BUILD LOG Ailsa Zircher BUILD LOG Ailsa Zircher

I spent half a day trying to set up a way to stay updated on SaaS

Turns out it's faster to build your own tool than to find, trial, and configure what's already out there — and the tool you build ends up asking better questions than the ones you'd have bought.

Turns out it's faster to build your own tool than to find, trial, and configure what's already out there — and the tool you build ends up asking better questions than the ones you'd have bought.

Doing it the old way versus the new way

I spent half a day trying to set up a way to stay updated on SaaS.

It took surprisingly long just to figure out what tools were even available — then trial them, set them up, and suddenly I had a ton of data coming at me. None of it filtered for my angle.

Then I built my own agent. Took 30 minutes.

It has a company list I control, sorted into cohorts — legacy enterprise, AI-native SaaS, vertical AI, infrastructure and dev platforms, craft-led SaaS — and a persona that knows my role. I can point it at a single company, run a full cohort scan, or ask for a two-week "what's new" pulse, and get something back that's actually filtered for what I care about: what shipped, what's real versus positioning, who's winning the AI transition.

A few of the signals a recent cohort scan turned up:

  • Plan Mode is becoming the standard trust interface for agentic UX.

  • The invisible AI interface is arriving faster than design systems are ready.

  • Outcome-based pricing is the new battleground for enterprise.

That's the kind of read I used to have to piece together myself from changelogs and Twitter threads.

Anyone else surprised by how far this has come? It's easier to whip up your own tools than wrestle with what's already out there — and once you've built one, it starts asking sharper questions than the tools you'd have bought.

Read More
BUILD LOG Ailsa Zircher BUILD LOG Ailsa Zircher

I gave the same prompt to 3 AI app builders

Lovable, Figma Make, and Google Stitch — same one shot prompt, very different distance to done.

I gave the same one shot prompt to three AI app builders.

Not multiple iterations. Not refinement. Just one detailed prompt for a personal workspace app. The UX they produced was actually quite similar.

But the difference was in how far each got me:

  • Lovable — felt like a usable product. Coherent UX, working flows, strong responsiveness.

  • Figma Make — solid prototype. Same structure, but would need more work to feel polished.

  • Google Stitch — a design starting point. Good exploration, but missing pieces and consistency.

View Results

The Takeaway:

it's not just about what gets generated, it's about how much effort is still required after the first prompt. Same input, very different distance to "done."

Read More