Happy Sunday, all.
There’s a saying I love: “to touch grass”. It can mean most literally going outside to touch grass, but it can also mean to temporarily disconnect from your existing pretenses and seek some kind of equalizing or grounding mechanism.
This week was full of touching grass in both senses. I went on a vacation with family spending some disconnected time away in a place that was not full of startup founders, technical engineers, or anything close to the sort.
This wasn’t new to me, either. Usually, I find myself "working" weekends unconsciously. People would be (mostly) relaxing at cafes whereas I’d be hunkered down in the same space tinkering with things, with a weekday intensity. So, I took the opportunity to understand more of what other people’s lives and grass looked like.
Most every time I sat down at a cafe in 2026, if someone was working next to me and looked like a remote worker, I asked them if they used AI (an anchor I’m familiar with) and if they did/did not, how they used it, as well as their reasons and motives for doing so. I'd not state my own work until the process was done.
Sometimes I'd ask this process outside of cafes, and to non remote-working folks, but the population in question (cafe remote workers) carries some selection bias. That being said, some interesting commonalities emerged. From over 100 semi-structured conversations, I gathered the following:
A majority of people use ChatGPT but do not pay for it. (this locks them out of strong reasoning model capabilities)
For people who felt the most pessimism towards AI, they often felt it was either forced on them (by their company), or that it posed an assault on their critical thinking. Environmental concerns surprisingly came up less (verbally) than creative concerns about artists and notions of human originality.
For people who felt optimism, they discovered its capabilities firsthand through hands-on experience. This group made up fewer of the people who were interviewed.
Most people who do some kind of flexible remote work also:
Tend to work off their laptop (e.g. no screens larger than 16" for their daily work - meaning app space and attention is limited).
Have no idea that file-modifying, internet-browsing agentic harnesses (e.g. Codex, Claude Code) exist or how to safely configure them (beyond the web version).
One particular person: Claire, a nursing student who used ChatGPT (free, web version), did not know that she was using a model to help her study that was significantly weaker than the ones a step up. She also didn't know she couldn't opt out of her inputs being trained-on if she was logged out.
This entire process made concrete just how far frontier discourse is from ordinary humans. And I do not speak pessimistically - I reject that. I see this as a wonderful opportunity for us deep in technology (and outside of it on the very fringes) to bridge understandings.
My additional rationale going through all that question-asking and getting outside of my normal setup is this: if you're going to be building things for other people (products, services, systems), it's best to get into their space and understand what's in their heads, well outside of your own setup. Sometimes, you have to slow down a little, to see what the bigger picture is.
Touch grass.
Be well,
Michael Kirsanov