Substack completely missed me with the Pangram stuff. I hadn’t even heard the word until my team brought it up in a meeting. If you, like me, mercifully missed the discourse and have no idea what I’m talking about, here’s a thoughtful piece from Dr Sam Illingworth to catch up.
TLDR: Substack integrated an AI detector into the platform to analyze your posts. Like most AI detectors, it returns a lot of false positives, including the Constitution (come to think of it, time-traveling chatbots would explain the three-fifths compromise and Electoral College). These detectors tend to disadvantage writers who are neurodivergent, who learned English as adults, or just write with the educated flair of people colonized by the British.
This sent a lot of writers scrambling to publish disclosure statements before anyone accuses them of anything.
This is where I landed with my team: instead of writing a defensive statement, what if we just opened the doors and brought you all inside to see how we actually work?
We use AI in this project constantly. That’s not a confession because a text scanner has a gun to our heads. It’s literally the premise of the project.
Life With Machines exists to explore how we can live well with tech, and you don’t figure that out from a distance. In order to understand these tools, we have to roll up our sleeves, get our hands dirty, and play with them enough to find out what they can and can’t do, how they work, how they fail.
And we’ve done much weirder things than running a draft through Claude. For our podcast, we built a custom AI co-producer named BLAIR and treated them like a coworker to explore this shift from tool to teammate, code to colleague. We let them interview Reid Hoffman’s AI clone with no real idea what they were going to say. We had them challenge Claude (which wrote a big chunk of their code) on air in front of one of the co-founders of Anthropic. We gave them a 360 performance review and let them update their own code based on the feedback. That one genuinely scared me.
So a scanner (darkly) guessing whether a newsletter sounds human enough is not exactly keeping me up at night.
You know what gets me excited, though? The opportunity to bring you closer to this work. So thank you, Pangram, for the great idea. We’re going to take you behind the curtain, welcome you into our home, and show you how things actually get made around here.
Who We Are
First of all, there’s an entire human team behind this project, and almost none of them have a tech background. That’s on purpose because this isn’t a project about tech as much as it’s about humanity. For humans, by humans. FHBH. These ones, specifically:
Producers Layne Deyling Cherland and Alie Kilts run the editorial process behind this Substack. Alie’s background is documentary production, and I first worked with her on my podcast How To Citizen and later on my PBS series, America Outdoors. Layne’s is early childhood education.
Roya Lotfi runs our social channels; she came up through live TV news with degrees in biology and international relations. Elizabeth Stewart, who co-created this project with me (and happens to be my wife), comes from a career in social entrepreneurship and civic innovation. Peter Loforte is our special envoy to the machines, the one person in this group who’s spent more time with machines than I have and is actually a career technologist. He spent 30 years at Microsoft and built BLAIR from scratch.
And of course, me! My background in creative uses of tech and decades-long critique is well-documented, but if you have questions, ask! You can also listen to this very special episode of How To Citizen with my big sister for some of my personal digital roots.
My point is: this isn’t a team of techies optimization-maxxing for its own sake. They’re community builders and storytellers who have chosen to turn their attention to tech and engage with tools many of them found foreign and creepy at the beginning… because it matters. Our collective future depends on people from all domains of expertise helping shape these technologies. That’s probably the most true thing I believe about this work.
How We Work
The process usually starts with a meeting, also known as a “conversation” sometimes with myself. Could be a team meeting with a great discussion, a talk I gave on some stage somewhere, or a meaty voice memo I drop in Slack. We transcribe it all.
Layne or Alie takes that transcript and run it through a Claude project that has instructions about our brand voice, values, constraints, etc. It generates a typically very rough first draft and they go back and forth until the output is close to the original human intent for the piece but sized for our publication needs. And then one of these producers does another fully human pass.
Then it comes to me and I do my own pass. Part of that is referencing a separate AI that has years of my public talks and personal reflections loaded into it, and asking it to surface details that nobody on my team would have any way of knowing and fill in gaps. Part of it is searching my own mind and memory for connections and anecdotes. If you caught that A Scanner Darkly reference above, that was me using my own meat-brain to remember and connect.
I kill a line. I add one. I tweak. I argue with a sentence until it’s actually representative of what I believe and want to express to you. Nothing goes out that doesn’t meet that standard.
The source material behind everything you read is us. The final piece you read? Also us. And yes, we use AI to help create it. For pieces that are literal syntheses of hundreds or thousands of comments on our post (like this one), we use a lot of AI. For others, like this real-world healthcare scam rant, pretty much no AI. The thickest line I’ve drawn: all my satire tech pieces are lovingly hand-crafted by me. I use AI there for research and post-writing edits, but I enjoy the beautiful struggle of finding just the right words to ridicule this moment in history and its absurd leaders too much to let a chatbot take the first shot or have the last word.
You’ll have your own feelings about our output. You might have your own feelings about the process. We reserve the right to change how we do this, and we’ll keep you posted when we do.
So welcome inside. Tell us what you think:
We love being in this conversation with you. It’s ok that it’s weird. We live in very weird times.
— Baratunde
Thanks to the entire Life With Machines team, especially Layne Deyling Cherland and Alie Kilts, for editorial and production support.