AI is fucking with us.

We’re probably not scared enough, or optimistic enough.

Bob Gower — Brooklyn NY, October 2026

One sunny afternoon in February 2025 I took a meandering walk through the Brooklyn Navy Yard. I kept an office there and found the quiet emptiness of the old shipyard a good place to wander and think. And this particular day I had a lot on my mind.

I’d built an AI tool a few months before to help me with challenges in my work as a managing director for a small consulting firm. That morning the tool had pointed out a pattern in me, not my work — rude.

As I worked through an issue the tool suddenly suggested I might be overfunctioning. I wasn’t sure what that meant at first, but as I reflected on the past year I — gradually then suddenly — realized I needed to quit my job.

I’d been taking on too much, receiving too little, and had lost enthusiasm. I was operating out of habit and anxiety and had lost faith in what we were building. I knew these things individually but the AI pointed out a pattern between them. Overfunctioning was the right word for what I was doing. My wife and therapist later agreed with the assessment.

Over the next 8 weeks I wrapped up my obligations, then stepped off into unemployment and contemplation.

For 20 years my work has been animated by one question: How can people work well together in the world as it is now? The most important story I could see was AI. Over the next few months I studied, built tools, and took every opportunity to talk with the smart people I knew and to meet more.

My experiments hinted at something deeper and stranger than I’d imagined. AI had slowly become a thought partner I valued. It hadn’t replaced my humans but it was changing what I brought to them. This was changing how I thought about myself and my work.

For a long time I’d wanted to start a company — and now, for the first time, that path seemed achievable.

It’s not an exaggeration to say that I would not have founded Splendid Torch without AI. But there’s a lot under that simple claim.

My mind is unruly and prone to boredom and frustration. I’ve rarely found a job, a title, or a role satisfying for long. And I tend to get hyperfocused for a couple of years before moving on to the next, sometimes wildly different, thing.

I’m a master of none who often steps away from something just before it takes hold. My resume reads like a pastiche. My investment portfolio is thinner than those of some childhood friends who put in 30 years with one employer and have comfortably retired to the golf course.

I could never comfortably retire to a golf course. It’s not in my nature.

In my 30s I moved to Silicon Valley where I worked in newspaper and early web design. I was in the newsroom of the SF Examiner on 9/11 and designed its front page — “Bastards!” in 220 point type, we got some notice for that one. And was for a time part of an organization best described as a cult.

I’d earned an MBA by then and been a startup product manager; the cult got me on my way out of business school. When I got out of the cult, I started the work I’ve done ever since: changing how organizations work, at places like Charles Schwab, GE, Spotify, General Assembly, and the Wikimedia Foundation.

What it costs

In the summer of 2024 I sat in on a presentation by an architect I know. His creative team had gone all-in on AI which allowed them to iterate quickly and cheaply — sometimes on stage in front of powerful community boards that could approve their work.

I’m a writer. I regularly publish business and personal essays and have edited one book and co-authored another. Like many word people I initially distrusted AI. It had stolen my books and the output was far from satisfactory.

Inspired by the architect’s process I returned to AI and by the next spring had carefully integrated it into my writing process. I found it helped me iterate more quickly too. I was intrigued.

But it was also horrifying. More than once I hastily hit publish on something that looked polished on the surface but underneath was a rhetorical turd. After receiving some valid and embarrassing pushback I re-tuned my process and increased my vigilance.

AI might help me write, but it’s my reputation that’s on the line.

I’m not much interested in automation. I can’t imagine trusting a machine to clean up my desktop, calendar, or email. And I dread the possibility that AI might think for me some day — here there be dragons. But AI helps my writing. It finds patterns I miss and darlings that need killing. That said, if you love automations please use them — and teach me.

In the summer of 2023 I had an important contract come across my desk and, like many idiots before me, figured I could use AI to review it and save some pesky legal fees. Context windows were small in that bygone era so I pasted the contract one clause at a time into GPT-4 asking it to “explain this to me in plain language.” It did a great job flagging concerning clauses. But it missed something important.

The contract sat in a context that AI couldn’t see. I had a long relationship with the person who sent it and also had other things going on in my life. My long-time lawyer knew me and would ask probing questions whenever he reviewed contracts. He was a personal friend who had my back.

GPT-4 was not a friend and did not care. It got the trees right but missed the forest. And I paid the price. Unwinding that contract turned out to cost me time, money, and sleepless nights.

I wish this hadn’t happened. But I also recognize it as a kind of tuition paid to learn something new. Sometimes you find a boundary by crossing it.

AI scares the hell out of me. It’s freakishly powerful and only seems to be getting better. I believe if current models didn’t improve at all for the next 10 years we would still be finding new ways to use them.

And AI is already good at analysis — my core work for the past 20 years. So yeah I’m scared for my ongoing relevance and market value. I’m also scared about damage to the environment and economy. And I worry I’ll be 80 and walking around in an AI-built environment I barely understand — like a monkey on the streets of NYC.

But am I scared enough? Probably not. And am I optimistic enough? Also no.

I’m not so worried about Skynet killing us all — it may happen but I don’t see much I can do about it. We may slow its development for a while, but I don’t think it will stop. Humans are forever the tarot fool looking at the stars while walking off a cliff.

That said there’s a more subtle fear that we can and should do something about.

Set and setting

Timothy Leary said that “set and setting” is the key to having a good LSD experience — an idea that’s central to modern psychedelic therapy. The gist is that your environment and frame of mind, not the substance, is what determines whether your trip will be good or bad.

And some are quite good — I’ve had them myself. A study at Johns Hopkins found that one session of psilocybin could create a lasting reduction in existential anxiety and depression in terminal cancer patients.

But here’s the thing: you don’t take a dose of psilocybin the same way you take aspirin. The value is in the experience the drug induces, not the metabolic impact it has on your body.

Today most organizations are taking AI like aspirin, when it’s more like psilocybin. We plug a tool into a painful spot in hopes it will heal the pain. But AI is a cognitive tool. It changes how we think. And organizations are groups of people that think together.

While I’m scared of AI, I’m mostly scared that we’ll use it unconsciously and without clear intention. But that’s what we humans usually do — leap first then wonder about the mess we find ourselves in.

Wednesday by 3pm

In ’98 I had to train my replacement running design and production for a weekly newspaper in San Jose. Because this was Silicon Valley in the 90s, it was hard to find someone who wanted the job — tech was hiring everyone, including me. We settled on an enthusiastic but inexperienced young man. And I had two weeks to teach him how to coordinate the work of 40 people to get a paper to press, in good shape, each Wednesday by 3pm.

During my tenure I’d built a lot of systems, mostly undocumented habits, in response to a crisis or a fuck up. The work is straightforward but chaotic and requires quick adjustments and thinking on your feet. As I trained the new guy I quickly realized how many workarounds and edge cases there were to the processes I’d built.

I was good at this job. But looking back I know that using AI to automate even small parts of it would be far more challenging than training a human.

Automation — AI running things for us — is a common promise from tool makers and pundits. We’ll have armies of tireless employees who never ask for a raise or take a day off. But to automate a process we need to understand it. And most companies don’t understand how they work.

Years ago I worked with a group at John Deere designing air and fuel systems for the next engine release. Their job was to get as much power as possible while staying emissions-compliant, and every design choice changed both. They were on a tight deadline and everyone felt slow. After many conversations we found that speed came down to one thing: their ability to book time in the engine bays to test prototypes. That was far from obvious until somebody named it.

Organizations don’t do what we want them to do. They do what they are designed to do. And most organizations haven’t been consciously designed at all. What passes for design is often accumulated habit built in response to a long series of fixes to the problem of the moment.

Adopting AI reflexively can be like taking a psychedelic in a bad frame of mind. And there are a lot of bad AI trips in the news.

Like the people who make them, AI agents don’t know as much as they think they know. They can be confidently wrong. Automation is a partnership between imperfect humans and imperfect machines. And even the best machine agents need humans. And humans need humans too.

Years ago when I was single and frustrated with dating, I asked three women friends to approve potential girlfriends before I got serious. I’d learned that I was not to be trusted and needed to outsource my judgement. A few months later they all approved of Alex and we’ve been happy for 15 years.

After spending time with several companies bringing AI into their workforce I’ve noticed that the way a company uses AI is often a reflection of how the leaders use it. It’s not something a CEO mandates, it’s something they model. Fear-based mandates go in and slop comes out.

It’s not malice, avarice, or Skynet we need to worry about as much as it is exhausted people making expedient decisions. This is the banality of organizational dysfunction. We fear barbarians at the gate but pay little attention to the team holding the gate together.

Outsourcing

Navigating NYC it’s important to know how to orient yourself when you get above ground from the subway. Spotting the Empire State Building or World Trade Center helps but you can’t always see a landmark.

When I started dating Alex she’d been here for much longer than me, and my sense of direction is shit, so I learned to trust her judgement. And when I was by myself Google Maps was getting good. But outsourcing directions to my wife and phone means I’m efficient but dependent.

I think AI might be the first chance we have to outsource shared cognition.

One summer in Kyoto I was on my motorcycle and crossed a double yellow to get around a line of stopped cars. A taxi behind me had the same idea and hit my back tire hard enough to launch me over the handlebars. My first instinct was to put my arms out to break my fall — and most likely my arms. Instead, I tucked, rolled, and walked away unharmed. My body remembered my martial arts training even though it had been years since I’d been in a dojo.

We’re heading into turbulence we haven’t trained for, and what’s here already has us flummoxed.

As mom told us, we should be careful about the company we keep. But now that company is machine intelligence sitting beside a lot of what we do. It’s part of the environment and is seeping into our organizations and our lives. It can take far more energy to resist than to absorb it.

I don’t know if things will get as bad as I fear or as great as I hope. But I’m pretty sure things will get weirder than I imagine.

The Temple

At a particularly low point in my life I went to Burning Man. One night, after a dose of a psychedelic, I talked with a friend about how difficult my life was, and how clear I was that “I made this.”

My friend tried giving me a pep talk. I needed truth. I wandered off alone into the night and ended up at The Temple — a quiet place for contemplation and grief. I lay back, looked at the stars, and let fear and sadness pass through me.

That night sucked. But I hit bottom so hard that I bounced.

The trip was bad but the truth was good. Taking a good hard look at myself and what I’d made gave me the clarity to know I could make something else.

I tell you this because I think we’re too caught up looking for “AI experts” or trying to be one ourselves. The people who build the models know the models, and the people who build the tools know the tools and a few use cases they’re good for. But no one knows what AI, and our weakness for shiny objects, will do to how we work together.

My company is called Splendid Torch, a nod to George Bernard Shaw who said: “Life is no ‘brief candle’ to me. It is a sort of splendid torch which I have got hold of for the moment, and I want to make it burn as brightly as possible before handing it on to future generations.”

We don’t own our organizations any more than we own our children. And my goal is to burn brightly and help us all build something worth handing off to future generations.

I can’t and won’t promise 200% growth this year. That’s cancer’s promise.