Crypto Activity to Posts

What this book is

The honest pipeline, not the assumed one

The pitch for a book like this writes itself before anyone checks the code: a VPS running around the clock, a cron job firing on schedule, an AI drafting fresh commentary for every alert. Almost none of that describes what actually runs. This book covers a real, deployed pipeline that turns Kaspa blockchain activity into automated posts on X, and the honest version is more useful to build from, because it's the version a reader can actually copy.

Nothing in this pipeline runs on a VPS — the thing that resembles a server is a GitHub Actions workflow, borrowed compute that exists for a few hundred minutes and then disappears, restarted on a schedule built to beat a platform timeout rather than to follow a calendar. Nothing is AI-drafted, either — every post this bot has ever sent came from the same fixed text template with different numbers substituted in. Kaspa Whale Watch is the real worked example: a bot that watches Kaspa's public blockchain for large transactions and posts an alert to X when one crosses a threshold, existing in two independent implementations using two different strategies to walk the same chain — a genuine before-and-after rewrite, complete with a real feature the newer version dropped.

What you’ll learn

Eight chapters, from the real pipeline to where an LLM would slot in

• the-real-pipeline — The full real pipeline end to end, before any assumptions get corrected.
• walking-the-chain-two-ways — Two competing implementations of the chain-walking logic, a genuine rewrite story.
• from-transaction-to-tweet — How a transaction becomes a tweet, and the rate limit that never made it into the second version.
• dry-run-as-a-design-feature — A dry-run mode that lets the whole pipeline be developed without live credentials ever touching it.
• where-scheduled-actually-runs — What a GitHub Actions schedule actually is, and why a stateless watcher fits it.
• the-scripts-that-never-post-anything — Calibration scripts that scan the chain but never post, and why that separation is deliberate.
• where-an-llm-would-actually-slot-in — Where an AI-drafting step would fit this pipeline, stated as a proposal rather than existing behavior.
• conclusion — Listener, classifier, template, rate-limited publisher, ephemeral scheduler — the general pattern this example teaches.

From the book

A preview

The version of this bot running in production today carries no rate limit protecting it from posting too often, even though an earlier version had exactly that protection.

Who it’s for

Anyone building an on-chain-to-social pipeline with Claude

This is for anyone building an on-chain-event-to-social-post pipeline with Claude who wants the real shape of program that fits ephemeral, restart-prone compute — including a credential-handling lesson learned firsthand when a live key turned up sitting in a plaintext file on disk.

A longer excerpt is available to newsletter subscribers.

Buy Ebook on Amazon