The AI Vampire — AI Is Slowly Draining Us All

This was an unusually hard post to write, because it goes against everything else happening right now.

I started noticing a new phenomenon just after the new year: people are overworking because of AI. This week, articles about it are suddenly everywhere. My theory is simple — AI is killing us all, Colin Robinson style. If you remember What We Do In The Shadows, Colin Robinson was an Energy Vampire. Being in the same room with him drained people. That's exactly what's happening with AI.

10x Productivity Is Real

Let's start with the root cause. AI genuinely makes you 10x productive once you learn how.

If you haven't used specifically Opus 4.5/4.6 with specifically Claude Code for at least an hour, you're in for a real shock. AI coding hit an event horizon on November 24th, 2025¹. Even Microsoft is openly encouraging employees to use multiple tools, and Claude Code has rapidly become dominant across their engineering org. When you give people open season, they find the path of least resistance fast.

The exact multiplier doesn't matter. The vampire effect kicks in at anything above 2x. The real question is: who captures that value?

The Value Capture Dilemma

Imagine you're the only person at your company using AI.

Scenario A. You work 8 hours a day at 10x productivity. Your employer captures 100% of the value. Your salary doesn't become 9x. Everyone hates you. You're exhausted. Congrats, you just got drained.

Scenario B. You work 1 hour a day and coast to peer level. You capture 100% of the value. Your company goes under — a competitor who's working harder takes you out.

The answer must lie somewhere in the middle, or we're all pretty screwed.

Involuntary Acceleration

The world is speeding up against its will. I grew up in the 1980s, when time moved more slowly — news was spaced out, society had room to reflect. Now we can't even keep up.

I've been experiencing "Nap Attacks" — suddenly falling asleep at all hours after long vibe-coding sessions. My colleagues at SageOx² are seriously considering nap pods.

We're addicted to a new drug and don't understand all its side effects yet. But one of them is massive daily fatigue.

The situation hits developers from all sides at once: addicted early adopters control the narrative, the news is inescapable, panicking CEOs are whiplashing AI into their orgs, and companies are extraction machines that literally don't know how to ease up. Use AI and get drained. Don't use it and get left behind.

Unrealistic Beauty Standards

Agentic software building is genuinely addictive — satisfying, frustrating, and exhilarating all at once. Every prompt is a slot machine pull with random rewards and occasional massive payouts.

People are shouting from the rooftops about what they built in 40-hour nonstop sprints. And other people are listening.

I need to take accountability here. I have 40 years of experience, have led large teams, read fast, and have essentially unlimited time, energy, and tokens. I am completely unrepresentative of the average developer. Yet I'm telling everyone "do it this way!" and even co-wrote a book about it.

Employers are looking at me and the other far outliers, thinking: "All my employees could be like that!" Dollar signs appear in their eyes. There's no reasoning with the dollar-eyeball stare.

Startups Are Poisoning the Well

AI-native startups are draining people faster than at any time in history, chasing instantly banal ideas. People come to show me their plans like end-of-movie villains — I've mastered the illusion of knowing what I'm doing with Gas Town³. Nobody actually knows what they're doing right now.

It's the same six tired pitches: sandbox systems, AI personas, agent memory, better RAG. The overwhelming majority won't sell a dollar of ARR⁴. Enterprises are quickly learning that Build is the New Buy⁵ — finance departments will start refusing to renew SaaS contracts this year.

But founders are running like it's a gold rush, setting the pace for everyone else. Enterprise executives see the horde coming and think they need to hustle too. Many are vibe-coding at home, dangerously, and reframing the problem as finding employees ripest for extraction.

The $/hr Formula

Back at Amazon, 2001–2003, everyone was working like sled dogs. One day during a heated grumble session, I walked up to the whiteboard and wrote:

$/hr

You can't control the numerator — Amazon pays a flat salary, no bonuses. But you have significant control over the denominator. I pointed at the /hr for dramatic effect. Wide eyes all around. Nobody had ever thought about it that way.

Over the next few months, I kept walking past offices where people were studying that same formula on their whiteboards.

That old formula is my proposed solution for the AI Vampire, a quarter century later.

Fighting the Vampire

Someone else might control the numerator. But you control the denominator.

Individually you may not have much sway. But collectively, the employees of your company hold all the power. Having been at the top, I've learned CEOs have surprisingly little.

Push back. Tell your leadership about the AI vampire. Educate them about sharing value capture between company and employees.

When I visited Combank⁶ in Sydney in December — their historic train station tech campus, open-plan, natural light, plants everywhere, people working, socializing, walking in the sun — it looked like the ideal balance of happiness and productivity. Just a normal Tuesday for them. That's where we need to aim.

Companies capturing 100% of AI value isn't remotely sustainable. Employees capturing 100% is temporary at best. The dial belongs somewhere in the middle. Companies will drag it higher. You need to fight to drag it lower.

The new workday should be 3 to 4 hours. AI automated the easy work and left us with all the difficult decisions. We've all been turned into Jeff Bezos. I can only sustain that pace for short bursts, even with lots of practice. 8 hours of hanging out, sure. But not 8 hours of this.

Building things with AI takes a lot of human energy.

I'm dialing back my own hours, saying No to meetings and podcasts, resisting the incessant demands. If this all comes crashing down, I won't have Regret Years to look back on.

I'm typing this at the mall, with Linh and Mozart. When I close the computer, we're going for a walk.

The date Anthropic released Claude Opus 4.5 — widely regarded as the inflection point where AI coding tools became genuinely transformative for professional software development. Two weeks later, on December 11th, OpenAI followed with GPT-5.2. ↩
Steve Yegge's current AI startup, focused on building AI-powered developer tools. ↩
An AI coding agent project created by Steve Yegge at SageOx. ↩
Annual Recurring Revenue — a key SaaS business metric measuring predictable yearly subscription income. ↩
The trend where enterprises stop purchasing SaaS products and instead build custom solutions in-house using AI coding tools, which has become dramatically cheaper and faster. ↩
Commonwealth Bank of Australia — one of the country's largest banks, known for its progressive tech culture and modern engineering campus in Sydney. ↩

Related Posts

Stay updated