Stop taking screenshots. Just point.

Ten spots, ten screenshots?

When I talk to AI about a screen, I take a screenshot and paste it in.

It works. But images fill the context window fast, and they burn tokens.

One day I had about 10 things to say about one screen.

10 spots. 10 screenshots?

That felt silly.

So I built a pointer

AI can't see my screen. "Make the button smaller" leaves it guessing which button.

So I made Momo PinPoint, a small layer on top of any web page.

Here is how it works:

• 📍Pin. I click to drop a numbered pin, or drag a box.
• ✏️Note. Then I type a note.
• 📋Copy for AI. It turns every note into text: which element, which section, what I want.

Instead of "make the button smaller", the note says: pin #1, this button, make it smaller.

Did someone already make this? Maybe. I never asked AI first.

I made a simple version and didn't polish it. I used it as it was. Done.

"I made this, it makes work easy"

Our office has a Show and Tell, a meeting where we show things we made. I showed PinPoint there.

Nobody said much. Only our CEO and CTO did.

They were glad someone had shared it, and glad someone had used AI to build an internal tool.

They said it could grow in a few directions.

Afterwards I shared the GitHub link in Slack and wrote: "It still can't do much yet."

A coworker replied that it looked nice and interesting.

Four days later, our CEO had put the idea straight into our internal ticket tool (think Jira, but ours).

He said the team could use it with a client the next day.

His version lets you draw boxes, arrows and pen marks on a client's site, write a bug report, and send it straight to our ticket tool as a ticket.

I only wanted to stop taking screenshots.

Now I use it for everything

• 🔎On reference sites. I drag a box over a section I like. The note goes to AI with that section attached. AI can read its code and UI, learn from it, and help me adapt it for my own project.
• 📄On my own pages. I comment on the whole page and send it all to AI in one go.

No screenshot. No paste. No explaining the screenshot.

Dragging images into a chat takes time, and the picture often misses part of the page.

Half the time I can't even tell if I dropped the right one.

Add-ons: plug it into your team's tools

PinPoint is small on purpose. But it doesn't have to stay small.

Once the comments are shared, they can go straight to where your team already works:

• 🎫A Jira ticket. Send the pins and notes straight to Jira, so your team gets one ticket they can pick up instead of a pile of screenshots.
• 🤖An AI agent. Hand the notes to an agent and let it fix the page itself.
• ✨Your own prompt. Add a prompt on top, like "fix these, then open a pull request", to take it further.

Today it stops at Copy for AI. What it connects to next is up to your organization.