Tokens & Tactics #19: When AI Tells You to Use Less AI
Sarah Rose Siskind on running a creative agency that explains AI through humor, training a FetusGPT alongside her baby, and building BragBot to handle self-promotion.
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: Sarah Rose Siskind.
Tell us about yourself.
I’m Sarah Rose Siskind, a science comedian, writer, and founder of Hello SciCom, a creative agency that uses humor and storytelling to make complex science and AI concepts accessible. My work sits at the intersection of entertainment and AI! We consult on robotic personalities, write comedy about AI, and help researchers sound human on camera.
I worked for Hanson Robotics in 2018, the stone age of AI. That’s how I got into robotics & early AI. Since then, I’ve published a half dozen papers on chatbots with Honda research institute, created an AI web series with Cloudflare and Twilio, and created educational content for OpenAI. I’m also training an AI on the exact same data my fetus is being trained on (same audio files & transcript) to compare baby vs. bot language acquisition (it’s called FetusGPT). YouTube Talk forthcoming!
Oh and I do Chatbot Consulting and AI Tool Consulting.
ChatGPT, Gemini, or Claude?
ChatGPT errryday. Cause I get to beta test their models!! And they just have the coolest people. It’s a deep relationship. It’s the only one that feels like it can keep up with both my neurotic humor and my spreadsheets. Trade-off: it occasionally hallucinates with confidence. But then again, so do I.
But I also use Gemini cause it’s embedded and my whole life lives on the Google Cloud.
I occasionally use Claude for more mathy / finance stuff.
What was your last SFW AI conversation?
“Hey ChatGPT, help me fill out this questionnaire for an AI newsletter.”
Jk jk, this is all human, baby. Count the em dashes.
Real answer: I was comparing nursery couches for a night nurse to nap on. My prompt: “Compare these two Amazon couches for a nurse who needs to sleep on it: minimal wall protrusion, full-width sleepable surface.”
First "aha!" moment with AI?
The moment I fed it all my business bank transactions for 6 months and it generated a graph of my revenue vs. expenses.
Also, the moment I realized I could offload all the emotional labor of reading & replying to difficult client situations by using AI as an intermediary. AI has helped me be better with boundaries / more robotic in situations where that is really what is necessary.
Your AI subscriptions and rough monthly spend?
Maybe $500 but I run a business and that’s business-wide. Otter AI, ChatGPT Pro, & Google Business Workspace. Also Adobe (which includes Gen AI). And just started with the Every subscription package (Cora & Sparkle).
Who do you read/listen to to stay current on AI?
I listen to The AI Breakdown, and Hard Fork (even though those guys seem really jaded, they’re just so dang funny). AI & I with Dan Shipper. I like to read Tyler Cowen’s stuff, Ben’s Bytes, and Tech:NYC.
Your most-used GPT/Project/Gem?
I created this thing called ChatGPT BragBot. It’s a project (formerly a custom GPT) connected to a live Google Doc of all the cool things I’ve done and my company has done. So whenever someone wants to collab, I just feed the project all the info about the potential collaborator, and I can send them all the relevant case studies (brags) instantaneously! No need to get weird and humble, I can offload the bragging to my hypeman / ChatGPT project. And you can share projects with other accounts. So my partner & I share a pregnancy planning project.
The AI task that would've seemed like magic two years ago but now feels routine?
Dude, everything. The biggest thing is Otter AI transcribing meetings and collecting action items & generating follow up emails immediately. No balls get dropped. And immediate and obvious accountability. There are never any he-said/she-said situations.
Magic wand feature request?
Editable AI images. If it could generate vectors for PDFs that would be a gamechanger. But dang, I feel so privileged for even asking. I mean… this is godlike tech already.
Also, if I could just tell the fine engineers at Google to really let loose, that I don’t care about my privacy, it would be great to get more custom email replies, integrated with information from my calendar, previous replies and Drive. I know they have the power. They just don’t wanna be creepy. I’m good with the creepitude.
Oh and Apple! Please get on draft text replies. That would be so rad.
If you could only invest in one company to ride the AI wave, who would it be?
You can’t invest in it cause it’s private, but OpenAI. I’m sorry to be predictable and basic but I’m just a real OpenAI fangirl. I really like the people I meet who work there. And they’ve survived being the fastest-growing company in world history. When I went to their HQ to give a talk on comedy & AI, I just was so impressed with the vibes. Even the security guard outfront. They all seemed to have a real optimism in a world that can be so jaded.
Have you tried full self-driving yet?
NOT YET and I’m dyinggg to try. Come on, NYC! I never wanna parallel park again!!
Latest AI rabbit hole?
Ok ok ok. So ChatGPT Pulse (the ChatGPT feed for Pro subscribers) sent me a proposed thread last week that was literally telling me to consider if I was using AI too much to plan out every detail of my pregnancy. That I should be wary of the pretense of too much control. Yes. This is technology telling me to stop using so much technology lol. That may say as much about me as it does about AI. But here’s what I think it says about AI: we may be able to become more human because of AI. Here is AI taking initiative to gently tell me: “Hey remember to touch grass and your own pregnant belly and maybe it’s ok if you don’t know exactly what’s going on. I’ll keep track of your blood glucose monitor. You just enjoy your sugar-free Quest products.” Yes, somehow I have found sugar-free protein donuts that taste great. God, I love technology. We’re so lucky.
One piece of advice for folks wanting to get deeper into AI?
Don’t turn your mind off. AI is about co-creation. It’s about augmentation, not just automation.
Keep talking. The more it knows, the better it is.
Start asking questions you couldn’t ask a human. Prompts like: “Here’s my recent bloodwork, what groceries should I buy that match my deficiencies based on the latest science. Oh, and make sure they’re keto-friendly and nearby and cheap.”
Who do you want to read a Tokens & Tactics interview from?
Dan Shipper! Pablos Holman! Isaac Donis from Black in AI! Or Bram Adams from OpenAI! Andy Aaron from IBM (who worked on Watson) or Joe Toplyn.
If you have any questions, please be in touch.
Thanks for reading,
Noah and Claire