An evening, not a quarter of roadmap

A pattern I keep running into: I’ll want some small piece of GTM machinery: a scraper that watches a set of subreddits, an enrichment step that stitches two tools that were never meant to talk, a script that turns a messy export into a clean segment. A year ago, each of those was a ticket. You’d write it up, argue for it against everything else in the backlog, wait a sprint or two, and maybe get a version that did 70% of what you meant. Now most of them are an evening. Not because I became an engineer overnight. Because the cost of building the unglamorous glue between systems fell through the floor.

That’s the quiet story under all the AI hype, and it matters more for operators than any single flashy feature: the tax on building your own tools has collapsed.

The glue was always the bottleneck

Growth work is mostly integration work. The clever strategy is the easy part; the grind is getting data out of one system, reshaped, and into another. This connective tissue never made a roadmap because it was never anyone’s headline feature. So it stayed manual, or it stayed undone.

LLMs plus a decent tool layer (the MCP stack, a scripting environment, a model that can hold the messy middle in its head) turn that connective work from a project into a conversation. The bottleneck was never the idea. It was the hours of finicky plumbing between having the idea and having the thing. Those hours are what got cheap.

The competitive gap is shifting from who has the best playbook to who can build the tools their playbook needs faster than the playbook changes.

What changes when tools are cheap

• You stop waiting. The distance between “I wish I could see X” and “here’s X” is now hours, so you actually check the things you’d previously have shrugged off.
• You build for one. It never made sense to put engineering time into a tool only you’d use once a week. Now it does, because it costs an evening.
• The moat becomes the operator. When everyone can build, the edge is taste: knowing which tool is worth building, what to measure, where the leverage is.

The trap

The obvious failure mode is building for the sake of it: a graveyard of half-finished scripts because everything is now buildable and nothing is prioritized. Cheap to build is not the same as worth building. The discipline that used to be enforced by scarcity now has to come from you. When the constraint disappears, judgment has to replace it, or you drown in your own tooling.

I don’t think “operator who can build” is a niche for much longer. It’s just what the job becomes.