Assistly

Hey IH 👋 first post here, been lurking for a while.

Every AI assistant shows up after the call ends, summarizing a conversation you already survived alone. Assistly shows up during it: live context and guidance on your screen, right when the conversation needs you at your best.

I’m a software engineer at a fintech by day, and for the past 3 months I’ve been building Assistly at night. It’s a real-time meeting assistant for macOS and Windows. No bot joins your call. It works from your computer’s own audio, in a private overlay that never shows in screen shares or recordings.

I am about to launch on Product Hunt and figured I’d share the three biggest lessons instead of just dropping a link.

1. I killed my most viral positioning on purpose

The obvious growth hack in this space is the “undetectable interview copilot” angle. Competitors lean into it hard and it clearly drives clicks.

I went the other way. Assistly is consent forward and privacy first: audio processed from your own machine, no bot injected into calls, and the overlay stays out of screen shares because those are your private notes, not because you’re hiding from anyone.

Short term, I probably gave up some easy traffic. Long term, I get to build something I can put my name on, sell to teams instead of hiding from them, and not wake up to a platform ban or a press piece with my face in it.

Lesson: if your positioning only works when nobody looks closely, it’s a liability, not a moat.

2. The hardest problem wasn’t the AI

Everyone thinks real-time assistants are all about the LLM. They’re not. The brutal part is making the whole thing feel instant on a live call. A couple of seconds too late and you might as well not exist.

I rebuilt the pipeline more times than I want to admit. At one point I spent days chasing a throttling issue that turned out to be a one line fix. One line. Days of my life.

Lesson: in real-time products, your architecture IS your product. Users don’t see “good AI.” They see “did it show up in time or not.”

3. The bet I’m making next: team spaces

This is the part I’d genuinely love feedback on.

Meetings are where team knowledge actually lives, but today it evaporates the moment the call ends, or gets buried in notes nobody reads. So the big bet is shared spaces: a space per team, project, or client, where every call adds to a memory everyone in that space can draw on.

That means when you join a project, your first call already comes with the context of every call before it. When a client asks “didn’t we agree on this last month?” you know, even if you weren’t in that meeting.

MCP (Model Context Protocol) is what makes the spaces powerful, in both directions:

• Assistly connects out to your existing tools (project trackers, docs, calendars, CRMs) for grounded context mid-conversation. When someone asks “where are we on that ticket?” what shows up on your screen is real, not hallucinated.

• Other AIs connect in to your team’s spaces. Any AI tool your team uses can query the same memory over MCP, so your coding agent knows what was decided on Tuesday’s call without anyone pasting notes around.

Nobody in this market is treating meetings as the shared memory backbone for teams yet, and I keep going back and forth on whether it’s the wedge or a distraction.

So, honest question for the folks who’ve shipped further than me: if you were me, would you double down on team spaces and shared memory as the differentiator, or is “fast, private, no bot” enough of a story on its own?

Happy to answer questions about the journey, the costs, or juggling this with a full-time job.