Companies saving money with AI including with pay cuts and layoffs… where are they redeploying the capital? More token budget or owned hardware for same market? Back to shareholders? Many people will likely see both income loss and capital gain. Depending on which is higher you’ll need to either find new income or new ways to deploy your now surplus capital.

One interesting effect I’m seeing is that when op-tempo goes up radically due to AI, rate of deploying capital can accelerate, but the cost of capital can go down in theory (you may be spending 10m in a month instead of a year because you’re moving so fast, but you’re saving on interest on that 10m. I think this means “speed of AI” businesses can survive significantly higher interest rate environments than speed-of-human businesses. Similar in some ways to crypto trading bots that take out big loans for seconds.

Two issues with this are a) if you’re selling into a speed-of-human value chain, wip inventory will just accumulate at the highest cost-of-capital bottlenecks, and eventually backpressure and supply choke downstream will slow everything down to bottleneck speed ie human speed b) this means, you can’t get true end-to-end acceleration except in fully AI-native value chains, which requires some consuming entity to be consuming at the speed of AI. This necessarily has to be AI agent consumers, but to close the economic loop, the human desires driving agentic consumption have to be fundamentally new. For eg I can’t read 100 books a day but my agent can. But the only value it can deliver for me is some bespoke compressed application (“read 100 books on ancient Egypt and write a custom book tailored to my interests”). And of course I’d have to be one of the people with capital gains > income losses. And unless my weird bespoke Egypt book can somehow create value for others, it’s a pure consumption sink. Like burning inflated dollars essentially. Like gas flaring.

A fascinating possible scenario here is, we can’t find fully end-to-end AI-native value chains fast enough that feed back to accelerate the whole economy, leading to the AI economy forking into solipsism like NFTs. To avoid this, the e2eAI-vcs should be directed to stuff like new pharmaceuticals or energy efficiency etc which can be discovered at the speed of AI but absorbed into the economy at the speed of humans.

There’s a bigger economic engine equation emerging here I can’t quite work out. I’ll call this the differential acceleration problem. If you can work it out, pitch me an essay for Protocolized about it