It’s 2:14 a.m. and you’re standing in a field with 20,000 strangers, bass hitting you in the sternum, phone dead in your pocket because you stopped checking it four hours ago. Nobody here is prompting anything. Nobody is reviewing a diff. The person next to you hands you a water bottle for no reason other than that you looked like you needed one.
Roughly a dozen venture capitalists looked at that scene and decided it was worth $9.5 million.
That’s the money that went into RASA World, a company that produces electronic dance music festivals and live events around the globe. The reasoning, as reported by Business Insider, is almost funny in its simplicity: the more of life that moves onto screens, the more valuable the stuff that can’t fit on one becomes. It’s a hedge. A bet placed deliberately against the direction most of the capital in this industry is currently pointing.
What this looks like from inside the build
I spend most of my working hours making software that talks to people. Routing intents, wiring tool calls, tuning retrieval so an agent doesn’t confidently hallucinate a refund policy. I like this work. I think it’s genuinely useful. And I also think the RASA World bet is correct, which is not a contradiction once you look at what bots are actually good for.
Every bot I’ve shipped that worked well did the same thing underneath the features: it removed a screen interaction that nobody wanted to have. The support bot that answers at 3 a.m. so a human doesn’t have to staff a chat queue. The scheduling agent that kills a four-email thread. The internal assistant that saves an engineer from grepping through a wiki. The product isn’t the conversation. The product is the time that comes back on the other side.
So what do people do with reclaimed time? Some of them go stand in a field at 2 a.m. The VCs funding concerts aren’t betting against the tools I build. They’re betting on where the demand goes once those tools work.
The money is sorting itself into two piles
The broader funding picture backs this up in an odd way. Capital is flowing toward specialized AI rather than generalist plays, plus defense technology, fintech, space technology, sustainable solutions, and health and biotech. Infrastructure is where the truly enormous checks land — corporate venture capital has become the single largest source of AI funding globally, outpacing traditional VC in total deployment, with Microsoft among the largest investors. Google and Blackstone are reportedly standing up an AI cloud business with $5 billion in initial equity and potential total investment up to $25 billion, built around data centers.
Set $9.5 million next to $25 billion and the concert bet looks like a rounding error. That’s sort of the point. You don’t hedge with the same size position you take on the main trade. You hedge with a small amount of money pointed in a direction that pays off precisely when the main trade goes weird.
The vertical AI trend matters for builders more directly. Generalist AI startups are having a harder time getting funded; investors want specific use cases with specific domain depth. If you’re building bots, that’s the clearest signal in the whole set of facts. A general-purpose assistant is a demo. An agent that knows the actual billing rules of one industry, wired into the three systems that industry runs on, is a business.
The design lesson I’m taking from a dance festival
There’s a failure mode in bot building I keep seeing, and the concert hedge names it better than I could. It’s the bot that tries to be the experience rather than clear the way to one. The companion app that wants you back on screen for another twenty minutes. The engagement loop dressed up as helpfulness. That kind of product is competing directly with the thing VCs just decided is scarce and appreciating, and it’s going to lose that fight on a long enough timeline.
The bots worth building do the opposite. They’re boring on purpose. They finish fast. They hand control back. If you’re measuring success by session length, you might be measuring the wrong thing — try measuring tasks closed per minute of human attention spent, and watch which design decisions suddenly look bad.
Practical version for anyone shipping this quarter:
- Pick a narrow domain and go deep. Solid retrieval over one well-understood corpus beats broad coverage over everything.
- Design for exit, not engagement. The best interaction is the short one.
- Instrument time saved, not messages sent. You’ll make different architecture choices.
- Assume the screen is the cost, not the product. Price your UX accordingly.
A dozen investors just put real money behind the idea that human presence is the appreciating asset. I’d rather build the tools that free up time for it than the ones competing with it for attention. The field at 2 a.m. isn’t going anywhere. Our job is to get people there faster.
🕒 Published: