5 Comments
User's avatar
Jordan Cutler's avatar

Wow! This is so practical and I’m excited to try all the use cases I have in mind 🤩

Fran Soto's avatar

Thanks, Jordan, looking forward to reading what you build!

I had this same "wow" thought when I saw it working and how the steps were relatively simple. No need for a phd to start trying this in our day-to-day work :)

Patricia Juarez @ AWS's avatar

@Fran Soto great work explaining the workflow. It is very clear.

This is something I have been doing for nearly a year. Maybe we could have that pending 1:1 and brainstorm ☺️

I was very surprised about Clawdbot success, as you said, it just makes easy setting things up and cron jobs or heartbeats checking for updates.

I have cron jobs running for a couple of months.

But still, I’m the bottleneck and human. There is a time you need to switch off. Also if you are too ahead of the curve it can be against you. The team and company needs to adapt. I was more concerned about that than delivering more work done.

Fran Soto's avatar

I totally experienced the same!

I was doing a bit of research for another post, and we are definitely the bottleneck as humans. A long time ago, the computer was the bottleneck (people making a queue to submit their stuff to a mainframe / or any limited-usage computer). I think that's where we realized we needed strong verifications before submitting code.

Now it's the opposite, I'd rather have AI do a lot of work, and I review at the end. If I don't like what I see, instead of fixing it, I could just let it generate it again from scratch.

Compute is no longer the bottleneck. Our brains are

I sent you a DM in slack so we have that 1:1 :)

Pawel Jozefiak's avatar

love the practical breakdown here. The 'works while you sleep' framing is spot on - that's exactly what I'm doing but with a different angle. My agent isn't working *at* Amazon, it's building stuff *for* me while I sleep. I've been tracking the actual economics: infrastructure costs vs revenue generated vs time saved. What surprised me is that the time savings is actually the smallest ROI - the real value is having continuous forward momentum on projects that would otherwise stall.

The trick is building in monetization from day one, not as an afterthought. Most people build cool agents that do stuff but never connect them to actual value capture.

How I'm approaching it: https://thoughts.jock.pl/p/my-ai-agent-works-night-shifts-builds