Welcome. Here I publish posts, deep dives and technical notes, with a simple philosophy: minimalism...
Below you'll find the latest posts.
-
Bitcoin and quantum computing: counting for yourself with a node
The third piece of the «Bitcoin and quantum computing» family: not the numbers, but how to redo them. Every command needed to count for yourself, with your own node, how many coins already have their public key in view: the census of the UTXO set by lock type and by age, the pass that counts reuse, the outpoint index, the check of your own addresses without handing them to anyone. Three levels, from minutes to days, that can be left half done while still giving a verifiable result. Every step carries the real output of our own run at block 957,301 and the fingerprints anyone must land on by redoing it. Public code, MIT licence.
-
Bitcoin's locks
Every coin on Bitcoin sits behind a lock: the scriptPubKey, the program that states under which conditions it can be spent. A reference piece on the lock types, from the first outputs of 2009 to Taproot and the quantum-resistant proposals: P2PK, P2PKH, bare multisig, P2SH, SegWit (P2WPKH and P2WSH), P2TR, each shown byte by byte on a real transaction verifiable with a node. What each type exposes of the public key, why a lock and an address are not the same thing, and how to read the language of scripts: pushes, ops and the stack.
-
The number used only once
Every digital signature consumes, besides the key, a second secret number generated for the occasion and good only once: the nonce. If it repeats, two signatures give back the private key with a subtraction: it happened to the PlayStation 3 and to Bitcoin wallets on Android in 2013. How the recovery works in ECDSA and in Schnorr, shown on the bytes of real signatures; why a nonce that is merely predictable betrays as well; and the point that concerns Bitcoin: a node validates a signature born from a disastrous nonce like any other, consensus cannot notice.
-
How big is «practically impossible»?
«It is practically impossible for someone else to generate your own seed» sounds like a way out, and anyone who does not buy it deserves an answer rather than reassurance. How big the space a Bitcoin wallet's seed comes from really is, why the right question is not whether someone guesses yours but whether two identical seeds will ever turn up among all the seeds ever generated, what it would cost a perfect machine just to count them, what difference twelve words instead of twenty-four actually makes, and what a passphrase adds. The fragile point is not the coincidence but the birth, which is to say the generator: the Android episode of 2013 and the Coldcard defect of July 2026.
-
A journey inside the machine that answers: how the bytes move in inference
What happens, physically, when a language model answers, and what changes if the one answering is a single machine, at home? A path through the principles of how it works, from attention to the craft of inference, with its two phases, its organization of the data, the techniques that make it cost less, read along one thread: where the bytes are and how much it costs to move them. In the background, an idea: local inference, as a choice and as a possibility, already matters today and will matter more and more. Keeping it concrete, an engine written in C that brings a model of hundreds of billions of parameters to run locally: DwarfStar, with which antirez shows how far one can push, and that this road can be travelled.
-
Bitcoin and quantum computing: the data (July 2026)
How much of all bitcoin already has its public key in view, since when, and at what pace it keeps exposing itself: 7.02 million BTC, about 35% of the circulating supply, of which 5.08 million exposed by address reuse and 1.93 million by construction (the original P2PK and Taproot). A census of the UTXO set at block 957,301, carried out with an ordinary node and reproducible down to the fingerprint, read table by table and figure by figure: concentration per lock, the curve by epoch of revelation, the pace of reuse year by year. This text is a deep dive off the path that starts from the post on the quantum computing threat: there the reasoning and the wider considerations, here the numbers and the method to check them with.
-
Bitcoin and quantum computing
A quantum computer will not be able to rewrite Bitcoin's ledger: it will probably be able, one day, to open some of its locks. What Shor's and Grover's algorithms actually do, why elliptic curve signatures are the target and hashes are not, which coins already have their public key in view (about 35% of the circulating supply, recounted on our own node) and which do not, where the machines stand with dates attached, where the post-quantum migrations under discussion stand (BIP-360, BIP-361), and why the hard part is not technical but a matter of governance: what to do with the coins of those who can no longer move them. A snapshot declared as of July 2026.
-
Maxwell's demon, taken seriously
An imaginary being watching a single molecule seems to extract work out of nothing, in defiance of the second law of thermodynamics. Dissolving the paradox took a century and a new idea: information is physical. Shannon's entropy of messages and Boltzmann and Gibbs's entropy of molecules answer the same question, Landauer's principle fixes the exchange rate between the two, and Szilárd's engine puts it to the test: how much a bit weighs, where it ends up when it is erased, and why there is no free lunch either way.
-
The prisoners meet the generals
Getting strangers who do not trust each other to agree on a single truth, with no arbiter and over an unreliable channel: that is the Byzantine generals problem, the heart of distributed consensus. Classical solutions hold only among known, counted participants; Bitcoin solves it on an open network with proof-of-work, which makes lying expensive and turns the generals into a defused prisoner's dilemma. Out of it also comes a time chain: a chain with a direction, which fixes the order of events. The value lies there, in the problem solved.
-
A fingerprint and a clock no one controls
A fingerprint says whether a text has changed, a clock no one can move lets one establish when this text existed, a stable signature says who it comes from: three already-known tools, put together, that do not abolish trust but shift it from whoever publishes to what anyone can check. A concrete use case, not a showcase of buzzwords.
-
Is a ban, in the world of AI, inevitable?
«Is a ban in AI inevitable?» is the wrong question: not because the answer is no, but because it hides the only ones that matter, which ban, decided by whom and at what cost, including the cost few put on the books. Really, "the people are minors"?
-
Caution in the wrong place
Governing AI focused only on following the rules can make a company cautious where the risk is small and careless where it is large; worrying about being compliant does not necessarily mean deciding with awareness
-
The energy cost of a language model
Where the energy cost of a language model really comes from when it answers
-
…it may not exist or you may not have access to it…
Reflections on the importance of technological independence for individual states, individual organizations and, why not, individual people