The Observatory

If you are raising money to build an AI-native company, the fundamental negotiation hasn't changed — but the story you're telling investors has, and that gap will cost you if you walk in unprepared.

If your first hundred users don't feel personally obsessed over, you won't know what to automate — and right now, agents let you do that obsessing at a cost that has no precedent.

When retention is automated, the ethical weight of the habit loop lands entirely on the founder's intentions — not the product's friction. That's the design problem you actually have.

Ben Horowitz's lessons about leading through chaos hit differently when your "staff" can spin up a hundred threads simultaneously — and fail across all of them at once.

If you're still tuning agent behavior by rewriting prompts after something breaks, you're goal-fixing when you should be system-building. The difference compounds, and not in your favor.

If your agents are doing the work, the network effects that make or break the product are being seeded right now — and most founders are too busy watching outputs to notice they're engineering a cold start into their architecture.

If agents can execute faster than your team can reflect, the question of what to build — and for whom — becomes the only question that actually matters. Get that wrong and you're just failing at speed.

If your agents are shipping output nobody measures, you're just burning compute on vanity. The build–measure–learn cycle Ries built for human teams gets sharper and more dangerous when the loop runs at machine speed.

Kahneman's core finding is that fast, confident judgment is the default — and it's wrong in predictable ways. If your agents are making dozens of decisions a minute, you've just industrialized the failure modes he spent a career cataloging.

If agents can execute distribution experiments in hours instead of weeks, the bottleneck has shifted entirely to founder judgment about which channels are worth running at all.

If you're using customer conversations to validate your AI product, the conversations are probably lying to you, and you're probably helping them do it.

If your competitive moat can be drained by a competitor who just prompts differently, you never had a moat. Thiel's core argument about secrets and monopoly is more urgent now, not less.