Crazy Ideas Beyond AI
We woke up and all 85 people we had requested were contacted, precisely. For people we had talked to before, the messages continued the earlier context.
A world where AI does almost everything
We needed to send LinkedIn messages to overseas buyers. We gave the agent little more than target countries and job titles and told it to go into the sales tool, find the people, and send tailored messages. By morning, all 85 requested people had been contacted, the messages read well, and for prior contacts the thread picked up where it left off. While we slept, about ten of them had even completed free sign-ups.
The same day we had to connect the office printer for the first time. Visiting the website, installing drivers, reading the guide: unbearable. So we threw the entire document at the agent and told it to install and print. It asked for permission to operate the computer screen directly. Permission granted, it went to the printer site and downloaded what it needed. It needed the model name, and checking the back of the printer felt like too much work, so we took a photo and threw that in too: model unknown, printer looks like this, figure it out. It did.
The whole thing took about fifteen minutes, during which we were doing other work. Throwing meaningless, annoying chores at an agent and watching them dissolve is genuinely satisfying.
The end of the software productivity business is visible
And right there a question appears: then what is left for us to build? There does not seem to be much opportunity left in AI software productivity. Where a niche remains, the market is too small; grow it slightly and the frontier labs swallow it within weeks or months.
The world changed too fast in two years. It feels like we need to jump one level up to a crazier idea, and the problem is that we may still be defining problems with pre-AI assumptions. Business is discovering and solving problems people have, and the way we discover and define them may itself be outdated.
A company that lets you order sunlight
In the middle of that conversation we watched a video: someone presses a button on a smartphone, and sunlight shines down in the middle of the night. For night rescues or moments when you urgently need bright light, an 18-meter mirror on a satellite reflects sunlight down, no giant crane, relatively clean. The company is Reflect Orbital; it raised a 20 million dollar Series A in 2025 and plans to launch two test satellites in 2026.
Whether it works is not the point. Saying you will build an AI agent has become boring, and investors probably place little hope in it. This idea, whose economics we cannot even begin to judge, hooks instantly. You want to hear more. How do ideas like this get born?
Practicing breaking assumptions
From the same angle came another idea: a service that makes you feel as if you had a happy memory. Not drugs, not a chip in the brain. Considering that hypnotherapy is actually used to treat post-traumatic stress, it is not entirely absurd.
If someone who never lived in the US received the memory of attending school there as a child, would it help them understand that culture? If someone short on confidence received a memory of boldness, would their personality change? Injecting bad memories, or applying this to children, raises obvious ethical problems. But if reality is a composition of past memories, and those memories are already less than fully accurate, the question itself is worth asking.
What matters is not whether this particular idea is feasible, but the practice of pushing thought in this direction. Treat AI not as a technology but as a mindset for breaking assumptions, and the questions change.
Join Bloom
Bloom builds offline rooms where people and technology meet. We run them in Seoul, and now beyond it.
- 💬 Discord community, where events are announced first
- ▶️ YouTube, full talks from past events
- 📸 Instagram, photos from our events
- 🎟️ Upcoming events
