Back to all posts
blog-postailearningai-ethicsjournalpersonal-journey

An Unfortunate Skill Issue

May 17, 20267 min read

NOT ME!

I'm posting this because I want to show the downsides of using AI for writing. I posted my Burn the Boats post because I thought it was a fun, interesting take on my writing. I enjoyed the post, it mostly felt like me, and I wanted to share both my original post and my AI written post to show people the difference. Unfortunately, this is a step in the wrong direction but something I want to show people if they don't proof read their AI's written content. When I used my blog publisher skill in Claude, it's supposed to just proof read and edit with ZERO changes in my writing. This below me, is the completely different post that it wrote up. I point this out for two reasons. Number one, this is NOT what my skill should do and is a frustrating example of how AI can be inconsistent. Number two, and somewhat more importantly to me, not only does this not sound like me, where the first time I tried this did, but this sounds 100% like AI slop. I'm seeing it daily online now and AI slop is very disappointingly everywhere. It's the verbiage, the bullet points, and there is just a certain feeling you get when you read the same generated style. So, as positive and hopeful I am about AI, I post this as a warning to not only read your outputs but also, take ownership of what you post.

The Next Step After the Conference

Coming home from AI Dev 2026 left me with a weird but useful problem. I felt more validated than I expected, but I also felt less interested in chasing whatever sounded the most technically impressive.

The conference theme was memory and agents, and I liked a lot of what I saw. I still think agents are important. I still think memory will be a big part of useful AI systems. But I also walked away realizing that most normal businesses are not asking for a multi-agent architecture. They are asking, even if they do not say it this way, "Can this help me with the work that is already driving me crazy?"

That feels like a better place for me to start.

I have spent the last year or so trying to catch up on AI, programming, and all the technical foundations I ignored for a long time. I have taken a ridiculous number of DeepLearning.AI courses. I won their most engaged learner award at the conference, which still feels strange to type. But the more important realization is that learning by itself is not the business.

At some point, I have to turn the learning into something useful for real people.

Why Small Businesses

I keep coming back to restaurants, wine, hospitality, and local businesses because that is the world I actually know.

I know what it feels like when a shift is understaffed. I know how much important information lives in someone's head. I know how often training is inconsistent because everyone is busy. I know how hard it is to make documentation when the work itself never stops long enough to document it.

That is probably a better starting advantage than trying to out-engineer people who have been coding since they were ten.

My edge is not that I am the best developer in the room. I am not. My edge is that I understand the problems of a service business from the inside, and I have been learning enough AI to start connecting those problems to practical tools.

What I Would Teach First

If I were sitting down with a small business owner, I would not start with agents, RAG, fine-tuning, transformers, or any of the words that make AI sound more complicated than it needs to be.

I would start with three things.

First, I would teach them how to use AI as a drafting partner. Emails, menu descriptions, staff notes, event copy, review responses, training material, and checklists are all places where AI can help without taking over the business.

Second, I would teach them how to use AI as a thinking partner. Not "make my decisions for me," but "help me organize this mess." A lot of business problems are not mysterious. They are just buried under scattered notes, half-finished documents, and things people keep meaning to clean up.

Third, I would teach them what not to put into AI. Customer information, employee issues, financial details, legal problems, and anything sensitive should be handled carefully. AI is useful, but it is not an excuse to stop using judgment.

The First Real Offer

The first thing I am building is an AI readiness audit for local businesses.

That sounds more formal than it is. Really, it is a structured conversation:

  • What work happens every week that is important but annoying?
  • What questions do staff or customers ask over and over?
  • What information lives in someone's head instead of somewhere useful?
  • What do you write, rewrite, summarize, or explain repeatedly?
  • What should stay human no matter how good the tool gets?

The output should be one practical quick win. Not a giant transformation plan. Not a pitch deck. Just one useful thing that could be tested quickly.

Maybe that is turning messy wine notes into staff training material. Maybe it is summarizing customer reviews. Maybe it is drafting private event replies. Maybe it is building a small assistant that can answer questions from a menu, wine list, or training guide.

The point is to stop talking about AI generally and start finding specific places where it actually helps.

What I Am Building First

The first demo will probably be a wine or restaurant knowledge assistant.

The simple version is this: give it a small set of notes, menus, wine descriptions, or training documents, then ask questions in plain English. It should help explain a bottle, summarize a menu item, suggest pairing language, or turn rough notes into something a server could use before a shift.

This is not meant to replace a sommelier, chef, manager, or experienced server. That is the wrong frame. It is meant to make the knowledge already inside the business easier to use.

That matters because good hospitality depends on details, but the details are often scattered everywhere.

What This Is Not

This is not me giving up on the bigger company idea.

I still care about Limbic Logic and the idea of using AI to predict fear-based freezing. That idea still feels meaningful to me. But I also have to be honest: it is a long, hard, research-heavy path. If I make that the only thing I work on, I can already feel the paralysis creeping in.

So I am splitting the work.

The main track is local AI enablement: teaching, auditing, building small demos, and talking to real businesses.

The side track is Limbic Logic: one focused block per week, with the next milestone being a clean report on the data I already collected.

That feels more sane. It keeps the big idea alive without letting it become an excuse to avoid doing the next obvious thing.

The Actual Goal

The goal for the next couple months is simple:

  • Publish three practical posts.
  • Build two small demos.
  • Talk to at least three real operators.
  • Find one person who says, "Can you help me with that?"

That would be a real signal.

Courses are useful. Ideas are exciting. But feedback from actual people is the thing I need next.

So that is the direction now: less abstract AI ambition, more useful work for real businesses.

I think that is where the next version of this journey starts.

Related Notes

  • Previous post: [[Post 6 Post Conference]]
  • Sprint hub: [[Local AI Enablement Sprint]]
  • Audit framework: [[AI Readiness Audit for Local Businesses]]
  • Demo planning: [[Demo Ideas for Local Operators]]
  • Long-term project lane: [[Limbic Logic Weekly Research Lane]]