A simple trading bot on Polymarket reveals the hidden force reshaping every business: AI is collapsing the inefficiencies your industry was built on.
Imagine two lemonade stands on the same street. One charges $3 and the other charges $1, but nobody walking by knows about the cheaper one. A kid with a bicycle figures this out, buys all the $1 lemonade, and sells it for $2.50 at the other end of the street. That kid is making money from a gap -- the fact that people at one end don't know the cheaper price exists.
The whole economy works like this. Businesses make money because there are gaps: information people don't have, work that takes a long time, or things that are hard to find. These gaps aren't cheating -- they're the structure of how business works.
Now imagine the kid gets a drone that flies between the two stands in 2 seconds. The gap closes almost instantly. That's what AI is doing -- to every industry, not just lemonade stands.
A real bot on a prediction market called Polymarket did exactly this. It noticed that Polymarket's crypto contracts updated their prices slower than the real exchanges. When Bitcoin moved fast on Binance, the bot bought the "obvious winner" on Polymarket before prices caught up. It turned $313 into $414,000 in one month with a 98% win rate.
In late 2025, a bot exploited a speed gap on Polymarket. The platform's short-duration crypto contracts repriced slower than spot exchanges like Binance. When Bitcoin moved sharply, the bot bought the nearly-certain outcome while Polymarket still showed 50/50 odds. It ran 6,600+ trades while humans slept.
A developer claimed to have reverse-engineered the strategy and rebuilt it in Rust using Claude in 40 minutes -- price monitoring, probability calculation, position sizing, risk controls -- all from a single prompt session. What once required a quant team, engineers, and risk managers now needs one person with a laptop and an API key.
Other bots followed: a Claude-powered system made $2.2M in 2 months using ensemble probability models on news and social data. A swarm model trained on 3 years of NBA data pulled $1.49M from sports contracts. The edge wasn't smarter strategy -- bots using identical strategies to humans captured roughly 2x the profit through flawless execution alone.
Nate B Jones lays out a taxonomy of arbitrage gaps that AI is compressing across the economy:
1. Speed gaps -- One system updates slower than reality. The Polymarket bot is the purest example. The business version: your competitor's pricing updates in real time while yours updates weekly.
2. Reasoning gaps -- Same public information, different interpretation speed. A Fed statement drops; LLMs can reason about implications in seconds while a human committee takes days. One Polymarket bot made $2.2M purely by interpreting public news faster.
3. Fragmentation gaps -- The same thing priced differently in different places. The "Big Four consultant" who charges six figures to synthesize five publicly available data sources is sitting on a fragmentation gap that LLMs close for free.
4. Discipline gaps -- Humans know the playbook but can't follow it at 3 AM. Bots don't get tired, don't oversize positions on confident bets, don't miss trades at lunch. This gap exists everywhere: sales teams, content pipelines, operations under pressure.
5. Knowledge asymmetry / intelligence gaps -- The old economy ran on labor arbitrage (SF engineer vs. Bangalore engineer). AI replaces that with intelligence arbitrage: the unit of value shifts from person-hours to outcomes. The companies winning are the ones whose people can best leverage AI tools -- not just use them, but rebuild workflows around them.
In the 1980s, CNC machining let a single operator produce in 45 minutes what took 10 hours of hand-milling. Smart shops hid the machines in the back room and charged the old rate. Then everyone got CNC machines, and prices collapsed 60-80%. The same arc is playing out in knowledge work right now. Agencies and consultants charging "bespoke" rates for AI-generated deliverables are on borrowed time.
On March 27, a configuration error at Anthropic leaked draft materials about Claude Mythos, described as a "step change" model with dramatic improvements in reasoning, coding, and cybersecurity. Markets moved immediately: the software ETF fell 3%, Bitcoin tumbled from $70K on hacking fears. OpenAI reportedly finished pre-training its own next-gen model the same week. Sam Altman told employees "things are moving faster than many of us expected."
The point: every model release rotates the arbitrage landscape. There is no post-AI equilibrium -- only permanent rolling disruption.
1. What inefficiency is my business/role built on? Name the gap. If you can't name it, you won't see it closing. (Example: product management was built on the gap that engineers were "too valuable for meetings.")
2. How fast can AI close it? Structural gaps stick around: regulatory moats, relationship trust, physical logistics, genuine creative taste, hard domain judgment. Informational and cognitive gaps are closing on a timescale of quarters.
3. What new gap does the closure create? Value always migrates upstream -- from production toward judgment, taste, relationships, and systems thinking. When code generation gets commoditized, system design becomes the gap. When content production is free, distribution and curation become the gap.