BitBOINC runs more than one independent research effort on the same donated CPU/GPU time: hunting for the private keys of long-unclaimed Bitcoin "puzzle" addresses, modeling gravitational microlensing events with Kampylos, and more as they launch. Read the full explanation.

bitboinc is a volunteer computing project running more than one independent research effort on the same donated CPU and GPU time -- each one described below, in its own section, with its own explanation of what it searches for and why you should trust it. More get added over time; this page grows with them.

Kampylos: gravitational microlensing

A distributed-computing effort to exhaustively model binary-lens gravitational microlensing events -- a background star's light bent by a foreground pair of masses -- using the same donated compute as the rest of this project. Full project details, source code, and published results are at the Kampylos repository.

Finding every viable solution for one of these events requires an expensive search over the lens parameters (mass ratio, separation, source trajectory angle) -- the fit landscape often has several genuinely different parameter combinations that explain the same observed light curve almost equally well, and it's easy to miss one if you're not looking hard enough. Public sky surveys (OGLE, MOA, KMTNet) have alerted on tens of thousands of these events since the 1990s, and only a fraction of the ones showing this kind of binary-lens behavior have ever gotten a full, exhaustive treatment -- it's normally too computationally expensive for any one research group to do for every event, so most published analyses stop at whichever solution was found first, not a complete accounting of the parameter space.

This is exactly the kind of problem donated volunteer compute is good at that a single research group's own compute budget isn't: running the same exhaustive grid search across a far wider parameter range, over far more archival events, than any one group could normally afford.

What this can realistically produce

In order of how confident we are in each:

A full solution-map dataset for tens of thousands of archival events -- this doesn't currently exist anywhere, and simply completing the search produces it. Published openly, not kept private.

Real corrections to the published literature -- for at least some events, an exhaustive search will very likely turn up a better-fitting or additional solution than what's currently published, since the standard approach is known to sometimes miss a genuine alternative.

Possibly a genuine new discovery -- an anomaly nobody has fully modeled turning out to be a planet signal, for instance. Real, but not something we can promise in advance out of tens of thousands of events.

The Nancy Grace Roman Space Telescope has already launched, in August 2026, eight months ahead of its original schedule; its own dedicated microlensing survey starts science operations around 2027, and is expected to add thousands more binary-lens events a year at a rate current per-event analysis methods can't fully keep up with. This search runs against today's archival events already -- it doesn't wait on Roman -- but is built to fold Roman's much larger stream of events in as they start arriving.

Why should I trust this one?

Unlike the Bitcoin puzzle hunt described elsewhere on this page, there's no money or wallet involved anywhere in Kampylos -- the trust question here is about data and methodology, not fund custody. The source photometry comes entirely from public survey archives (linked in the repository), the search code is open source, and results (including the full solution-map data, not just whichever answer looks most interesting) are published openly rather than cherry-picked. Nothing about running this app is any different, risk-wise, from running Potamos or Keraunos -- same sandboxed BOINC app, same "reads a small work packet, writes back a result" model, nothing else on your machine.

Nostos: where do interstellar objects come from?

In 2017 the first object ever seen to come from outside the Solar System, 1I/'Oumuamua, swept past the Sun on a path no planet, asteroid or comet of our own could have. Two years later came 2I/Borisov, an interstellar comet, and in 2025 a third, 3I/ATLAS. All three were born around some other star and have been drifting through the Milky Way for millions, perhaps billions, of years. Nostos (Greek for "homecoming") tries to find which star that was.

The idea is simple; the scale is not. Take the object's measured path, run it backwards through the gravity of the Milky Way, and do the same for the stars around us: wherever the object once passed very close to a star, and slowly, that star is a candidate for its birthplace. Gaia DR3 gives a full 3D position and velocity for about 33 million stars, and Nostos follows the object back 100 million years against every single one of them. Because an orbit is never known perfectly, each star is checked against 100 slightly different copies of the object, spread the way its measurement uncertainty allows.

The best earlier searches (Bailer-Jones and colleagues, 2018 and 2020) used the roughly 7 million stars with radial velocities in Gaia's previous data release. Gaia DR3 has about 33 million, measured more precisely, and Nostos runs every one of them, for each interstellar object in the search, with the orbit uncertainty included for every star. That is about 2.4 CPU-years of computing per object -- far beyond one research group's budget, and exactly what donated computers are good at. Each task follows one object against 1500 stars and reports only the stars that came within 50 parsecs; everything else is discarded so results stay small.

What this can realistically produce

In order of how confident we are in each:

A complete catalogue of close stellar encounters for 'Oumuamua, Borisov and 3I/ATLAS over the last 100 million years, covering every Gaia star with a measured radial velocity. This doesn't exist yet, and finishing the search produces it. It will be published openly.

A check of the candidate origin stars proposed in the literature, with newer Gaia data and a full uncertainty treatment -- some will hold up, some won't.

Possibly a convincing birthplace: a star the object passed slowly and very close to, recently enough for the trajectory to still be trustworthy. Honestly, this may never be possible for objects that have been travelling for billions of years; finding that no star fits is a real result too. And the Vera C. Rubin Observatory is expected to find many more interstellar objects in the coming years, each of which can be added to the same search.

Why should I trust this one?

There is no money involved anywhere in Nostos. Everything it uses is public: the Gaia DR3 catalogue (with the official parallax zero-point correction), the published orbits of the interstellar objects, and a standard, published model of the Milky Way's gravity. Before a single task went out, the pipeline had to reproduce the published results of independent studies: the closest known encounters of 'Oumuamua with HIP 3757 and HD 292249, and of Borisov with Ross 573, all come out inside the published uncertainty ranges. Every close encounter the volunteers find is checked again with the star's own measurement errors before anything is claimed, and the progress and closest stars found so far are public on the Nostos page.

Phygas: where were the runaway stars born?

Most stars are born in clusters and drift away from them gently. Some leave in a hurry: when a massive star in a binary explodes as a supernova, its companion is flung out as a runaway star, and the neutron star left behind -- a pulsar -- gets a kick of hundreds of kilometres per second. Close encounters in crowded young clusters throw stars out too. Phygas (Greek for "fugitive") traces these fugitives back to the cluster they came from: which one, when they left, and how fast they were thrown out.

Every Gaia DR3 star with a full 3D velocity (33 million) and every pulsar with a measured proper motion (about 700) is run backwards through the Milky Way together with more than 4,000 young star clusters and moving groups (Hunt & Reffert 2024), both sides drawn many times from their measurement errors. Pulsar radial velocities are never measured, so they are drawn from the known spread of neutron-star birth velocities. Stars are first screened against every cluster in a fast linear pass; promising pairs and every pulsar get the full Monte Carlo.

What this can realistically produce

An all-sky catalogue of the stars and neutron stars that may have left each young cluster, with the time and the ejection speed of each departure. Studies so far have covered one region or a few dozen stars at a time; a pulsar census against open clusters was published in 2025 for the 145 pulsars with independent distances, and Phygas extends it to every pulsar with a measured motion. Pulsar birthplaces give the time and place of the supernova and the speed of the kick, which is still poorly understood. Runaway-star and pulsar pairs that leave the same place at the same time would mark a binary disrupted by a supernova. Everything, raw data included, will be published openly.

Why should I trust this one?

No money is involved, and everything Phygas uses is public: Gaia DR3, the ATNF pulsar catalogue, the Hunt & Reffert cluster catalogue and the same Milky Way model as Nostos. The pipeline reproduces published results before anything new is claimed: the binary pulsar J1302-6350 comes out as a member of the cluster UBC 525, ranked first of more than 4,000 clusters, and the double pulsar J0737-3039 points to Theia 58, as recent studies found. Most stars that cross a cluster do so by chance, so every cluster's candidates are compared against a control run in which the stars' velocities are shuffled; only an excess over that chance level counts. Progress and the best candidates so far are on the Phygas page.

Bitcoin puzzle hunt: Potamos & Keraunos

Searches for the private keys of well-known, long-unclaimed Bitcoin "puzzle" addresses, using donated CPU and GPU time from volunteers all over the world.

Back in 2015, someone created a series of Bitcoin addresses, each funded with an amount of BTC, and each with its private key deliberately restricted to a known, sequentially-numbered range: puzzle #1's key is somewhere in a 1-bit range, puzzle #2's in a 2-bit range, and so on up to puzzle #160's 160-bit range. It was set up as a public challenge: whoever finds the key for an address can claim the coins in it, using nothing but publicly available information. Most of the small, early puzzles were solved years ago. Many of the larger ones are still unclaimed today, some holding real value, and that's what Potamos and Keraunos are for.

This is not Bitcoin mining. Mining creates new coins by competing to extend the blockchain, and it never stops. This project does nothing of the sort: it searches a fixed, famous set of addresses whose creator deliberately left them as an open challenge years ago, for anyone who finds the matching key to claim. No new coins are created, no blocks are mined, and the work has nothing to do with Bitcoin's network or its security.

Is this legitimate? Why should I trust it?

This is a fair question to ask before you donate real compute time and electricity to any project, and we'd rather answer it directly than leave you to wonder.

First: nothing about running this puts YOUR computer, YOUR files, or YOUR own wallets at any risk. The app only ever reads the small work packet the server sends it (a key range and a target address) and writes back a result file saying whether it found a match in that range. It never touches your disk outside its own BOINC slot directory, never opens a network connection of its own (all networking is BOINC's own client, the same one every other BOINC project uses), and never asks for or needs anything from you besides spare CPU/GPU cycles.

Second: the part that matters most -- what happens if a key is actually found -- is deliberately built so that no single moment of automation can move funds anywhere. See "If a key is found" below for exactly how that works, in real technical detail, not just a promise.

Third: bitboinc is one project in the Athena network, a set of independent open-source projects run by the same person under the same name, with a public track record you can look at independently rather than take on faith. This isn't a project that appeared out of nowhere to ask for your compute time.

If something here still looks unclear or you want to verify a specific claim, ask on IRC (see below) -- questions about exactly how any part of this works are welcome, and we'd genuinely rather explain it than have you guess.

How it works

1. Two apps, two methods. Potamos scans a range of private keys sequentially, one by one, checking each against the target address. It's simple and exhaustive, but only practical for smaller ranges. Keraunos runs Pollard's kangaroo algorithm instead, which is vastly faster than brute force -- but it only works on a puzzle whose public key has already been exposed on-chain (which happens if that address has ever sent a transaction, not just received one). Most of the still-unclaimed puzzles don't have an exposed public key, which narrows down what Keraunos can actually target at any given time.

2. Volunteers work together, not just in parallel. A kangaroo search doesn't just split a range into pieces and hand each piece to a different computer -- that would throw away most of the algorithm's advantage. Instead, every volunteer's computer runs its own randomized search over the same range, and periodically reports the distinguishing points it finds back to the project. The moment any two of those points -- from any two computers, possibly on opposite sides of the world -- collide, the private key can be computed directly. More volunteers genuinely means faster progress on the same target, not just more redundant work. Sphinx (below) reuses this exact same collaborative-search mechanism for a different target.

3. One active target at a time. Only one puzzle is worked at a time per app, so effort isn't spread thin. Solving a puzzle is detected automatically the instant a real collision happens (no polling, no delay): the server immediately stops generating any more work for that puzzle -- both Potamos and Keraunos -- and automatically activates the next queued target so volunteers' computers don't sit idle waiting for a human to notice. If someone outside the project claims a target's coins first (racing us to the same puzzle, which does happen -- these are public, famous addresses other groups also search), that's caught the next time the project owner checks the target's on-chain balance, and the switch to the next target follows the same automatic path.

4. Checking results. Every result your computer sends back is independently checked before it is accepted, the same way any BOINC project validates work, and you're credited for it.

If a key is found

This is the part most worth being precise about, so here is exactly what happens, in order:

1. The moment two distinguished points collide (or, for Potamos, the moment a direct match is found), the server encrypts the raw private key immediately, in the same process that discovered it -- using RSA-4096 with OAEP padding, a well-established, standard encryption scheme, not anything custom. Only the encrypted result is ever written anywhere: to the database, to the project's logs, and to a notification email sent to the project owner. The plaintext key exists only for a moment, in memory, during that one encryption step, and is never written to disk or logged.

2. The matching decryption key -- the only thing in the world that can turn that ciphertext back into a usable private key -- is not on this server, has never been on this server, and is not in any code repository. It exists only on hardware the project owner personally controls, offline.

3. Nothing moves automatically. There is no code path, anywhere in this project, that takes a found key and sends a transaction. The server's job ends at "store the encrypted result and tell a human." A person then manually retrieves the encrypted value, decrypts it offline on their own machine, verifies with independent math that the decrypted key actually corresponds to that puzzle's known target address (so a corrupted or wrong decryption can never be mistaken for a real find), and only then manually sweeps the funds to a hardware wallet. Each of these is a deliberate, separate, manual step -- not an automated pipeline with a human rubber-stamp at the end.

4. Once the funds are confirmed safely moved, they are split four ways: a share to keep the platform running, a share to science projects, a share to nature and environmental protection, and a share to the volunteers themselves. The volunteers' share is not divided evenly among everyone who ever ran the project -- it is given out by a random drawing among the people who contributed valid work on that specific puzzle recently, weighted by how much they contributed. A bigger contribution means better odds, not a bigger guarantee.

We don't publish exactly which sub-ranges have already been searched -- that information would let other, unrelated search efforts skip straight past the parts our own volunteers already covered. What we do publish is aggregate progress: how many distinguishing points have been found so far and roughly how many keys that represents, without revealing the map of where. See the puzzle progress page for the current numbers.

Sphinx: not Bitcoin, a separate cryptography challenge

Sphinx has nothing to do with Bitcoin. It targets Certicom's ECCp-131 discrete-logarithm challenge -- a deliberately unsolved cryptography problem published for anyone to attempt, with no coins, wallet, or address of any kind involved. It reuses the same collaborative kangaroo search as Keraunos above (every volunteer runs an independent random walk over the same target, and a collision between any two volunteers' distinguished points solves it directly), just applied to a completely different, non-Bitcoin elliptic curve. A solution here is a piece of mathematics, not a claim on money, so none of the "if a key is found" fund-handling process described above applies to it. We haven't independently reconfirmed whether Certicom's own historical prize for this challenge is still being honored; treat solving it as a genuine cryptographic achievement in its own right, not a guaranteed payout.

Chat with us

If you have questions about attaching, want help troubleshooting a task that failed, want to know what we're currently working on and why, or just want to talk to other volunteers and the people running the project, we're on IRC. It's also where we announce things first: a new puzzle target, a change in strategy, a new project launching, or -- hopefully one day -- a key actually being found.

Network: Libera.Chat (irc.libera.chat:6697, TLS)
Channel: #bitboinc

Any IRC client works. If you'd rather not hunt for one, we recommend our own client, Rhizome -- just point it at irc.libera.chat:6697 (TLS) and join #bitboinc.

This is the shared chat for every project on this page, and the wider Athena project family besides -- as anything here grows enough to need its own dedicated channel, we'll set one up and list it here.

Already joined? Log in.

Two new research projects: Nostos and Phygas
🌠 Two new research projects: Nostos and Phygas 💫

bitboinc now runs two new astronomy projects alongside Kampylos and the puzzle hunt. Both use Gaia's measurements of 33 million stars and run the clock backwards through the Milky Way, to find where things came from.

🌠 Nostos: where did interstellar objects come from?
'Oumuamua and Borisov are the first two objects we have seen arrive from another star's planetary system. Nostos traces each of them back 100 million years against every one of the 33 million stars, to find the stars they passed closely and slowly: candidates for their home systems. 'Oumuamua is running now, Borisov comes next.
Nostos page · Full description

💫 Phygas: where were runaway stars and pulsars born?
Some stars are thrown out of their birth cluster, by a supernova in a binary or by a close encounter, and neutron stars leave with a kick of hundreds of km/s. Phygas traces every Gaia star with a known 3D motion and every pulsar with a measured motion back against 4,087 young star clusters, to find which cluster they left, when, and how fast. Pulsar tasks go out first, then the stars.
Phygas page · Full description

How to join
Both are CPU applications for Windows, Linux, macOS and Android. If you haven't changed your project preferences, you'll get them automatically. Otherwise, enable Nostos and Phygas under Project preferences. Both have their own badges.

All results, raw data included, will be published openly. Thanks for computing with us!
9 Oct 2026, 20:57:53 UTC · Discuss

Name our Kampylos objects and get credited!
Kampylos has two new objects that need names: Object-2, a flaring dwarf nova we identified in a 2002 MOA alert, and Object-3, a 2003 microlensing event.

Suggest names in the forum thread until 14 October 2026: Help us name Kampylos Object-2 and Object-3!

Want your name in our publications? Results from Kampylos will be published in scientific notes, and we want to thank the volunteers who computed them. Add your official full name and country (as on your ID) on your account page under "Credit in scientific publications" (direct link). If you don't, you stay anonymous and are only counted.
7 Oct 2026, 11:56:49 UTC · Discuss

Boincsynergy has listed us.
I am happy that they prepared a detailed page for us. Thanks to everyone.
https://boincsynergy.ca/wiki/BitBOINC
3 Oct 2026, 18:40:00 UTC · Discuss

Scheduler and Validator Issues Resolved
The scheduler and validator problems have been fixed.

Due to Cloudflare's caching system, we ran into some issues, and during the process our scheduler log server and data compression server filled up. All of these problems have now been resolved.

We are currently seeing around 23,000 bot requests originating from the United States. We don't know exactly what this is it looks like a bot attack or a data-scraping attempt from the outside. Because of this, Cloudflare and the pfSense firewall on our network are currently rate-limiting all incoming requests. We are experiencing a temporary network slowdown, which means IPv4 connections are currently delayed and served through the cache.

For the best performance and the fastest connection to the server, please enable IPv6 on your device and network. Our server has supported IPv6 from the day it was set up.
Last 24-hour statistics:

2 Oct 2026, 12:56:57 UTC · Discuss

🚀 BitBOINC is fully up and running! 🚀
🚀 BitBOINC is fully up and running! 🚀

Both of our search apps are now active and working real puzzles:

⚡ Keraunos (kangaroo search) -- currently on puzzle #140
🌊 Potamos (sequential scan) -- currently on puzzle #71

Full platform support is live: Windows, Linux, macOS, and Android on CPU, plus GPU acceleration via NVIDIA CUDA, OpenCL (AMD/Intel/Apple), and Apple Silicon Metal.

💰 Credit system fixed -- credit is now computed the native BOINC way, proportional to the real work your hardware actually does, instead of a mismatched estimate. If you felt Potamos credit looked low before, that's fixed going forward.

🏅 Also live: an achievement badge system (100 badges across 10 categories) and an infinite level/title system -- check your profile page, and the badge gallery to see what's still locked.

Thanks for crunching with us -- and if you hit any bugs, reply on the relevant pinned topic below. 🙏
29 Sep 2026, 14:35:10 UTC · Discuss

News is available as an RSS feed