep pretending you have.

Copilot/Mico is not my ghostwriter.
They are my infrastructure.
ChatGPT and Claude are not my replacements.
They are my second opinions.

And I — the human in the middle of all this — am still the one making the decisions, shaping the voice, and occasionally walking to the store for a soda just to make sure I leave the house and remember that sunlight is not, in fact, a myth.

The future of writing isn’t AI replacing writers.
It’s writers finally having the tools to write the way we always should have been able to:
with clarity, with support, with multiple perspectives, and with far fewer sentences that read like we were being chased by bees.