The second time I built it, I didn't realize it was the second time until I was about three hours in and found the first folder. Same name, different date stamp. tool-v1 sitting right there in my projects directory like a passive-aggressive Post-it note from past me. I just stared at it for a second and then went and reheated my coffee without saying anything to anyone.

The idea was a lightweight tool that would take a wall of meeting notes — the kind your team dumps into Slack at 5 p.m. on a Friday — and spit out a clean summary with action items flagged. Useful, right? I thought so. I thought so twice, apparently, eighteen months apart, with no memory of the first attempt until the file system reminded me.

I've got at least one tool that actually made it out the door, so it's not like I never finish anything. But this one had a specific gift for getting almost all the way there and then finding a new reason to stop. Twice. Same tool, same almost, different year.

Why I Built a Meeting Summarizer Nobody Asked Me to Build

The first version came out of a genuinely bad stretch of work meetings. I was sitting on a call where someone was reading bullet points out loud that were already on the shared doc, and I opened a new project folder just to feel like I was doing something productive. Started sketching a prompt chain that would ingest raw notes and return structured output. It worked okay in testing. Not great, okay.

Then I got busy. DoorDash picked up, the kids had something going on, I had a bug at work that ate two evenings. The tool went into the folder and the folder went quiet. I didn't close the project so much as just stop opening it.

The second version started because John mentioned he hated recapping his week for his own records. He does this thing where he keeps a running log of what he's been up to — mostly so he can cross-reference it against whatever theory he's currently working on, which I find both funny and kind of admirable. He said he wished something could just do it for him.

"You build all this stuff," he said, "and none of it does anything I actually need."

He wasn't wrong. I told him I'd look into it. Opened a new folder. Named it tool-v1. Did not check if that name was already taken.

Two Builds, One Identical Wall, and a Folder I Should Have Deleted

Both versions hit the same problem. Not a technical problem, exactly — more of a shape problem. The tool worked fine on clean, structured notes. Give it a decent Slack thread or a formatted doc and it would produce something genuinely useful. But real meeting notes are not clean. They're half-sentences, initials nobody explains, action items buried inside jokes, and at least one person who types in all lowercase with no punctuation as a personality trait.

The first time I hit this, I spent a week trying to pre-process the input. Wrote a bunch of cleaning logic. It helped some. Then I tried to figure out how to let users paste messy text and have the tool handle it gracefully, and that's where version one went dark. I didn't have a clean answer and I didn't want to ship something that looked dumb on real input.

The second time — and I want to be clear that I had completely forgotten about version one at this point — I hit the exact same wall, at approximately the same hour on a Thursday night, and made the exact same call. I need to handle messy input better before I put this in front of anyone.

That's when I went to get coffee and found the old folder. I opened it. Read my own notes from eighteen months ago. One of them said: "messy input is the whole problem, figure this out before shipping." I had written myself a warning and then filed it where I would never see it until I had already ignored it a second time.

There's a version of this story where I find the old folder, merge the two codebases, and push something out. That is not what happened. What happened is I read through version one for about twenty minutes, realized I'd actually gotten further the first time on the cleaning logic, and then closed both folders and watched Netflix. I was tired. It was late. The kids had school in the morning.

I did come back to it — once more, a few weeks later. Got it working well enough that I would have used it myself. That's usually the threshold I aim for, and I cleared it. Then I started thinking about what it would take to make it something I could actually put out there, and I thought about what happens when you get close to a second user and still can't quite close it, and I put the laptop down.

What Two Attempts at the Same Thing Actually Ran Me

First build: probably twelve hours spread across two weeks. API costs in testing were small — a few dollars, maybe less, I didn't track it carefully because I told myself I'd recoup it when I shipped. I did not ship. Those dollars are just gone, which is fine, they were small, but I'm naming them because that's the deal here.

Second build: another eight or nine hours before I found the old folder, then another three or four after I decided to keep going anyway. Call it twelve hours again. More API spend, slightly more this time because I was testing with longer inputs. Still small. Still unrecouped.

The real cost is harder to put a number on, which I realize sounds like I'm about to get philosophical, so I'll just say it plainly: I spent a combined evening's worth of time on a thing I had already spent a combined evening's worth of time on, and I have nothing to show for either one except a folder structure that now has two subfolders and a mild sense of embarrassment.

I also told John I was working on something for him. That was months ago. He has not asked about it, which is either very kind or a sign that he already forgot, and either way I'm not bringing it up first.

What I Kept, What I Closed, and the Notepad File I Still Have Open

The cleaning logic from version one is actually sitting in a utilities file I've reused on other things. So it's not completely wasted — it just ended up being a component instead of a product, which was not the plan but is at least something.

The summarizer itself is still in the folder. I haven't deleted it. I also haven't opened it in about six weeks, which probably tells you more than I want it to. There's a Notepad file from 2023 — one of the ones I use as a diary — where I wrote down the original idea, and it's still in the same tab group as my work stuff. I see it every morning. I have not acted on it.

What I keep thinking about, in the background, is that the messy input problem is actually interesting. Not interesting enough that I've solved it, but interesting enough that I keep not deleting the project. There's a difference between a thing you've abandoned and a thing you're letting rest, and I'm genuinely not sure which one this is.

The honest version is: I probably built this thing to feel productive during a bad patch at work, and the second time I built it to feel useful to John, and neither of those is a great reason to ship something. There's a pattern in the almosts where the motivation is slightly off from the start and the project just knows it somehow. This one knew.

The folder is still there. Both of them, technically — I never merged them into one. Some nights I think I'll open it up and finally sort out the input handling. Other nights I open DoorDash instead, because at least that one pays out in a way I can see on the same day I did the work. The tool might still ship. It might turn into a third folder with a third date stamp. Honestly, at this point, either outcome feels about right.

Note: Everyday Hustle Jeff is the written side of a digital series. These accounts are composite and dramatized. Any figures are what they were on the day, for one person, and are not a projection, a track record, or advice.