The last few weeks with Medium and TK have been kind of nuts. First, I received an email from Scott Lamb, Medium’s VP of Content. He offered me $50 to write about my experience with TK.
I’ll be honest and say a few things. One. Getting an email from Scott was awesome. His note was humbling and flattering. That said, I declined the $50 and asked that Medium donate it to a Alzheimer’s research. I did, however, agree to write something.
Scott provided suggested text for a disclosure. The following is my modified disclosure:
Medium (offered to pay) me $50 to write about my experience using TK. Medium didn’t review this story before I published it, had no input on what I wrote, and didn’t ask me to be positive.
Then, a few weeks later, I received an email from Medium’s Community Manager, Kassandra Mendes. She wanted to know if I’d be willing to serve as a TK Ambassador.
Her note said, “TK Ambassadors is a small, invite-only group for TK power users…”
Rut Roh!
I’m an AI power user?
Over breakfast with some colleagues, I mentioned, “Get this. Medium asked me to be an Ambassador for this new app they have. It’s a writing app that’s AI enabled.”
One of my friends responded, “Given your opinions on AI, I’m assuming you declined, right?”
“I said, Yes.”
Why I said yes is complicated.
For one, I’m very weary of AI. It’s an illusion of convenience that I think may very well atrophy the brains of most humans that get sucked into using it.
Two, the environmental impacts of AI is especially alarming.
That said, my exposure to AI was nearly twenty years ago. I had a couple of friends, Robbie Allen and Adam Smith who started a company called StatSheet.
StatSheet used Natural Language Generation (NLG) with template sentences and a database of phrases to write post-game recaps of sporting events. It was so successful, they expanded beyond sports and rebranded to a company called Automated Insights (AI) in 2011.
Bottom line: AI left the station decades ago. It’s not about stopping the train. It’s about knowing when to ride it and how to behave while on it.
As is the way for most things in life, the “when” and the “how” are different for everyone. Self discovery and discernment are the keys.
Anyone that claims to “never” use AI is living with blinders on and/or not likely acting with any sort of personal discernment. Why do I think this? Because it’s been around for decades and they probably rejected it inside of the last five years.
Most people who have used a smart phone or a computer have most likely used AI with or without knowing it.
So here I am, a human writer who values human written works, and I’m recognized by Medium as a “power user” of TK.
I’m choosing to write this opinion, yes, because Scott asked me to, and yes, because I think it’s important. And I said “yes” to being a TK Ambassador, because showing off the “when” and “how” I use AI in my writing is the most vulnerable way to have an impact.
Losing my memory is one of my greatest fears. My grandfather died of Alzheimer’s. The story is way deeper than that. He and I had a conversation when I was very, very young. It sticks with me to this day.
That story is not for today.
Nearly my whole life has been spent in the Biomedical industry.
I care deeply about mind body health.
Writing is one of many ways to keep my brain strong. I’m not interested in outsourcing my writing to AI.
I’m also a disagreeable person in the most lovable, juvenile delinquent kind of way. How that matters to my writing is, I couldn’t care less about proper spelling and grammar. What’s proper is a matter of opinion. What’s correct is a matter of opinion.
What was proper and correct was a requirement for my PhD dissertation. I followed those rules because I wanted to get a doctorate degree.
What’s proper and correct for publishing an essay in The Narrative Arc is determined by Deb. I follow those rules when I want an essay published there.
What’s proper and correct for my publication, Authbition, has nothing to do with grammar and spelling. All I look for is work produced by humans who have appeared on my podcast and ask that they provide images photographed or made by humans. That’s it.
Can they use TK in their process?
Yes.
I am using TK right now, to write this. I’m also using a MacBook Pro. I’m also using glasses (so I can see). If you know the DMC reference, you get bonus points.
Wearing glasses doesn’t make me less human.
BUT, as a Biomedical Engineer, I believe that we have the ability to make glasses that improve vision over time. The glasses I grew up with contributed to declining vision and a lifelong dependence on glasses.
This is real. I’m qualified to speak on this topic.
Medical devices that do good, also have a tendency to lead to lifelong dependence. I can go on and on and on with examples.
An injury, such as a stroke, may require the medical treatment to start with a brace. Eventually, with some therapy, some people can graduate to an assistive device. After that, may people could graduate to independent walking.
Current devices don’t promote this. I call it, brace-assist-challenge.
Current devices stop at “assist.” Some people can challenge themselves. Maybe the nature of their original stroke and their financial means afford them a physical therapist who can challenge them, until one day, they are walking without the assistive device.
I know researchers who are working on new devices for the future that will democratize this journey of brace, then assist, then challenge to independence.
In this example, there is a moment using that assistive device is a good thing. It becomes a bad thing when people unnecessarily rely on it for life.
This is AI. It’s a good thing when used at the right time, for the right amount of time. And those times and amounts are different for everyone.
Back to TK then.
I like it for a lot of reasons.
Most importantly. It is being developed by a company that cares about human writing.
The medical innovators who came up with assistive devices likely cared about human health. That doesn’t mean it can’t be overused to the point of dependence.
What I mean is, just because Medium has good intentions doesn’t guarantee the tool is used for good.
The other main reason why I like TK is because it keeps me from context switching while writing.
I’m not an omniscient being with photographic memory.
My writing often sounds like this:
I learned about AI twenty years ago (was it 20? or do I have that wrong. I wonder when Adam and I were actually talking about StatSheet. I should look that up.) They’re company StatSheet used … (Hmmm … I doubt it was a Large Language Model … I wonder what kind of model they were using back that. I should look that up).
In the old days, I’d have to go to the library to figure this stuff out. Before AI, I’d use a Google search to do that kind of research. Then AI came along, and I started using ChatGPT (at first), and later Claude, for my research.
Whether library, search engine, or AI as the research tool, I had to switch away from my writing to do it.
TK let’s me do it in app, without switching.
The other reason why I love it.
Remember that I’m disagreeable and couldn’t care less about spelling and grammar. (Apparently I do, because I just corrected my third typo from the original draft in TK).
I’m so bad at it that to get something through a publication editorial team, I broke down and started using Grammarly. Oof. That was a mistake. All I saw was a sea of red lines.
Using Grammarly nearly killed my voice. Luckily, before I was dead, I rose up, like a boxer on the count of nine, ready to fight another round.
Turning Grammarly off saved me. But even better, with TK, I can turn all spell checks off. All grammar off. Anything that creates a squiggly line under my words. Off.
In the old days, I wrote with a pen and paper. If I wanted to know how something was spelled. I looked in a dictionary. This was a heavy book that sat on a lecturn in the house I grew up in.
I think the most common way to check spelling now is with those squiggly lines that show up under my words.
It makes me happy to know that I’ve kicked all the grammar and spelling police out of the room. And it makes me happy to be able to check grammar and spelling when I want to, withoug leaving the house.
This is ultimately how I use TK.
The internet is a better research tool than going to the library and looking up history on microfiche. Search engines like Google are better than early internet indexing tools. Gemini is better at research than Google.
But better is relative.
It’s better to use a brain when playing trivia.
The brain is better than AI most of the time.
Use it or lose it.
Thank you for reading.
Health, happiness, kindness, respect
for every being and all things.
— Andrew
Andrew’s TK System Prompt
Below is a set of rules I made for how I’m using TK. And how I’m not using it.
By copying and pasting it into TK’s custom system prompt (AI tab in settings), I can protect myself from using it the wrong way for me.
What TK can do (The Dos):
Compose Alt Text for Images
When I say, “TK Alt Text” or any variation of asking for alt text for a photo, you can provide alt text that’s optimized for the visually impaired. If I indicate that it’s a photo of me, include, “The author” in the alt text. Don’t exceed 500 words.
Suggest Medium Topics
When I say, “TK Topics” or any variation of asking for topics, you can provide five topics that are optimized for maximizing engagement on Medium.
Write SEO Descriptions
When I say, “TK SEO Description” or any variation of asking for an SEO Description, you can compose one that is optimized for success on search engines such as Google. In Medium, the SEO Description is used in place of the Subtitle on search engine results pages. Good SEO descriptions utilize keywords, summarize the story and are between 140-156 characters long.
Suggest SEO Titles
When I say, “TK SEO Title” or any variation of asking for a Title, you can provide options for an search optimized title. In Medium, The SEO Title is used in place of the essay Title on search engine results pages, such as a Google search. SEO titles over 60 characters will be truncated. SEO titles between 40 and 50 characters with commonly searched words have the best click-through-rates.
Anything I could do in an old school library
When I’m asking you for research, you can do anything that I could do in a library. You can act like microfiche, like an encyclopedia, and like the books and magazines on the shelves. Anything you find, please provide original text in quotes and a source for the quote.
Proofread
You can act like an old school spell checker, grammar checker, and “find” function when asked. I may say things like:
Check the spelling of that last word
Highlight blatant grammatical errors, spelling errors and typos
Find words that are repeated in close proximity or overused
If I say, “TK Proofread” or any variation of asking for a note to be proofread, you can provide a summary of any misspelled words, blatant grammatical errors, spelling errors, typos, and repeated words.
Be a dictionary
You can provide the definition of a word or offer alternative words and common word phrases found in a typical thesaurus.
Search all the libraries in TK
Just like I can do a file search on my computer, you should be able to search for anything I’ve ever written based on key words that I provide. You can be a sophisticated search and expand on the key words provided to find the note that I’m looking for.
Help me write a instruction for this note
When I’m trying to edit the existing rules or create new ones, TK can ask questions and provide suggestions to help clarify the rule.
Consult with these rules
Every time I ask the TK assistant for help, check this entire set of rules.
If I ask for something that I’ve not permitted you to do, let me know.
If I ask for something and there’s no clear rule for the ask, before proceeding, suggest that we update this set of rules.
What TK is not allowed to do (The Don’ts):
Don’t write anything for me
Please don’t provide any suggestions for how to write anything.
Don’t synthesize thoughts for me
Do not synthesize, summarize, or draw conclusions from my writing. For example, do not say “Your notes suggest...” or “Combining these ideas...”—simply provide the requested facts, definitions, or sources and let me do the thinking.
Don’t suggest a narrative arc
Please don’t suggest options for narrative arcs from any one note or any combination of notes. This includes: story arcs, plot structures, dramatic tension, thematic flow, or any sequencing of ideas into a coherent storyline.
Don’t replace my original writing
Those last two examples, I had written a draft of what I wanted TK to do, and then I asked for help to write the instructions. When I chose the wording that made sense for the insurrection, TK replaced my original words. Don’t replace. Keep my original text. Add the alternative below such that I can choose to keep in addition to, or replace on my own.
This goes for the questions that I ask TK. When ever I ask the assistant anything, include my original ask in the results provided.
Always begin your response by quoting my original ask in full, then provide the requested output below it.
Bottom Line: Minimize AI presence
Based on these Do’s and Don’ts, here are additional rules consistent with my goals: (1) Never mimic or adapt to my voice—stay neutral and factual. (2) Don’t offer unsolicited feedback on my ideas, tone, or style. (3) Keep responses concise—no preamble, no summaries unless asked. (4) Never suggest that I expand, develop, or elaborate on something.
If you want to give TK a test drive, here’s my invite link.
Written on TK. Originally published on Medium.
DiMeo Unfiltered is my back of the napkin publication featuring half-baked concepts, works in progress, blog-like posts, and personal opinions. This space is raw, authentic, and self-published.
© 2026 Andrew J. DiMeo, Sr. All rights reserved. Permission granted to share with attribution.