My bookmarks folder is where good ideas go to die

A while back I sat down to clean out my Pocket queue. Three hundred and something saved things, most of which I had no memory of saving. Buried in there was a Reddit thread I’d bookmarked maybe eight months earlier: a group of people in a niche trade forum, complaining in real detail that no software did the one thing they all needed. I remember the exact thought I had at the time. There’s a business in that. Then I saved it, closed the tab, and never opened it again.

By the time I found it, someone else had built the thing. Good for them. Losing it wasn’t an idea problem. I’d had the idea, and it was a good one. The problem was that I had nowhere to put it except a pile, and a pile never hands anything back.

If you save things because “there might be something here,” you already know the pile I’m talking about. The teardown of some small SaaS that’s quietly doing fine. The essay about a workflow you could probably automate in an afternoon. You read it, you nodded, you filed it away. On Tuesday it felt sharp. By Thursday it was gone.

I got sick of that, so over a couple of weekends I built something to fix it for myself. It worked well enough that I cleaned it up and put it online for twelve dollars.

It’s a decision problem, not an idea problem

The mistake I kept making was thinking I needed more inputs. I didn’t. I had hundreds. What I was missing was a fast, honest way to look at any single one of them and answer the only question that actually matters: is there a business in here, or am I about to spend a weekend finding out there isn’t?

The usual move is to paste the article into ChatGPT and ask for a summary. I did plenty of that. The trouble is a summary is just the thing you already read, made shorter. It gives you more words back when what you wanted was a verdict. And that little gap, the one between “huh, interesting” and “okay, actually do this,” is exactly where my good ideas kept dying.

What it actually is

Idea Harvester isn’t an app, and there’s nothing to install. It’s a skill file. You paste it once into whatever AI assistant you already use, Claude or ChatGPT or Gemini, it doesn’t matter which. From then on you hand it a link, a PDF, or some pasted text, and instead of politely describing the thing back to you, it runs it through an actual assessment.

It reads what you’re asking for and answers in kind. Say “review this” with a link and you get a 30-second worth-it-or-skip call, nothing heavier. Say “harvest this” over a full article and you get the breakdown: the core thesis, who would genuinely pay for it, whether one person could build it, roughly how long to a first dollar, and who’s already competing. Ask “what can I sell here” over a report and it comes back with specific revenue paths, ranked by how fast each one would get you to cash.

Underneath all three modes, it puts every input through the same five questions before anything reaches you:

1. Who specifically needs this? A person, not “small businesses.”
2. Could one person realistically build it?
3. How long until the first dollar?
4. What already competes, and where are the gaps?
5. Is there an angle nobody’s working yet?

Then one last gate runs at the end: is this real, is it tested, would you actually put your name on it? If the answer is no, it doesn’t dress the idea up and hand it over anyway. That gate is the part I care about most. It’s the difference between a tool that tells you every idea is brilliant and one that will occasionally look at your pitch and suggest you go do literally anything else.

What comes back

Not a wall of text. A position. In practice it reads something like:

“Build the niche calculator. Single page, live in about a week, and here’s the angle the current players keep missing.”

Or the other way: “Skip the directory idea. Too much ongoing curation for too little early money. But that automation workflow sitting underneath it, the boring thing you already do by hand every week? That’s the actual product.”

Every answer lands on one of three words. Do this, look into, or skip. No “it depends.” You can act on it or you can argue with it, and honestly both of those beat the pile.

I’ll be straight about the limits. It gives you structured judgment and estimates, not guarantees. It won’t promise you a revenue number. What it will do is tell you where the opportunity probably sits, how hard it looks from where you’re standing, and whether it’s worth your next free Saturday. That’s the whole job.

Who it’s for

If you’ve got a saving habit and no system to process it, this is pointed at you. Solo founders. Indie hackers who don’t want to lose two hours researching every “is there a business in this” itch. Creators who clip everything and action none of it. You don’t need to be technical either. If you can paste text into an AI assistant, you can run it. No signup, no server, no subscription. You own the file outright.

Why twelve dollars and not free

There are free prompts floating around that promise roughly this. I’ve used them. Most hand you a confident-sounding summary with no framework underneath it. This one got built and run against real articles before it ever went up for sale: three detection modes, a proper assessment matrix, research folded in, and that quality gate on the end. It’s made for one specific buyer instead of trying to be everything for everyone.

Twelve dollars, paid once. A read-it-later app costs more than that every month and never makes a single call for you. This pays for itself the first time it talks you out of building something nobody wanted. You get the skill file, a full worked example so you know exactly what you’re buying, a quick-start that takes about two minutes, and every v1.x update. If it’s not for you, there’s a full refund inside 30 days.

Go dig something out of the pile

Your backlog is not going to process itself. Next time you catch yourself saving something because “there might be something here,” that’s the moment this was built for.

Idea Harvester — $12, one-time: atomic24.gumroad.com/l/pxcegm

Paste it in, point it at the most interesting thing you’ve saved this month, and see what it says back. Worst case, you clear one item off the graveyard. Best case, you find the one that was worth saving in the first place.