I built an AI “chief of staff” for myself — people kept asking to use it, so I’m opening the beta

I’ve been building Struxy, and I’m finally at the point where I’m comfortable letting strangers use it.

The funny part is that Struxy wasn’t originally supposed to be a startup.

I was working on a completely different product and constantly juggling conversations, people I needed to follow up with, random ideas, promises I’d made, things I needed to remember, etc.

Task managers weren’t really solving it because the problem happened before the task manager.

I would have a thought like:

“Need to send Sarah the deck Friday, João mentioned that startup programme again, and I should probably book Lisbon before prices go insane.”

…and then I was expected to manually decide what was a task, what was a person, what belonged to what project, set dates, organise it, etc.

Half the time I just wouldn’t.

So I started building something where I could basically dump the thought as-is and let the software do the organisational work.

That became Struxy.

What it does right now

You type or speak naturally.

Struxy keeps the original thought, interprets it, and tries to pull out the useful structure underneath it — tasks, reminders, people, context, follow-ups, etc.

The goal is eventually a proper AI chief of staff for your personal life: something that understands your context well enough that you can actually delegate things to it.

But I’m deliberately not pretending we’re there yet.

The current beta is focused on getting the foundation right:

Can you talk naturally → can Struxy understand what mattered → can it remember it → can it make that information useful later?

That sounds simple, but I’ve discovered it gets surprisingly difficult once people stop writing clean AI prompts and start talking like actual humans.

I want people to break it

The beta is now live at:

https://struxy.app

I’m particularly interested in people trying it with genuinely messy inputs.

Rambling. Typos. Multiple unrelated thoughts in one sentence. Mention someone you talked about three days ago. Change your mind halfway through.

Basically: don’t adapt the way you think to the software.

If Struxy only works when people learn how to talk to Struxy, I’ve built the wrong thing.

One slightly unusual thing: I’m charging for the beta

I went back and forth on this.

In the end I made Founding Beta access pay what you want — €1 minimum, €5 suggested.

It’s a one-time payment, not a subscription.

I’m not expecting beta revenue to fund the company. The €1 is mostly a tiny commitment threshold and an experiment in whether people will actually pay anything for the problem, rather than collecting another giant waitlist of people who clicked a button once.

People who join now get permanent Founding Beta Member status and access to the private feedback group on Slack.

What I’d love feedback on

Not really the landing page.

I’d much rather know:

• Did you instinctively understand what to tell Struxy?

• Did it correctly understand messy real-world thoughts?

• Did anything it created actually make your life easier?

• Where did you suddenly think, “why the hell do I have to click this?”

• What did you expect Struxy to do next that it couldn’t?

This is the first version where I feel the core loop is reliable enough that feedback from real use is more valuable than me continuing to build in isolation.

Would love to hear what breaks.

— Kush