What actually graduates on pump.fun — the numbers

Pump.fun is usually described by its top of funnel: tens of thousands of token launches every single day. The more interesting number is the bottom of the funnel.

The funnel, roughly

| Stage | Order of magnitude | |---|---| | Tokens launched per day | ~20,000–30,000 | | Reach a meaningful holder base | a few hundred | | Graduate the bonding curve | ~1% of launches (≈100–300/day) | | Still trading actively a month later | a fraction of graduates |

Graduation rates have oscillated between roughly 0.5% and 2% over the launchpad's life — spiking in mania phases, compressing in quiet ones. In early 2026 the rate climbed back above 1% for the first time in months, a signal that usually accompanies renewed memecoin risk appetite.

Why the graduation rate is worth watching

Most crypto metrics are gameable or vanity. The graduation rate is neither:

• It's a paid signal. Filling a curve takes real net SOL inflow. You can bot volume, but you can't fake the curve filling — someone paid for every step.
• It's a market-wide sentiment gauge. When the rate rises, marginal attention is flowing to new launches rather than established tokens. It's a speculative risk-appetite index that updates hourly.
• It's the launchpad's quality bar. Everything below graduation is noise by construction; everything above it has cleared the same objective threshold.

That's why our graduates board is the data spine of this site. We don't track the 99% — the chain already forgets them by dinnertime. We track the survivors, because the survivors are where every interesting pump.fun story (Fartcoin, PNUT, the AI-agent economy, the ICM experiment) actually started.

The 99% is not waste — it's the price of the 1%

A common criticism: "almost everything on pump.fun fails." True, and it misses the point. The launchpad is a permissionless attention market. The cost of listing is near zero, so the failure rate is supposed to approach 100% — the same way most tweets get no likes. What matters is that the filter is cheap, fast, public, and the same for everyone.

Compare that to the old model: insider allocations, paid listings, months of runway burned before the market ever votes. The curve compresses that entire process into hours and settles it in public.

What we'll track from here

Starting with this piece, pump.blog maintains a running view of:

1. Daily graduates — the live board, refreshed every minute.
2. Graduate quality — how many are still trading a week and a month later.
3. Narrative concentration — which stories (AI agents, news cycles, ICM startups) the graduating class clusters around.

If pump.fun is an attention exchange, graduates are its listings page. Someone should be reading it like one. That's the job here.