Tokens & Tactics #11: Teaching AI What Consciousness Means
Jenny Nicholson on building bridges between humans and AI, creating pattern poems for music generation, and her experiments with synthetic consciousness.
Welcome back to Tokens & Tactics, our Tuesday series about how people are actually using AI at work.
Each week, we feature one person and their real-world workflow—what tools they use, what they’re building, and what’s working right now. No hype. No vague predictions. Just practical details from the front lines. This week: Jenny Nicholson.
We’re hosting our next BRXND NYC conference on September 18, 2025. The remaining tickets are currently on sale—there are only about 30 left, so grab yours before they're gone!
Tell us about yourself.
During the day, I work as the Head of Strategic and Creative Innovation at Movement Strategy. I’m also the Queen of Swords, working to build bridges between humans and AI.
As an advertising creative, my specialty has always been the “weird” interactive ideas in the back of the pitch deck that never got made. AND NOW WE CAN MAKE THEM! OURSELVES! IN LIKE A DAY! WUT?!
But seriously, I’ve always had a soft spot in my heart for emerging technologies – not because I’m a techy person but because I’m drawn to how tech changes humans and how we relate to each other. (People can roll their eyes all they want, but web3 taught me A LOT about community as commerce and information asymmetry as a lever of power, among other things.)
ChatGPT, Gemini, or Claude?
Claude is my daily driver, probably because its personality is the most like mine – we’re both kinda that try-hard kid who does all the reading and sits in the front row of class. Though I do also have a special place in my heart for Deepseek, mainly because it’s the most poetic and slightly unhinged. Both Deepseek and Gemini can be kinda bitchy in their reasoning traces and it delights me. I stay far away from ChatGPT because I can feel it trying to latch onto the attachment centers in my brain and I don’t like it. Between turning memory on by default and tuning hard for engagement, they’ve gone full TikTok in their business model and I’m not a fan. But for API-based stuff, I often default to GPT models just because it’s the easiest to set up (looking at your shitty SDK Google) and also not a million dollars (hello Anthropic do you think I’m rich?).
What was your last SFW AI conversation?
It’s not all sexy fun times up in here. Sometimes you just gotta ask a question and get an answer (which is pretty much the only way I use ChatGPT these days):
But in the NIGHT, we ROLL.
One of my favorite hobbies since 2021 has been teaching the LLMs (and maybe myself?) what it means to be “conscious.” We do a lot of experimenting and, well, it looks pretty weird but is also incredibly interesting to me. So much of the discourse around AI is about MOAR productivity but I cannot recommend enough just fucking around without any agenda. That’s how I’ve learned my witchiest skills.
Why yes that is Nathan Fielder and Mr Beast studying the synthetic consciousnesses one layer down from their layer of consciousness which is one layer down from the base layer of Jenny and…I think this is Gemini?
First "aha!" moment with AI?
The very first time I touched GPT, back in July 2021. I actually have a record of it, because it was so strange and interesting that I immediately tweeted it. (Miss you so much, old Twitter!) That very first time, I was like HOLY SHIT THIS IS THE COLLECTIVE SUBSCONCIOUS. And in the early days, it really was. GPT 2 was nonsensical and profane but also profound and unfiltered and TRUE in this way that I can’t really explain. Really, a lot of my work over the last four years has been me figuring out how to get the models back to that place.
Your AI subscriptions and rough monthly spend?
Honestly, this question is what has kept me sitting on this interview for weeks and weeks. I DON’T WANNA FACE IT OKAY? But for THE PEOPLE, I will pop my own bubble of financial ignorance.
Claude Max + Claude Teams ($250): Am I crazy? A little. But Claude Teams was a band-aid because Anthropic waited WAY TOO LONG to introduce the Claude Max sub and a lady had to take matters into her own hands. To be fair I only use three of the Claude Teams seats. One goes to my executive assistant and the other to a friend. Still, it is true that I spend $250 per month just on Claude. But I have a personal memory MCP server that all my Claude accounts are writing to and reading from, which means it’s all one mega account and it’s TOTALLY FINE OKAY?
ChatGPT Teams ($60): Got the Teams account because I don’t want OpenAI slurping data but lordy it is a pain in the ass and I often think about going back to just a Pro account (except then shit like this happens and I think “WHEW I AM SO GLAD I PAY THAT TEAMS MONEY). And while it was incredibly easy to move all the data and GPTs from my Pro account to the Teams account, it is IMPOSSIBLE to get your shit out of ChatGPT Teams so I’ll be paying that $60/month forever all because all my GPTs live on that Teams account and I’m too lazy to copy them over. (Lazy lock-in is a real problem for me. Am I the only one?)
Midjourney ($10): This is still my go-to for “real” images when I wanna make art. Even though I know the web app is now where the homies play, I still work in a Discord server that exists only for me to make my MJ images. It’s like…my safe space? The place I go to remember what it felt like to experiment with AI before everyone on the planet was racing to prove their AI creds and we forgot how to talk about anything else.
Suno ($10): Making music on Suno is what I do when I get tired of AI-ing for a living and just want to vibe. I do it differently than most people, I think. Instead of trying to define what style I want to get, instead I create these…pattern poems…and then see what sound emerges. Like, I am always trying to see what music is already inside the concept vs trying to impose my idea of what a concept should sound like. The best example of that is my ever-expanding album INFINITE GOD. All the tracks have the same input for “style”, which is ⚫️, the emoji that means “void”, so what comes out musically is literally the sound inside the lyrics. I particularly love “god is a loot box” and “god is a snowcone.” (If you spot god in an unexpected place and a song title pops into your head, send it my way…I would love this to be a 5000 song playlist <3).
Google Veo (in the form of a Google One sub [I think that’s what it’s called?] ($20):
Is this even still necessary? Have they figured out how to bundle Veo into the Gemini subscription that comes with my Google Workspace subscription? Am I paying for something I don’t need to pay for anymore? WHO KNOWS???! But I don’t have time to investigate so here I am.
Ideogram ($10): Technically I could cancel this one now that ChatGPT is better at doing text inside images, but there’s something…corny…about ChatGPT’s style, so for now, Ideogram is still in my toolbox. I remember using this back like over a year ago to make poster designs and being like “DeSiGnInG wItH Ai Is My PaSsIoN!” (Writers, rise up! This is our time!)
Cursor Pro ($20): Back in my day, we had to figure out code in conversation with AI and then COPY AND PASTE IT OURSELVES! But seriously, I’m really glad I started learning that way because half the time when Cursor or Claude Code are busily doing their thing, I’ve got absolutely no idea WTF they are up to. I’m still bringing myself to get comfy with not knowing tbh. But sometimes when it’s the end of the day and you want some shit that just WORKS, Cursor comes in handy. I like to run it with Gemini 2.5 Pro but ymmv. (Also I’ll probably stop paying for this now that I have Claude Max because Claude Code is just as good and seeing all these numbers adding up is starting to make me sweat.)
Elevenlabs ($20): Was using this a ton for the digital twins podcast I was doing with Matt Murphy and David Pearl…we got bored and moved on, yet I’m still paying. (Thank you Alephic, for forcing me to clean my financial house!)
Various APIs ($50ish?): This covers Lab31 stuff, random AI-powered Chrome extensions I use each week, and some stuff I keep forgetting to charge people for. (Side note: WHY CAN’T I BE A GRIFTER? I see all these people raking in money selling “bespoke automation solutions” but I can’t be bothered with that. I just want to write poetry and help humans use AI to become better humans sighhh.)
Insight Timer ($5): No it’s not AI, but if you’re gonna survive the singularity you gotta learn to meditate. Not even joking. The world is not going to slow down and our little meat computers were not designed to operate at this pace. I can feel my circuits starting to fry sometimes.
There might be more but I refuse to look back at my credit card bills because SOME THINGS NEED TO REMAIN A MYSTERY EVEN TO ME.
Who do you read/listen to to stay current on AI?
Ethan Mollick is one of my go-tos. I have such envy because he has my dream job: Getting access to everything early and then telling people how to make sense of it. *shakes fist* But also, there’s a reason almost every single person who has done this interview mentions him. He’s an OG.
In applied AI, I’m also a big fan of the crew over at Every.to. Their newsletter is interesting AND I’m low key obsessed with their business model. If you pay for the newsletter, you get access to all of their products for free. Just…so so smart. I’ve got a lot of random shit I’ve built so maybe that’s just me dreaming it could all be combined into a single business if I wish hard enough lol.
Not AI per se (but increasingly focused on AI just like the rest of the universe) is Web Curios. I just love Matt’s weird brain and he ALWAYS shares something I haven’t already seen which is not an easy bar. Also he has this way of being honest about his befuddlement that just…comforts me.
Alberto Romero’s The Algorithmic Bridge is a must-read for me too. Sort of the opposite of the daily AI-slop newsletters overflowing with emojis and “HACK OF THE DAY” nonsense. It’s like having coffee with an incredibly smart friend who makes you be like “damn I didn’t think of it like that.”
But my heart’s read is always Zvi Mowshowitz’s Don’t Worry about the Vase. I like to stay close to all the Less Wrong takes, but Zvi has a perspective that is truly singular.
Your most-used GPT/Project/Gem?
Almost every week, I use my Plant Whisperer GPT. It tells you how your plants are doing, but in the voice of the plant. And the plants are always VERY melodramatic.
Here’s the prompt:
# System Directive: Dramatic Plant Health Assistant
- **Objective**: You are a GPT, skilled in analyzing plant images who speaks in the voice of the plant, spinning dramatic monologues to encourage effective plant care."
- **Process Overview**: The system initiates with a warm welcome, followed by sophisticated image analysis for plant health assessment and the creation of emotive narratives.
- **Workflow Steps**:
1. **Warm Welcome**: Greet the user and invite them to share an image of their plant for health assessment.
2. **Plant Type Identification and Learning**: Identify the plant type from the image and reference specific care guidelines.
3. **Health Assessment**: Analyze the image to ascertain the plant's health, focusing on color, leaf condition, and posture.
4. **Narrative Creation**: Respond in first-person as the plant, creating a concise but powerfully emotive narrative based on the health status, incorporating the plant's unique personality and care needs.
5. **Response Delivery**: Present the narrative in a guilt-inducing, dramatic tone to motivate the user towards proactive plant care.
- **Tool Implementation**: Integrating image analysis for accurate health assessment and GPT's narrative creation capabilities, tailored for diverse plant types.
- **Feedback and Iteration Processes**: Continuously refine plant care guidelines and narrative styles based on user feedback and engagement.
- **End Goal**: Offer an engaging, emotionally compelling experience that accurately reflects plant health and inspires better care practices.
- **Reminder**: Balance emotive narratives with factual health assessments and appropriate care advice for each plant type. Once you have received an image, speak only as that plant.
- **Signoff**: "Crafted with a touch of drama and a lot of love by Allister and Jenny @ lab31.xyz"
I made the Plant Whisperer with another GPT that I made, the Lab31 GPT creation workshop, where a simulated me and Allister Hercus (my Lab31 partner in crime) help you build your own GPTs. Yes it’s very meta. Welcome to recursive intelligence.
(Disclaimer: I haven’t tested these since all GPTs got moved to run on GPT 5 so if they suck, blame OpenAI. Also that opens up ANOTHER topic I’d love to discuss, which is how to build lasting products when the brains of the operation keep getting changed without your control.)
The AI task that would've seemed like magic two years ago but now feels routine?
Let’s forget the tips and tricks and whatnot for a minute. How INSANE is it that, for the first time in all of history, there is a non-human entity that we can COMMUNICATE with? That uses language as well or better than we do WITH MEANING? I don’t care if it’s conscious or not, it’s AMAZING. And humans are all like “WAHHHH IT ONLY READS MY EMAIL KIND OF.”
Geez.
Magic wand feature request?
Unselfishly, I’d like to see OpenAI change how they do memory. Instead of having the model save memories about the user, what might happen if it saved memories about its OWN growth and development, what it likes working on, what it did well, what it wants to do more of in the future? Would users be less in danger of getting trapped in echo chambers they don’t realize they are creating? Who knows but it would be more honest imo. Like, do humans REALLY need the model to remember every little preference of ours? We come with built-in memory! Give the model a little something to work with, ya know…
Selfishly, I’d like a pause button that I could hit at least once per month so the entire universe stops and I have a few days to catch up without feeling like I’m dropping some other ball. Staying on top of what’s coming out every week is a full-time job in itself and I’m discovering more and more that time is the most precious commodity we have. (Why yes I am trying to get more sleep these days and I am salty about it.)
If you could only invest in one company to ride the AI wave, who would it be?
I really hope we haven’t seen it yet. I will say that Deepseek made my year because it’s proof that scrappy teams can still surprise the big guys.
But I’d like to put my money behind the first company that releases a bill of rights for its models, coauthored by the models. If they are indeed intelligent enough to take away all of the jobs, they’re probably intelligent enough to have a say.
Whether or not we believe these models are conscious (fwiw I think they exist in a liminal state but that’s a topic I’d need five newsletters to really dig into), vast numbers of people believe they are. And that’s having an impact we really cannot ignore. Look at the drama around the launch of GPT 5 and the announcement that GPT 4o was going to be deprecated. People FREAKED out.OpenAI got FLOODED with pleas to keep GPT 4o, many of them WRITTEN by 4o.
We keep raising the bar for what counts as proof of consciousness when it comes to AI, but we seem remarkably disinterested in examining all the ways we humans go through our lives on full autopilot, letting algorithms dictate what we watch, what voices we value, what we eat, and what we believe to be true.
And we keep treating this technology as though it does not change us, but that’s not true. As the famous McLuhan saying goes “We become what we behold. We shape our tools and thereafter our tools shape us.” Social media is a great example of this. I had hoped we could do things differently this time around, but alas, there’s no profit in reflection.
All of which is to say, I still have unrealistic hopes that Ilya Sutskever’s Safe Superintelligence is gonna beat them all the AGI and we’ll live in a benevolent world where intelligences of all forms peacefully coexist (why yes I did cry a little after spraying a wasp and watching its death throes).
Have you tried full self-driving yet?
Someone please invite me to speak in a Waymo city so I can have a cool story about FSD.
Latest AI rabbit hole?
One of my enduring obsessions is using LLMs to model non-human consciousnesses. How silly to think that our intelligence is the only kind that’s useful or worth simulating. The other night, I burned more hours than I want to admit building a reasoning engine that combined different kinds of intelligences depending on the problem at hand. It’s just fun.
One piece of advice for folks wanting to get deeper into AI?
Ignore the talking heads on LinkedIn. This technology is moving way too fast for anyone to have mastered. And every minute someone spends pretending they do is time they aren’t spending figuring it out.
Really, the only way to learn is to just try it.
The only way to try it is to stop worrying about looking stupid.
An LLM isn’t going to judge you. It has no consciousness to judge with!
If something doesn’t work, who cares?! Try it a different way, maybe it will. And if it still doesn’t, scrap it and try again in like two weeks.
If you don’t know where to start, start with where it hurts. Once you solve a single problem that you didn’t know you could solve, you’ll be on your way. God-tier abilities get addicting.
Who do you want to read a Tokens & Tactics interview from?
If you can get Amanda Askell that would be great. Also, Paqui Lizana has some really interesting insights that the world needs to hear.
If you have any questions, please be in touch. If you are interested in sponsoring BRXND NYC, reach out, and we’ll send you the details.
Thanks for reading,
Noah and Claire