Who is actually launching these coins?

Most launches are manufactured, not made.

The biggest creator wallets are factories: hundreds of launches per day, the same ticker relaunched over and over, almost none completing the curve. If you've ever felt like the trenches are a conveyor belt — here is the conveyor belt.

"Dev sold" isn't the scandal. The repeat business is.

Selling the initial dev allocation early is standard practice — many buyers read a clean early dev exit as a good sign (no overhang left to dump), and some devs buy back in later to re-commit. So we don't score a fast sell by itself. We measure it in context: how often, across how many launches, at what graduation rate, and whether the dev ever comes back. A trader exiting one launch and a factory exiting four hundred a day are different animals — the table makes the difference visible. Recent-window samples per wallet, not lifetime audits.

Dev history persists.

We split every prolific dev's launches chronologically — classified them only on their early launches, measured only their later ones (7,412 devs, 167,723 later launches). Devs with at least one prior graduation completed future launches at 2.139% vs 0.307% for devs who had never graduated anything — and the gap widens when the big farms are excluded. Past KOL performance didn't survive this same test. Dev history did.

Honest limits: graduation measures completion, not your PnL — and a dev can buy their own graduation. Whether good-dev tokens pay buyers better is still accruing data (directional, underpowered so far). We publish that either way when it resolves.

Creator = the wallet signing the create transaction on our live event stream (mis-parsed feed rows are filtered and the filter is regression-tested). Graduation = bonding-curve completion (migration event). Sell/buyback timing from on-chain wallet activity, sampled over each wallet's recent window. Full method on the methodology page.