How Hatch started
While running Indie Hackers a year ago (in May 2025), we noticed two things that convinced us to start Hatch (the most productive AI assistant):
AI was becoming the most important shift in software, but
the interfaces around it still felt
shittyreally basic.
That felt wrong to us.
Serious work happens across projects, notes, files, drafts, calendars, emails, decisions, and weeks or months of accumulated context, etc., not in one-off chat conversations. So if AI was gonna become part of daily work, it needed a better environment around it.
We framed the first version of Hatch as an AI-powered second brain for founders, aimed at memory, synthesis, and action. But the vision quickly got bigger and we quickly started throwing around the term AIOS ("AI operating system").
One early design idea stuck: that the window, instead of the chat, should be the primitive. This way, chats and docs (etc) could standalone and freely reference each other.
Ultimately, this would mean that Hatch could become more than a "better note app" or a "better chatbot." It could become a place where chats, documents, files, memory, and connected tools all lived together, and where the assistant could actually help carry work forward.
That's still the ambition!
The evolution from “AI second brain” toward a broader AI operating system is a pretty significant shift in scope. Curious how Hatch has progressed since this post — has the product ended up validating that broader vision, or has usage pulled you in a different direction?
The "window as the primitive instead of chat" framing is a genuinely different take, most AI assistant products just wrap a chatbot in nicer UI and call it a second brain. I'm building in the adjacent space with FounderFlow (mentioning since it's relevant), more narrowly focused on daily triage than a full AIOS, so I've been wrestling with a smaller version of the same problem: how much should live inside the assistant's memory vs. just get referenced from where it already lives. Have you landed on a rule for when something graduates from "referenced context" to something Hatch actually stores and reasons over?
Hi, I'm very new to all this and just about to launch a SaaS app in the accountability /PT space, so sort of related but not directly. That research has got me more interested in AI Assistants and where that is heading, particularly with the arrival of Tomo, but also with the precedent / vision of Kubrick's HAL. Bit of a side question to your product, but how close do you think you could get Hatch to operate as HAL ?