Two straight Product Hunt #1s pitch agent memory as infrastructure

Unabyss took the number one spot on Product Hunt's July 17 leaderboard with 598 votes, pitched as shared memory across every app and LLM a user touches. The next day, ZooData took number one with 606 votes, calling itself the data layer for AI agents. Neither one is an app you'd open and use. Both are the thing other apps plug into.

That's the detail worth sitting with. Two days running, the most-voted product on Product Hunt wasn't a feature, it was plumbing. And it's the same plumbing: a place for agent context to live outside any single agent.

The stakes are for anyone building an agent product right now. If the memory layer is separate from the agent, users can swap the agent and keep their history. That makes the agent itself replaceable, and it makes whoever holds the memory the one everyone else has to build around. It's the Segment playbook from the SaaS era, one layer up: Segment won by sitting under the apps, collecting the data every app needed, and becoming too embedded to rip out.

Other Product Hunt entries from the same window point the same direction. Albato AI shipped agent-driven workflows across more than 1,000 apps, rebuilding the Zapier model around intent instead of triggers. That kind of orchestration only works if there's a shared memory underneath it feeding context to whichever agent is running.

What I don't know yet: whether Unabyss or ZooData actually get adopted as infrastructure by other builders, or whether this is two products racing for a category that doesn't have a real winner yet. Product Hunt votes measure launch-day attention, not retention. Worth watching whether either one shows up as a dependency in other tools' changelogs over the next few months.