Wednesday was the second meeting of the Dallas AI Club. The room felt different from the first — looser, more willing to drop the work talk and get to whatever was underneath it.
The week’s confession came early. One person said, with the kind of relief that arrives once you finally name a thing, that he had a dozen viable business ideas, was excited about all of them, and had not actually started any of them. He kept opening a tool, getting curious, opening another tab, getting tired, closing everything. The room exhaled because half the room had been doing the same thing that week.
Someone in the room gave the feeling a name: opportunity paralysis. The phrase landed because everyone recognized it. Another voice added what he called the boot-up cost of switching between AI projects — the mental load required to re-enter each thread, with its own context, its own tools, its own state of half-finishedness — and pointed out the irony that this cost is so high that the rational response is, more and more often, to do nothing at all. The shame, several people said, was in calling that laziness, when what it actually is, is overwhelm wearing a different costume.
This was the through-line of the night. We are not drowning because the tools are bad; we are drowning because the tools are so good that every direction is plausible.
The hardest skill anyone in this room is currently being asked to develop is the skill of knowing what *not* to build.
We held the meditation, as we did last week. Before we did, a small bit of news from the group: the phone that led us through the practice in the first meeting had been run over by a car in the days since. Memento mori, someone said, and we laughed at how on the nose it was. Another member had sensed the gap and loaded the meditation onto his own phone before he arrived, so the medium failed and the practice continued — which felt, without anyone making a thing of it, like the answer to something we had not yet asked.
Knowing we will die, how do we want to live the next hour? The question landed in a room full of people who had just admitted, more or less out loud, that they were not in fact doing the things they cared most about. Nobody had to say much about that. The room got quiet for a beat, and then the conversation kept going.
Several people had landed on a useful working frame for AI: a super glorified intern. Incredibly intelligent, capable of doing in an hour what would take a person a week, but lacking wisdom, common sense, and any reliable sense of when it is confidently making something up. You do not hand the keys to your business to an intern. You give them a task, check their work, and stay the manager.
The practical version of this looked like cross-checking. One person described feeding the same draft through Claude, Gemini, and ChatGPT to see what each one flagged. Another described loading the model with strict personas — act as a CFO, a lawyer, the harshest critic of this idea — and using the friction as a stress test. None of it was glamorous, which I think was the point. The work of being a good manager of a glorified intern, it turns out, is mostly editing, doubting, and asking the same question three different ways.
The most concrete picture of what these systems can do in the right hands came from someone who walked us through the agent stack he had built to find his next job. A small suite of automated bots — one that summarizes his incoming emails, one that drafts LinkedIn messages to specific people in his network, one that combs through years of his project files in a Dropbox folder and assembles a custom technical biography tailored to each role he applies for. While most of us had been describing the paralysis of having too many ideas, he had quietly built a workflow that was already in motion. The gap between his clarity and the rest of our paralysis was not really a gap of skill — it was a gap of decision. He had picked one thing and let the system follow.
The conversation turned, as it does in this room, toward what all of this becomes. Someone joked — half-joked — that AI is an alien intelligence. The room laughed and then stopped laughing. The worry that surfaced was specific: a future in which agents write the blog posts that other agents read, in which résumés get drafted by bots and evaluated by bots in pursuit of jobs that may already be a closed system, with the human removed from the loop at exactly the points that used to be the human’s job.
That was the moment two postures came into view, side by side. One was to seek higher ground — move to an intentional community, learn to grow food, climb out of these waters entirely and watch the wave from a distance. The other was to learn to surf — stay in the water, read the swells, and ride what is already in motion. Nobody argued for one over the other, and some people in the room were clearly entertaining both, which I think is honest.
We held the practice again to close the night. The question lands differently when the room has just been talking about agents writing for agents in a world where humans may be optional. The meditation does not answer it. It just keeps the question alive, which is what it is for.
What I took home was that the comfort of this room is not in mastery; none of us has mastered any of this. The comfort is in not having to navigate these waters alone. The wave is moving faster than any of us can read on our own, and this is one of the few places I have found where people are willing to admit that out loud and keep showing up anyway.
We meet again next Wednesday.
If you are not yet in the room and want to be, you can request an invitation.
— Jeremy