"I'm tired of boring SaaS and boring stories."
A friend said that. I agree. This is the first time a small group of people can build what used to require a whole industry. Someone vibecoded a consumer app that turns pet photos into Renaissance portraits. It reportedly makes $100,000 a month. A funny idea, shipped in a weekend, and people pay for it because it makes them smile. That's not boring. The boring part is when the entire conversation about these tools stays stuck on chatbots and wrapper apps that nobody loves building or using.
If you're tinkering, scratching an itch, keep going. Most of the interesting things I've seen started as a weekend project. But if you're trying to build something that lasts, build toward the world that's coming, not the one we lived in two years ago.
The bets
I'm betting on interfaces that disappear. The screen should be where you go when you want to, not where you're forced to be. Most of the time we spend on screens isn't worth the attention. Voice is the input layer that changes this. Text only gets you so far. Voice carries tone, hesitation, emphasis: the parts of a sentence that change its meaning without changing its words. The interfaces that feel natural in three years won't be the ones you look at. They'll be the ones you talk to.
I'm betting on systems that earn the right to remember you. The product that remembers you well enough to act on your behalf, while giving you a real way to leave, is the one that won't need to trap you.
I'm betting on one person plus agents as the default team. The minimum viable unit for building something real just shrank to one. That changes what gets built, who builds it, and how fast the cycle turns. As screens retreat to only where engagement is earned, the products that remain are starting to look more like games. Orchestrating agents, adjusting in real time, verifying outputs: it already feels closer to a real-time strategy game than a dashboard. The tools that make this feel like play will be the ones people actually learn.
The distance between someone who actually uses these systems and someone who reads about them is already enormous, and it's going to get bigger. Information asymmetry is the real advantage right now, but it's not about knowing the news. It's knowing what works in practice because you've tried it. It's being able to assemble context, notice what's missing, and adapt your setup as the models and tools change under you. You get that by building, not by watching.
When AI can generate anything, the scarce thing isn't content. It's who made it, who checked it, who's accountable. I'm betting on verification as the hard problem: evals, audit trails, provenance. The boring infrastructure that makes it possible to let systems act on your behalf without losing sleep. Everybody will need it.
I'm betting on physical presence as a product. For builders who now ship entire products alone, co-presence is a real win. Hacker houses, curated meetups, Cafe Cursor-style gatherings where people show up with laptops and swap workflows. Not "community" in the vague Discord-server sense. Compressed pockets of people who chose to show up, organized around increasingly narrow shared obsessions. Attendance is proof of commitment no algorithm can fake.
I'm betting on physical agents. Robots, embodied systems, the software and safety layers that make them actually work. The gap between a robot that demos well and one you'd trust in your home is the same gap as between a chatbot and a system you'd run your company on. That gap is closing.
I'm skeptical of thin productivity wrappers that labs will absorb within a cycle or two. Skeptical of "AI for X" pitches where X has no defensible data or regulatory moat. Skeptical of products that assume people want more screen time. Skeptical of one-size-fits-all assistant visions; I think context fragmentation wins.
Most predictions about where this goes will be wrong. Mine included. But these aren't predictions. They're where I'm putting my time when I can't be certain. They've held through every update I've made to my thinking in the past year. New information keeps confirming them. That's enough to act on.