–
blockchain height
TESTNETTESTNET
A blockchain where hashing only ever runs on hits that already cleared real radio astronomy data triage — four deterministic rules, recomputed by every node, before any nonce is even tried.
A Bitcoin blockchain spends electricity solving puzzles that serve no purpose beyond securing the network itself. Projects like SETI@home showed, over two decades, that millions of people are willing to donate processing power just to help sift through real radio telescope data — with no reward at all. Galaxy SIGNAL combines the two ideas: the "proof of work" is real scientific work, and whoever does it gets rewarded for it.
Every block starts from a genuine Breakthrough Listen hit — ~28.9 million real radio detections, L band, hundreds of real observed stars.
Four deterministic rejection rules -- real cadence position, coincidence with the telescope's own reference pointings, coincidence with other targets, and drift -- recomputed independently by every node from the same public data. This is the real gate; a hit either clears it or it doesn't. A classifier also scores each hit, but that score is declared by the miner and never independently verified -- it's published on the block as metadata, with no effect on whether the block is valid or what it pays (see the whitepaper for the full mechanism, and /audit for the 2026-08-26 correction that established this).
Only after a hit clears those rules does the final step of Galaxy SIGNAL's Proof-of-Useful-Verification (PoUV) kick in: SignalX, a hashing algorithm designed and built specifically for this project — a memory-hard virtual machine (32 general-purpose registers, a rotating 32MB memory table, real BLAKE2b hashing) designed to narrow the usual GPU-vs-CPU hashing gap, with real-hardware test vectors rather than a resistance claim taken on faith. Whether it holds up against every modern GPU's cache hierarchy is an open question this project doesn't consider closed. Without clearing the rejection rules first, no nonce turns a rejected hit into a valid block — not even with all the hashpower in the world.
The search for signals of technological origin in space — so-called technosignatures — today generates volumes of data that far exceed manual analysis capacity. Breakthrough Listen, the largest SETI (Search for Extraterrestrial Intelligence) program ever funded, publicly makes available petabytes of radio observations collected at telescopes like the Green Bank Telescope and the Parkes Observatory — it's exactly this open archive that feeds every block on this network: ~28.9 million real L-band detections, hundreds of real observed stars.
The public data exists, is free, and is ready to use — but analyzing it requires
specific tools and methods. Telling a real signal apart from human-made interference
(RFI) is a classic pattern-detection problem, well suited to data-mining and
machine-learning techniques. That's the reason this network's Proof-of-Useful-Verification
(PoUV) draws on the same kind of pipeline an active community already maintains in the
open (blimpy, turboSETI, the UC Berkeley SETI Research Center) —
not by running that pipeline live, but by mining against its already-published output,
with results that can be independently validated and compared against it.
This network doesn't re-run these five stages — it mines against their finished output, a table of already-detected hits. Shown here so it's clear what kind of processing already happened before any of this data reached the network, not as a description of this network's own live pipeline.
Searching and selecting observations via the Breakthrough Listen public API, filtering by target, frequency, and file type.
Reading filterbank/HDF5 files with blimpy, converting to
time-frequency spectrograms, and data cleaning.
Detecting narrowband signals and Doppler drift with turboSETI, producing
the hits table this network actually mines against.
ON/OFF cadence — alternating observation of the target with a reference — recorded
directly in that hits table (the status field), and what this network's
own four rejection rules recompute independently for every hit.
This network's own corrected gate and the original survey's published results reach the same conclusion by independent methods — see /audit for how that convergence was checked, not assumed.
| Source | Data type | Use in the project |
|---|---|---|
| Breakthrough Listen turboSETI hits table (Enriquez et al. 2017 corpus) | A CSV of already-detected hits — frequency, drift rate, SNR, cadence position, sky position, per-row file position | Primary source — the actual data this network mines against. The raw filterbank/HDF5 observation files aren't part of this corpus at all; see /audit for that boundary. |
| Enriquez et al. 2017 | Published paper (692 stars, L band) — the corpus's own drift-rate search parameter matches this paper exactly, confirmed independently; see /audit | Validation baseline — this network's corrected gate and the paper's published vetting reach the same conclusion by independent methods, not by replicating the paper's procedure |
The webminer is the recommended way to start — nothing to install, and no unsigned binary to trust. A desktop app is also available if you want mining to keep running in the background.
Open a page and start mining right from the browser — Windows, Mac, Linux, doesn't matter. A free account (email/password or Google) is required — the wallet and mining controls only appear once you're logged in, so a backup of your key is never optional or an afterthought. Creating one takes a few seconds, and the wallet itself still generates client-side, in your browser, the moment the account exists — see section 8 for exactly what that account does and doesn't change about who holds the key. If you'd rather not create a server account, the CLI and desktop app below use a local, password-protected wallet file instead — a different setup, not a simpler one, and their builds are currently unsigned with no published checksum (see below), so that trade isn't automatically the more cautious choice either.
/miner and log in or create a free accountWallet, mining dashboard, and a 3D signal waterfall — the same self-custody wallet as the webminer, running natively.
Freshly built by CI on every push — unsigned test builds, but each one's SHA-256 above is computed live from the file actually being served (never a hardcoded value) so you can verify what you downloaded matches. If you don't need mining to run in the background, the webminer above needs nothing installed.Same mining engine as the desktop app, without the wallet UI or the waterfall — for a
headless server, a script, or a background process managed some other way. Takes an
existing wallet (create one with the desktop app, the webminer, or
galaxysignal-wallet) and a --wallet-password-env flag to run
unattended.
On 2026-08-26 we corrected a real error in the mining gate — 120 of the network's first 263 blocks were paid on a miscalibrated filter. Read the full record.