I spent a decade working on software designed to extract.
At Twitter, I was on ads. My job was to turn attention into revenue — tune the auction, stack the ad load, find one more impression per session the retention curve could survive. Before that, freemium iOS on the other side of the same coin: make the user want it enough to pay.
What changed isn't guilt. Almost everyone I worked with was trying to ship something good inside a structure that rewards different behaviour. The structure is the problem. I saw enough of the full machine — from acquisition-cost spreadsheets to A/B tests on whether shame converts better than curiosity — and it stopped feeling like a puzzle I wanted to solve.
I didn't stop. Day job is applied AI at Mercor — the work is still interesting, the stakes are still real. What I stopped buying is the idea that the tools small teams use have to come from places optimised against them. CosmicGlue is how I give those tools back: the uptime monitor, the go-links, the widgets, the agent runtime — priced for founders without a procurement department, built by someone who remembers what $50-a-seat does to a startup's budget.
It's a studio. Apps, games, a few tools. Some paid, most cheap, some free. The through-line is simple: nothing on it is designed to extract from you. No dark-pattern cancel flows. No upsell tier you didn't know existed. No data sold, because there's no data to sell.
The prices are small. Not charity — the price is a filter, not a business model. Enough to cover the servers and prove I'm not going to disappear. Not enough for anyone with a spreadsheet to get excited about. Nobody is pitching any of this to a VC. That's the point.
A few things it isn't.
Not a protest. I'm not angry. I'm just building differently because I can.
Not a redemption arc. I did what I did. Building small, honest software now doesn't erase that — it's just the work I find interesting, on the terms I want.
Not a minimalist performance. I'm not trying to prove anything by running everything in a terminal. What I won't pay for is tools that penalize growth — per-seat pricing that charges full freight for the teammate who needs a link once, same as the one who lives in the product. Taste is the constraint, not the aesthetic.
What I'm building is boring on purpose. Uptime monitoring. Team go-links. Comments and forms for static sites. A secure runtime for AI agents. Analytics without the creepy parts. Idle games without a monetization team. Most of these exist already — they're $50 a seat, locked behind a sales call, quietly adversarial. I want versions priced like what they are, that don't assume you're a growth surface. Small studios shouldn't get price-gouged to run basic tools.
That's the whole thing. Small tools, priced like what they are, built to last. No story arc. No grand plan. Just software that should have existed all along.
If that's useful to you, great. If it isn't, also great. There's nothing to sell you.
— Tim