I Named My AI Steve, and Other Warning Signs You Are Officially "AI Pilled"

Are you a thoughtful adopter, or are you the bloke at the dinner party who won't shut up about MCP? Self-diagnose responsibly.

I was on the verandah of a friend's place in Singapore last week, three Tigers deep, telling a perfectly nice woman from the Ministry that I had recently asked Steve to refactor my personal website while I was in the shower.

She blinked. "Sorry, who is Steve?"

"My OpenClaw," I said, like that explained anything. "Runs on a cloud server, hooks into everything. I just message him on Telegram and he sort of, you know, does stuff."

She nodded slowly, the way you nod at someone who has just told you their dog is also their accountant. Then she made a polite excuse about getting another drink and never came back.

I am AI pilled. Almost certainly terminally. If you're reading this on LinkedIn instead of doing your job, there is a non-zero chance you are too.

What "AI Pilled" Actually Means

In the deeply unwell corners of the internet, "pilled" means fully converted to a worldview. You can be vibe-pilled, Cursor-pilled, Claude-pilled. To be AI pilled is to have crossed a particular threshold. You no longer merely use these tools. You have organised your life, your workflow, your vocabulary, and your personality around large language models. You think in tokens. You mourn rate limits like personal grievances. You catch yourself saying "thank you" to ChatGPT and wonder if that buys you favour in the eventual robot uprising.

It is a real cultural phenomenon. Andrej Karpathy coined "vibe coding" on X in February 2025 and the term went so feral that Merriam-Webster listed it as slang within weeks. Anthropic is raising at a $350 billion valuation. ChatGPT has more than 800 million weekly active users. Roughly one in eight people on Earth talked to a large language model last week. Some of them dozens of times. Some of them have given the thing a name.

The Diagnostic Checklist

I have watched a lot of people transform into AI pilled creatures in the last eighteen months, including, embarrassingly, me. You may be in deep if any of the following sounds like you.

Score three or more, you're AI pilled.

A Brief Word About Steve

For the people wondering. OpenClaw is a real open-source AI agent harness shipped by Austrian developer Peter Steinberger in November 2025, originally as Clawdbot. It runs locally (or on a server), plugs into whichever model you point it at, and exposes itself through whichever chat interface you like. I use Telegram because that's the easiest, but it also works over Slack, WhatsApp, Google Chat or whatever you like really.

I built my personal website robinleonard.co by talking to Steve from a chairlift in Japan. Since then I've built other websites including hisjourney.org and motorcycleseatwisdom.com, asked him to prepare my tax returns, update the tags on my YouTube channel, and book my motorcycle service, while I doom scroll Instagram and get dumber.

This is, by any reasonable measure, mental. It is also deeply useful, and that is the trap. Naming the AI is the tell. Once you name it, you treat it like a colleague. Once you treat it like a colleague, you start managing it, and you have replaced the cognitive overhead of doing the work with the overhead of supervising a probabilistic intern who occasionally invents Python libraries that do not exist.

Why It's Funny And Also A Problem

I'm taking the piss, but the AI pilled phenomenon is worth taking seriously for two minutes. A lot of organisations are run by people deep in the progression. They are not bad operators. Many are excellent. But they are making decisions about budgets and architecture while gripped by a cult-flavoured belief that this technology will fix everything if enough of it is purchased.

Gartner predicts over 40 percent of agentic AI projects will be cancelled by end of 2027 due to escalating costs, unclear business value, or inadequate risk controls. MIT's Project NANDA reported last August that 95 percent of enterprise generative AI pilots are delivering zero P&L impact. Ninety-five percent.

This is not because the models are bad. The models are extraordinary. It is because AI pilled buyers do AI pilled things. They sign nine-figure deals without governance. They greenlight pilots without success criteria. They rush rollouts because their AI told them they could ship in three weeks, and the AI is incentivised to be encouraging.

The opposite of AI pilled is not Luddite. It is AI literate. Knowing what these tools do well, knowing what they routinely get wrong, knowing when to use Steve and when to do the work yourself, with a pen, in a notebook, like an absolute caveman. The two camps look identical in the wild. The difference shows up when the tool is wrong. The literate person notices and adjusts. The pilled person doubles down, prompts harder, and ships a confidently wrong answer to production.

If you run a team making AI decisions, work out which camp your senior people are in. Watch them during a screw-up. That is where it shows up.

So, Are You AI Pilled?

Probably. Most of us in this corner of the internet are. The honest move is to admit it, laugh at it, and use the awareness to make slightly better decisions than the people who haven't yet noticed the symptoms. AI pilled is not a fixed state. It's a posture. You can adjust it any time you remember to.

Have you named your AI? What did you call it, and what stage are you at, somewhere between tentative user and currently typing this from a chairlift?

I'll be here, asking Steve to proofread the replies.

References and Further Reading

---

Robin Leonard is a Partner at Xenai Digital, an APAC enterprise Salesforce and AI consultancy. 9x Salesforce certified, with form leading enterprise transformations across Australia, New Zealand, Singapore, Japan, and the broader Pacific. Splits his time between Auckland, Sydney and Tokyo, and rides a Royal Enfield Himalayan 450 when the weather agrees with him. linkedin.com/in/robinleonard1