Blog

Notes from the tape.

What we recorded, what we tested, and what we got wrong along the way.

• October 14, 2026Filters people swear by, and how to check them

  Bundles, fast curve fills, the dev wallet, time of day. Every bot builder hears filter rules. Here is how to test one against outcomes instead of trusting it.

• October 14, 2026The first 2 minutes after graduation

  What happens right after a pump.fun coin graduates, why your bot's reaction time matters so much there, and how to measure your own latency.

• October 14, 2026How to give your AI the right data

  How to prompt ChatGPT, Claude or Cursor to build or test a trading bot on real files, what to paste, how big a file to start with, and how to make it show its assumptions.

• October 14, 2026Historical pump.fun data: Bitquery vs Birdeye vs Solana Tracker vs Dune vs Launch Tape

  A plain comparison of five ways to get pump.fun and Solana launch data for building or backtesting a bot: live APIs, a query engine, and flat files.

• October 14, 2026Your backtest is lying: 5 things to fix before you trust it

  Most first backtests of a memecoin bot look great because they test a bot that could never exist. Here are five fixes that make the numbers honest.