Working in tech means drowning in nuggets. Thought-leader feeds, marketing that sneaks past your filters, a colleague dropping a link in Slack. The real question was never “follow this or ignore it.” It’s what you do with a nugget when you’ve got a coding agent sitting open in another tab, ready to just build the thing.
Turns out you can just copy the idea into a running system. Not read about it — run it.
I downloaded someone’s “agent in a markdown file” for automated outreach. By the end of the day I had a working system sending tailored emails. Within hours of that, I hit a problem I didn’t know I had: how many of these should we actually be sending? No purchasing decision. No HR. No committee, no business case. Just tokens — and a novel scale problem I’d never have met if the system hadn’t existed yet to surface it.
Second one: a voice recorder plus Zapier feeding a Google Drive folder, an agent that sweeps the files, pushes into the CRM, drops tasks in Slack, and keeps a real-time ledger finance can actually see. Stitched together from about six different people’s posts. I’m a scientist who still doesn’t think of himself as a salesperson, and I built a finance ledger before lunch.
I’m not special here. The judgment is in the markdown, the error handling, the routing, the voice — none of which felt hard to write. What’s actually scarce isn’t the building anymore. It’s noticing the problem the build surfaces the moment it’s alive, and knowing when to stop copying and start thinking.
The cost that never shows up in a “look what I built” post is the cost of exiting wonder — the gap between “it works” and “I’ve hit the real question.” Get through that gap fast, or you’re just collecting demos.