Jeremy Rubin's Blog

Here you'll find an assorted mix of content from yours truly. I post about a lot of things, but primarily Bitcoin.

categories: Bitcoin, Shenzhen Journey.

FE'd Up Covenants

Covenants are a way of expressing restrictions on Bitcoin. Covenants, while possible to implement as an extension to Bitcoin, do not exist natively. To enable them requires the Bitcoin community to agree upon upgrades such as CTV, CAT, CSFS, and more. This paper serves to demonstrate at a high level how covenants could be introduced to Bitcoin without a soft fork using Functional Encryption and Zero Knowledge Proofs. Read the full paper.

Spookchains: Drivechain Analog with Trusted Setup & APO

This post draws heavily from Zmnscpxj’s fantastic post showing how to make drivechains with recursive covenants. In this post, I will show similar tricks that can accomplish something similar using ANYPREVOUT with a one time trusted setup ceremony.

This post presents general techniques that could be applied to many different types of covenant.

note: I originally wrote this around May 5th, 2022, and shared it with a limited audience

Spooky Chains

Peano Counters

The first component we need to build is a Peano counter graph. Instead of using sha-256, like in Zmnscpxj’s scheme, we will use a key and build a simple 1 to 5 counter that has inc / dec.

Assume a key K1…K5, and a point NUMS which is e.g. HashToCurve(“Spookchains”).

Generate scripts as follows:

<1 || K1> CHECKSIG
<1 || K5> CHECKSIG

Now generate 2 signatures under Ki with flags SIGHASH_SINGLE | SIGHASH_ANYONECANPAY | SIGHASH_ANYPREVOUT.

Rule Increment

For each Ki, when i < 5, create a signature that covers a transaction described as:

Amount: 1 satoshi
Key: Tr(NUMS, {<1 || K{i+1}> CHECKSIG})

Rule Decrement

For each Ki, when i > 1 The second signature should cover:

Amount: 1 satoshi
Key: Tr(NUMS, {<1 || K{i-1}> CHECKSIG})

Are these really Peano? Sort of. While a traditional Peano numeral is defined as a structural type, e.g. Succ(Succ(Zero)), here we define them via a Inc / Dec transaction operator, and we have to explicitly bound these Peano numbers since we need a unique key per element. They’re at least spiritually similar.


Publish a booklet of all the signatures for the Increment and Decrement rules.

Honest parties should destroy the secret key sets k.

To create a counter, simply spend to output C:

Amount: 1 satoshi
Key: Tr(NUMS, {<1 || K1> CHECKSIG})

The signature from K1 can be bound to C to ‘transition’ it to (+1):

Amount: 1 satoshi
Key: Tr(NUMS, {<1 || K2> CHECKSIG})

Which can then transition to (+1):

Amount: 1 satoshi
Key: Tr(NUMS, {<1 || K3> CHECKSIG})

Which can then transition (-1) to:

Amount: 1 satoshi
Key: Tr(NUMS, {<1 || K2> CHECKSIG})

This can repeat indefinitely.

We can generalize this technique from 1...5 to 1...N.

Handling Arbitrary Deposits / Withdrawals

One issue with the design presented previously is that it does not handle arbitrary deposits well.

One simple way to handle this is to instantiate the protocol for every amount you’d like to support.

This is not particularly efficient and requires a lot of storage space.

Alternatively, divide (using base 2 or another base) the deposit amount into a counter utxo per bit.

For each bit, instead of creating outputs with 1 satoshi, create outputs with 2^i satoshis.

Instead of using keys K1...KN, create keys K^i_j, where i represents the number of sats, and j represents the counter. Multiple keys are required per amount otherwise the signatures would be valid for burning funds.

Splitting and Joining

For each K^i_j, it may also be desirable to allow splitting or joining.

Splitting can be accomplished by pre-signing, for every K^i_j, where i!=0, with SIGHASH_ALL | SIGHASH_ANYPREVOUT:

Input: 2^i sats with key K^i_j
    - 2^i-1 sats to key K^{i-1}_j
    - 2^i-1 sats to key K^{i-1}_j

Joining can be accomplished by pre-signing, for every K^i_j, where i!=MAX, with SIGHASH_ALL | SIGHASH_ANYPREVOUT:

    - 2^i sats with key K^i_j
    - 2^i sats with key K^i_j
    - 2^i+1 sats to key K^{i+1}_j

N.B.: Joining allows for third parties to deposit money in externally, that is not a part of the covenant.

The splitting and joining behavior means that spookchain operators would be empowered to consolidate UTXOs to a smaller number, while allowing arbitrary deposits.

One Vote Per Block

To enforce that only one vote per block mined is allowed, ensure that all signatures set the input sequence to 1 block. No CSV is required because nSequence is in the signatures already.

Terminal States / Thresholds

When a counter reaches the Nth state, it represents a certain amount of accumulated work over a period where progress was agreed on for some outcome.

There should be some viable state transition at this point.

One solution would be to have the money at this point sent to an OP_TRUE output, which the miner incrementing that state is responsible for following the rules of the spookchain. Or, it could be specified to be some administrator key / federation for convenience, with a N block timeout that degrades it to fewer signers (eventually 0) if the federation is dead to allow recovery.

This would look like, from any K^i_j, a signature for a transaction putting it into an OP_TRUE and immediately spending it. Other spookchain miners would be expected to orphan that miner otherwise.

Open States / Proposals

From a state K^i_1, the transaction transitioning to K^i_2 can be treated as ‘special’ and the OP_RETURN output type can be used to commit to, e.g., the outputs that must be created in when the Terminal State is reached. This clarifies the issue of “what is being voted on”.

This method does not lock in at a consensus layer what Terminal State is being voted on.

In certain circumstances, without violating the one-time-setup constraint, if a fixed list of withdrawer’s addresses is known in advance, the Open States could cover withdrawals to specific participants, which then must collect a certain number of votes from miners. However, it seems impossible, without new primitives, for an arbitrary transaction proposal to be voted on.

Setup Variants


Instead of using randomly generated keys for each state, define each to be an xpub and derive a path where it is k/i/j for each state/satoshi amount. This saves some data, and also requires less entropy.

Trustless Data Commit:

commit to the hash of the entire program spec as a tweak to the xpub, so that someone can quickly verify if they have all the signatures you are expected to generate if honest.

One way to do this is to convert a hash to a list of HD Child Numbers (9 of them) deterministically, and tweak the xpub by that. This is a convenient, yet inefficient, way to tweak an xpub because the child has a normal derivation path for signing devices.

Single Party

A single party pre-signs all the transactions for the spookchain, and then deletes their xpriv.

You trust them to have deleted the key, and signed properly, but you do not trust whoever served you the spookchain blob to have given you all the state transitions because of the trustless data commitment.

MuSig Multi-Party

Define a MuSig among all participants in the setup ceremony, N-of-N.

Now you simply trust that any one person in the one-time-setup was honest! Very good.

Unaggregated Multi-Party

Allow for unaggregated multi-sig keys in the spec. This grows with O(signers), however, it means that a-la-carte you can aggregate setups from random participants who never interacted / performed setup ceremonies independently if they signed the same specs.

Can also combine multiple MuSig Multi-Parties in this way.

This is nice because MuSig inherently implies the parties colluded at one point to do a MuSig setup, whereas unaggregated multi-sig could be performed with no connectivity between parties.

Soft Forking Away Trust

Suppose a spookchain becomes popular. You could configure your client to reject invalid state transitions, or restrict the spookchain keys to only sign with the known signatures. This soft fork would smoothly upgrade the trust assumption.

Symmetry of State Transition Rules & DAG Covenants

We could have our increment state transitions be done via a trustless covenant, and our backwards state transitions be done via the setup.

This would look something like the following for state i:

Tr(NUMS, {
    `<sig for state K_{i+1}> <1 || PK_nonsecret> CHECKSIG`,
    `<1 || Ki> CHECKSIG`

The advantage of such an optimization is theoretically nice because it means that only the non-destructuring recursive part of the computation is subject to the one-time-setup trust assumption, which might be of use in various other protocols, where recursivity might only be unlocked e.g. after a timeout (but for spookchains it is used at each step).

A compiler writer might perform this task by starting with an arbitrary abstract graph, and then removing edges selectively (a number of heuristics may make sense, e.g., to minimize reliance on one-time-setup or minimize costs) until the graph is a Directed Acyclic Graph, consisting of one or more components, compiling those with committed covenants, and then adding the removed edges back using the one-time-setup key materials.

Commentary on Trust and Covenantiness

Is this a covenant? I would say “yes”. When I defined covenants in my Calculus of Covenants post, it was with a particular set of assumptions per covenant.

Under that model, you could, e.g., call a 7-10 multi-sig with specific committed instructions as 4-10 honest (requires 4 signatories to be honest to do invalid state transition) and 4-10 killable (requires 4 signatories to die to have no way of recovering).

For emulations that are pre-signed, like the varieties used to emulate CTV, it is a different model because if your program is correct and you’ve pre-gotten the signatures for N-N it is 1-N honest (only 1 party must be honest to prevent an invalid state transition) and unkillable (all parties can safely delete keys).

I model these types of assumptions around liveness and honesty as different ‘complexity classes’ than one another.

What I would point out is that with the counter model presented above, this is entirely a pre-signed 1-N honest and unkillable covenant that requires no liveness from signers. Further, with APO, new instances of the covenant do not require a new set of signers, the setup is truly one-time. Therefore this type of covenant exists in an even lower trust-complexity class than CTV emulation via presigneds, which requires a new federation to sign off on each contract instance.

With that preface, let us analyze this covenant:

1) A set of sets of transaction intents (a family), potentially recursive or co-recursive (e.g., the types of state transitions that can be generated). These intents can also be represented by a language that generates the transactions, rather than the literal transactions themselves. We do the family rather than just sets at this level because to instantiate a covenant we must pick a member of the family to use.

The set of sets of transaction intents is to increment / decrement to a successor or predecessor, or to halve into two instances or double value by adding funds. Each successor or predecessor is the same type of covenant, with the excetion of the first and last, which have some special rules.

2) A verifier generator function that generates a function that accepts an intent that is any element of one member of the family of intents and a proof for it and rejects others.

The verifier generator is the simple APO CHECKSIG script.

3) A prover generator function that generates a function that takes an intent that is any element of one member of the family and some extra data and returns either a new prover function, a finished proof, or a rejection (if not a valid intent).

The prover generator is the selection of the correct signature from a table for a given script.

Run the prover generator with the private keys present once to initialize over all reachable states, and cache the signatures, then the keys may be deleted for future runs.

4) A set of proofs that the Prover, Verifier, and a set of intents are “impedance matched”, that is, all statements the prover can prove and all statements the verifier can verify are one-to-one and onto (or something similar), and that this also is one-to-one and onto with one element of the intents (a set of transactions) and no other.

At a given key state the only things that may happen are signed transactions, no other data is interpreted off of the stack. Therefore there is perfect impedance match.

5) A set of assumptions under which the covenant is verified (e.g., a multi-sig covenant with at least 1-n honesty, a multisig covenant with any 3-n honesty required, Sha256 collision resistance, Discrete Log Hardness, a SGX module being correct).

Uniquely, that during the setup phase at least one of the keys were faithfully deleted.

The usual suspects for any bitcoin transaction are also assumed for security.

6) Composability:

The Terminal State can pay out into a pre-specified covenant if desired from any other family of covenants.

7 Theses on a next step for BIP-119

Martin Luther had 99, I'm going to give you 7

Warning: this post assumes a great deal of context on CTV is already understood by the reader. If you are not familiar, you may wish to start with, the advent calendar, and the BIP-119 text and reference implementation, as those may provide much needed context about what CTV is and why a next step is being discussed. If you only have a little time, minimally I would advise these two from the advent calendar as the minimum required context: Contracting Primitives and Upgrades to Bitcoin and RoadMap or Load o’ Crap?.

This post starts with a conclusion:

Within a week from today, you’ll find software builds for a CTV Bitcoin Client for all platforms linked here:

  • Mac OSX TODO:
  • Windows TODO:
  • Linux TODO:

These will be built using GUIX, which are reproducible for verification. The intended code to be built will be which is based on Bitcoin Core v23.0 release candidate 5, with commit hash dd9a4e0ea8a109d1607ca1ec16119b1bc952d8b0. You can begin testing this immediately, and even producing your own GUIX builds as well.

Signatures for the builds will be available below:

  • TODO: … .asc

The source tarball:

  • TODO: … .tar.gz

The client has a Speedy Trial release similar to Taproots with parameters proposed to be:

  • Signal Start MTP: 1651708800 (May 5th, 2022, 00:00 UTC)
  • Signal Timeout MTP: 1660262400 (August 12th, 2022, 00:00 UTC)
  • Activation Height: 762048 (Approximately Nov 9th)

See the appendix to verify these parameters.

This ensures 6 signalling periods to activate CTV. The Start and Timeout are targeting mid-period (if hashrate stays steady) times to ensure that it is unlikely we would have more or fewer periods.

The week delay between this post and builds is to provide time for review on the selection of parameters as well as ability to rebase onto a final v23.0 release, should it become ready within the week. Backports are in the works for v22.0, but release builds may not be made available as Bitcoin’s release build processes have changed since v22.0 to use GUIX. The branch for backports is available here: with current commit hash 4d2c39314834a28cd46da943a12300cca8ffcb10, if you would like to help with testing.

Why this, why now?

I’ve just returned from the Bitcoin Miami “Bacchanal”. Personally, I had a couple different goals for being there1. One of my primary focuses was on talking to as many people as possible about BIP-119 and the future road to take.

While consensus has to happen among a much broader set of people than can fit in a conference in Miami, the reality is that there were more than 20,000 Bitcoiners at this event and a good representation across industry, developers, journalists, podcasters, plebs, whales, pool operators, miners, venture capitalists, and more. To say it was a representative sample wouldn’t be fair, but it certainly was not a homogeneous crowd. And I spoke to as many people as I could.

There were a couple common threads across the feedback I received:

  1. Agree or disagree with CTV in particular, folks generally liked how/that I was driving a conversation forward, and respected the hustle required to do so.
  2. A lot of people felt that CTV would help them in a tangible way, and more than a few times, individuals approached me with a concrete use case they needed but had not fully written it up yet.
  3. A lot of people wanted to know what the next step was and what I was planning to do to get it activated and when.

Some people had some suggestions on what I should do as a next step:

  1. Some folks said I should just do a UASF and rally the users.
  2. Some said I needed to organize a summit for developers to explore covenants2.
  3. Some said I didn’t need to do a UASF, nor advocate for it, but I did need to decide on exact release parameters and distribute a reproducible binary so that it was clear what should be run and when so that end-to-end activation testing could proceed.

It’s (un?)remarkably difficult to integrate all feedback on a complex topic like a Bitcoin upgrade coherently. But, having thought it through, I decided that the approach above was the correct next step. Below, you’ll find some reasoning on why I believe this to be proper and not out-of-line with how soft-fork development should go.

However, if I’m wrong in your view, consider me a mere messenger and please don’t shoot the messenger. You just need to communicate clearly to the community why they should not run and signal for CTV and I’m confident that the wisdom of the consensus set will decide in it’s best interests.

So why ship a binary and release parameters?

1) CTV passes a basic pre-flight checklist.

This discussion has to start anchored in a “pre-flight checklist” for CTV. These are fundamental questions that we should be able to tick boxes for for any proposed upgrade… Sadly, the community at large doesn’t have a codified checklist3, but personally I tick off the following boxes:

  1. No material changes to the BIP/Spec/Ref Impl necessary for ~2 years (beyond rebases).
  2. A reasonably well reviewed and tested PR-implementation exists.
  3. ~5 Months of a 5.5 BTC Bounty for bugs in CTV
  4. I socialized a similar roadmap 5 months ago, which received a reasonably warm response so there are ‘few surprises’ here against previously communicated intent.
  5. A community of supporters: breakdown, 16 supporting orgs, 109 individuals, 3 mining pools (totalling about 15-18% depending on when you look).
  6. Only 3 individual NACKs + 1 org (owned by one of the individuals). You should read them yourself, but I think the NACKs are summarizable as “it’s too soon” and “there should be a different process” rather than “I have identified a flaw in the proposal”. See section 4 for more on this. The NACKs are linked below:
  7. Ample time to have produced a nack with a technical basis.
  8. 7 regular meetings over ~16 weeks to discuss the upgrade.
  9. There exists software from multiple different parties for using CTV to accomplish various tasks. None of these users have uncovered issue or difficulty with CTV.
  10. Many in the community are arguing for more functionality than what CTV offers, rather than that the functionality of CTV might be unsafe. CTV ends up being close to a subset of functionality offered by these upgrades.
  11. CTV does not impose a substantial validation burden and is designed carefully to not introduce any Denial of Service vectors.
  12. There exists a Signet with CTV active where people have been able to experiment.
  13. Backports for current release and prior release (will be) available.

2) Unforced errors

In tennis, an unforced error is a type of lost point where a player loses because of their own mistake. For example, suppose your opponent shoots a shot that’s a lob high in the air and slow. But it looks like it’s going out, so you do a little victory dance only to uncover that… the ball lands in. You had enough time to get to the ball, but you chose not to because you didn’t think it would go in. Contrast this to a forced error – your opponent hits a shot so hard and fast across the court no human could reach it let alone return it4.

What’s this got to do with Bitcoin?

Community consensus is the ball, and we don’t know if by August it will be in or out.

Getting to where the ball might land is preparing a software artifact that can activate.

If an artifact isn’t prepared that does this, even if community consensus is ready by then, it’s an unforced error that it wasn’t ready which precludes us from being live in August.

When should you avoid an unforced error like this? When the cost of getting to the ball is sufficiently small.

I’m already maintaining a backportable to Bitcoin Core 23 and 22 patchset for CTV. It’s not much work to set parameters and do a release.

Which brings us to…

3) Product Management is not “my Job” – it’s yours.

Devs don’t swing the racquet, we get to the ball. It’s the community’s job to decide to swing or not. One might rebutt this point – the community isn’t well informed to make that call, but fortunately devs and other tuned in individuals can serve as “coaches” to the public and advise on if the swing should happen.

Producing the code, the tools, the reviews, the builds, the dates, these are all getting to the ball activities.

It’s possible that as a developer, one could say that we should not get to the ball unless we know the community wants to swing.

But that’s false. What if the community wants to take a swing at it but developers haven’t gotten to the ball? What if developers refuse to get to the ball because they don’t want the community to take that shot? Well, tennis can be a game of doubles (I’m really sticking with this metaphor), and maybe your teammate – the community itself – strives to go for it and sprints cross court to make up and take a shot. Maybe that shot looks like a UASF, maybe it looks like a hard fork, maybe it’s lower quality since there was less time to make the right shot placement (worse code if usual devs don’t review). But ultimately, the economic majority is what backstops this, the devs just have the opportunity to help it along, perhaps, a little smoother.

Largely, the formal critiques of CTV (the 3 NACKs) are based on topics of whether or not to swing the racquet, not if we should be at the ball. There are other critiques as well, about the generality of the upgrade, but we’ll discuss those later in this post.

I’ll excerpt some quotes from the NACKs below:

Michael writes,

I also think attempting regular soft forks with single features is a disturbing pattern to get into having just activated Taproot. As I say in the linked post it means less and less community scrutiny of consensus changes, it becomes a full time job to monitor the regular soft forks being activated (or attempting to be activated) and only a tiny minority of the community can dedicate the time to do that.

So this seems to be a point largely about product management – we should only take a shot when we can line up more than once, due to the cost of swinging the racquet. Not really my job, it’s the communities.

Hence I’d like to register a “No” soft signal as I fundamentally disagree that a CTV soft fork should be attempted in the near future and my concerns over premature activation outweigh my enthusiasm for digging into the speculated real world use cases of CTV.

If you disagree that it should be attempted, that’s fine. Your time to voice your concerns is in making the swing.

John writes,

Generally, I do not think Bitcoin is ready for any new soft forked features at all in the short term. Taproot just arrived and there is already so much work to be done to adopt and utilize it.

This is a product management point. “Work on feature X should block all work on other features”. It’s not a point on if CTV is a feature of independent merit.

Further, it’s a layer violation. Wallet progress is wholly independent of consensus upgrades, and generally, we don’t operate on a waterfall development model.

Any improvements CTV-eltoo may bring to LN are not significant enough to claim they are urgent for adoption or R&D for LN progress and adoption.

Again, a product management point. How do we measure what is important to the Lightning Community? Oh, on, there are multiple Lightning Network companies and individuals (Lightning Labs, Muun Wallet, Roasbeef, ZmnSCPxj, LN Markets, Breez, fiatjaf, and more). So if that represents the community, then it seems like a greenlight.

Since I am not qualified, nor are 99% of Bitcoiners, to evaluate this deeply, I feel much more time and scrutiny is required to accept CTV and the related projects Jeremy has prepared with it.

If you’re not qualified, remind me why we’re listening?

Sorry, I couldn’t help the snark. Graciously, let’s accept the framing for a second – who are the stakeholders who need to sign off? What is this process like, concretely?

Is this a process that happens before or after we ‘get to the ball’?

It can definitely be after we get to the ball, and the decision to swing or not is a bit too product-management-y for how a dev should engage.

Also, related projects are a bit like the mix-ins at Cold Stone Creamery™. You are free, of course, to just get ice cream! That there are a myriad of uses doesn’t mean you need to accept all of them, it’s sufficient to just consider the one or two you care about.

I am currently happy with what we have in Bitcoin, and would prefer Core development prioritize optimizations and cleanup work over more and more features that have no urgent need or chance of adoption anytime soon.

This belies a basic misunderstanding of FLOSS:

  1. People work on what they want to
  2. CTV is already ‘finished’

A non-dev’s preference to spend time on cleanup or optimization doesn’t make any dev write that code or shift focus. Most core devs don’t have a boss, and if they do, it’s probably not you! It’s structurally impossible to direct the attention of developers.

And it so happens that I am not happy with what we have in Bitcoin, so I did something about it. With respect to adoption, people will likely be using CTV pretty soon after it’s available, since it is a big step up for a number of critical uses like vaults. These applications are already being built. They can be used on signet which can be deployed to mainnet immediately5. The implementation details for basic custody contracts are pretty simple and don’t require the level of coordination for support that other contracts like lightning or DLCs, so adoption can be at the individual level.

The path around prioritization remains a product management question, and not something devs can be compelled to follow.

4) There are other things to work on.

I can get to the ball for this shot, but then I’d like to work on getting in position for the next shot on time.

There are other important technologies to work on, keeping covenants in limbo ties up a lot of human capital in trying to solve for getting something, vs. having something and being able to work on building solutions using it plus designing new technologies that make Bitcoin even better in different or complimentary ways.

What’s the right amount of rumination (chewing) to swallowing? Eventually, the mouthful you have is stopping you from taking the next bite.

5) Consensus is memoryless

A memoryless process is something that “never makes progress”. For example, consider a board game, where you need to roll a 6 to win. You expect to need 6 rolls to win. You roll a 5. How many more rolls do you need? It’s not 5. It’s 6 – the process is memoryless.

Clearly consensus isn’t entirely memoryless. Something that is a concept only obviously has to be turned into a hard artifact.

CTV has been a ‘hard artifact’ for 2 years. 2 years ago I took a poll of 40 or so developers who attended my workshop in Feb 2020. An average sentiment was that maybe we try to do CTV in a year or so, and that we could definitely do it maybe in 2 years.

I hear the same today from people who advocate a slower process. Maybe a year from now, we could definitely do it in maybe 2 years.

In 2 years, if we wait, won’t we hear the same?

Here’s a few reason why we might hear the same complaints in the future:

In 2 years, suppose Bitcoin is ~2x the price.

Shouldn’t we acknowledge twice as much at risk and do twice as much work to security audit things going into it? What if it’s not just El Salvador with Bitcoin as a national currency, but now Guatamala too?

Suppose we want to get to a point where 50% of the community has reviewed a change. In 2 years, what if the community is 2x the size? Then even if we hit 50% of today’s community, we only have 25% of the community read up.

Hopefully we keep innovating. And if we do keep innovating, we’ll come up with new things. If we come up with a new thing that’s 2x as good as CTV in 2 years, but it takes another 2 years to implement it concretely, should we wait till we finish that? But what happens when we come up with something 2x better than that? Wait another 2 years? Now we’re 4 years out from when the swing to get CTV over the net was doable, and we’d have 0 in hand solutions for the problems CTV tries to solve.

All this points to the nature of the memorylessness of trying to get consensus in an open network.

The best I can do is to get to the ball and let the community decide to take the shot.

Concretely – what is the cost to having CTV in bitcoin during this time while the “better” alternative is being worked on, if we do decide to activate knowing we might one day obsolete CTV? What are the benefits to having 3-5 years of basic covenants in the meantime? On the balance it seems, to me, net positive. It also seems to be a decision that the community at large can judge if the costs are worth the benefits.

6) You can fight against it.

A criticism of the soft fork process is it’s not safe with Speedy Trial (ST) and ST is bad so we shouldn’t do it. This is not a strong criticism: with Taproot, our most important upgrade in years, went smoothly even though there was sharp disagreement over the safety of the mechanism at the time.

Here’s a breakdown of why Speedy Trial is OK from the perspective of different participants:

You want CTV and won’t take no for an answer.

Start with a ST. After 3 months, it either goes or it doesn’t. At least we were at the ball. Now, let’s do a UASF with a LOT=true setting. Because ST is a fail fast, it’s in theory using time where otherwise we’d have to spend coordinating for the harsher realities of a LOT=true effort, so it’s a happy-path optimization.

If you’re a miner, you should signal during the period.

You’d like CTV and might take no.

ST is for you. Can always follow up with another ST if the first fails, or another option if your opinion changes.

If you’re a miner, you should signal during the period.

You do not want CTV, but if others do want it, whatever.

ST is for you – it only goes through if others signal for it.

If you’re a miner, you can signal no during the period, and you can write a blogpost on why you don’t.

You do not want CTV, and will not take yes for an answer.

I’ve written forkd software in 40 lines of python which guarantees you to be on a non-activating chain. Resist my evil6 fork!

If you’re a miner, you signal no during this period. You may want to optimize the forkd code for never building on the chain you don’t like.

7) Bitcoin Core is not Bitcoin

Bitcoin Core is a ‘reference client’ for the Bitcoin network, but it is not the job of any of the maintainers of Bitcoin Core to decide what happens with respect to consensus upgrades.

When I’ve previously asked maintainers for clarity on what ‘merge rubric’ they might apply to the CTV pull request, I’ve been effectively stonewalled with no criterion (even for things that were historically merged) and claims that soft-fork discussion is outside the purview of maintainership. To be clear, I’m not asking maintainers to merge, merely when they do make the decision to, what they are evaluating. The reticence to make clear guidelines around this was surprising to me, at first.

But then I understood: it isn’t the maintainer’s fault that they cannot give me any guidance, it’s how it must be.

The idea that Bitcoin Core might serve as the deciding body for consensus upgrades puts developers of Bitcoin Core into a dangerous position in the future, whereby various agents might wrongfully attempt to compel Core developers (e.g., using the legal system) to release a soft-fork client for whatever nefarious goal. Making it clear that soft-forks are released by independent efforts and adopted by the community at large is the only process we can take that keeps Bitcoin Core apolitical and unexposed.

We’ve seen in other communities what it looks like when lead devs exert too much influence over the protocol and reference clients directly. Not good. We do not want to have a similar precedent for Bitcoin.

While previous soft-forks have generally been released by Core, I have no qualms with leading by example for how future soft-fork development should be. And if Core wants to merge CTV and do a release with compatible parameters, they are welcome to, without such a release being driven by the project maintainers directly, but rather to maintain compatibility with the will of the community.

Thus, Alea Iacta Est.

Bonus: What do I do now?

I believe the community’s next steps are:

  1. Evaluate the software proposed above and find any bugs (claim 5.5 BTC Bounties?)
  2. Discuss vociferously through the next few months if BIP-119 should be activated or not (that means you should e.g. post publicly if you/your org endorses this particular path, cover it in your news org, etc).
  3. Before the end of July, Miners should signal if the speedy trial should succeed
  4. Before November, if Speedy Trial passes, then all users should ensure they upgrade to validate CTV
  5. If Speedy Trial fails, at least we were at the ball, and we can either try again next year, meaning CTV would be availble for use in at minimum 1.5 years, or we can re-evaluate the design of CTV against alternatives that would take more time to prepare engineering wise (e.g., more general covenants, small tweaks to CTV).

What is Jeremy Rubin going to do?

Well, at this point I am unemotional about any outcome. Judica, my startup, is focused on Bitcoin infrastructure that can be used to great impact with or without CTV, so I’ve explicitly positioned myself to be personally indifferent to outcome. I personally think that CTV is ready to go, and will deliver immense benefits to the network, so I’ll advocate for signalling for it.

However, in my advocacy, I’ll be careful to note that it’s not a must. All actors must decide if it’s in their own rational self-interest to have the soft-fork proceed.

But what about UASF?

Regrettably, no one has produced the ST compatible UASF code since last year, for various reasons. I understand the motives and tradeoffs of a UASF out the gate, but I still personally believe a UASF is best done as a follow-up to a ST, as I detailed in my mailing list post on the subject here.

Appendix: Parameter Check!

$ gdate -d@1651708800 -u
Thu May  5 00:00:00 UTC 2022
$ gdate -d@1660262400 -u
Fri Aug 12 00:00:00 UTC 2022

The below script simulates the passage of time and confirms that we are beginning at an expected mid-period time, and we are also ending near a mid-period time, given the assumed number of SIGNAL_PERIODS. This technique should guarantee with high certainty at least SIGNAL_PERIODS - 1, and repeating the simulation with up-to-date numbers as the signalling window progresses will produce more accurate forecasts.

import datetime
current_time = 1650301349
height_now = 732450
minutes_till_may_5th = 23457
height = int(height_now + minutes_till_may_5th/10.0)
print("Expected Height", height)
start_height = height + (2016 - (height % 2016))
print("Expected Start Height", start_height)
print("Start is mid period: ", (start_height-height)/2016.0)
minutes_from_now = 10*(start_height - height_now)
print("In This many Minutes", minutes_from_now)
print("In This many days", minutes_from_now/60.0/24.0)
stop_height = start_height + 2016*SIGNAL_PERIODS
print("Stopping at height", stop_height)
total_blocks = stop_height - height_now
end_time = (total_blocks - 2016/2)*10*60 + current_time
print("End of signalling at expected time", datetime.datetime.fromtimestamp(end_time))

active_height = 762048
secs_till_active = (active_height - height_now)*10*60
print("Active at", datetime.datetime.fromtimestamp(current_time + secs_till_active))
Expected Height 734795
Expected Start Height 735840
Start is mid period:  0.5183531746031746
In This many Minutes 33900
In This many days 23.541666666666668
Stopping at height 749952
End of signalling at expected time 2022-08-10 23:02:29
Active at 2022-11-09 23:02:29
  1. including roller skating along the beach of course… 

  2. in theory, was intended to be part of this… 

  3. I think defining such a process more formally would be great, but it’d be controversial. 

  4. technically, if you don’t touch it, this is a winner. A forced error, as Mike West notes, is where you still got your racquet to the ball. However, winners and forced errors are pretty similar in contrast to unforced errors. 

  5. probably following a thorough security review 

  6. sigh, if you really think I’m not working in good faith it’s a lost cause… 

Sapio Studio Tutorial for CTV Meeting

On Your Mark Get Set GOOOOOO

Other people have followed this tutorial with some success:

Common sticking points include:

  1. Not building latest sapio binary before starting
  2. Not building latest sapio-studio project before starting
  3. Not having the correct clang version (>=12)
  4. brew installing clang somewhere else (try /opt/homebrew, but also /usr/local/Cellar)

  1. Set up a signet node.

Build this branch

You’ll likely want settings like this in your bitcoin.conf too:

server = 1
  1. You’ll need to create a new wallet, if you’ve not done it before
./bitcoin-cli -signet createwallet sapio-studio-tutorial # If you've done this before fine
./bitcoin-cli -signet getnewaddress
  1. Get coins to your address / DM me for more

  2. Follow the install instructions on You can skip the the sapio-studio part / pod part and just do the Local Quickstart up until “Instantiate a contract from the plugin”.
  3. run cargo build –release from the root directory to build the sapio-cli.

Start Up Sapio Studio

  1. Get Yarn
  2. Install Sapio Studio
    git clone --depth 1 && cd sapio-studio
    yarn install
  3. Start react server
    yarn start-react

    Leave that running in tab

  4. Start electron server
    yarn start-electron

Once it launches this is what you should see: Go ahead and click on settings to configure the sapio cli for the first time. First, let’s configure our sapio-cli. Go head and do “Configured Here” and leave everything blank. Click on the select file and point sapio studio to your (freshly done) release build of sapio-cli.

Click Save Settings.

Now, you should see the following confirmation.

Next let’s configure your node. Clikc select file and point it to your .cookie file and then configure the node as Signet & set the RPC port to whatever you used. Save it.

Click test connection – it should tell you… something.

Navigate back to wallet. You should see some balance or something (assuming you did the faucet).

Next, click “Load WASM Plugin” and find your plugin-example/target/wasm32-unknown-unknown/debug/*.wasm files. Go ahead and pick the jamesob vault one.

Next, click Create new contract.

You’ll see a little applet for the JameVault. Look at that handsome guy!

Next, we need to create some Taproot keys for this – gonna need some command line action:

./bitcoin-cli -signet getaddressinfo $(./bitcoin-cli -signet getnewaddress "vault_project" "bech32m")

and then copy the witness program. do this twice – one for cold one for hot.

Now click on your applet and start filling it out. You can get new addresses by clicking on Bitcoin Node (just not keys yet) n.b. sometimes things are in btc, sometimes in sats. c’est la vie. finish up and click submit If all is right, you should see a contract get created! To actually create it, click on the parent txn and then sign it… and then broadcast it (with real money you’ll want to verify before doing this) it’ll now pop up in the mempool.

n.b. BUG ALERT: the viewer on the right hand side is a little glitchy. before playing with the psbts/txn signing logic, always close it and re-open by clicking on the txn you want. Now you can click on the begin redeem txn and broadcast and sign that one so it’s in the mempool Click on the updatable output, and you’ll see some options. click on spend hot, and fill out the form. Hit submit, and then hit recompile.

Be sure to pick an amount that is amount in input - fee (todo: make this smarter!). if all is well, you’ll see a hot spend.

if you picked a relative timelock > 1 block, try playing with the simulate tab :D.

If you navigate back to the CONTRACTS tab you can pull up contracts you previously made / recompilations with effects, even across sessions.

London Advancing Bitcoin Tutorial

Content for Following Along!

Other people have followed this tutorial with some success:

If you’re having a problem, see above log where people have had issues. Common problems include:

  1. Not building latest sapio binary
  2. Not having the correct clang version (>=12)
  3. brew installing clang somewhere else (try /opt/homebrew, but also /usr/local/Cellar)
  4. xargs not liking something (working to debug it, you can open the JSONs with vi and do by hand some of the steps).

  1. Install JQ (json manipulating tool) if you don’t have it / other things needed to run a bitcoin node.
  2. Set up a signet node.

Build this branch

You’ll likely want settings like this in your bitcoin.conf too:

# generate this yourself                                                                                                                    rpcauth=generateme:fromtherpcauth.pyfile     

Get coins / DM me

  1. Follow the install instructions on You can skip the the sapio-studio part / pod part and just do the Local Quickstart up until “Instantiate a contract from the plugin”. You’ll also want to run cargo build –release from the root directory to build the sapio-cli.

  2. Open up the site
  3. Run sapio-cli contract api –file plugin-example/target/wasm32-unknown-unknown/debug/sapio_wasm_plugin_example.wasm
  4. Copy the resulting JSON into the RJSF site
  5. Fill out the form as you wish. You should see a JSON like
      "context": {
     "amount": 3,
     "network": "Signet",
     "effects": {
       "effects": {}
      "arguments": {
     "TreePay": {
       "fee_sats_per_tx": 1000,
       "participants": [
           "address": "tb1pwqchwp3zur2ewuqsvg0mcl34pmcyxzqn9x8vn0p5a4hzckmujqpqp2dlma",
           "amount": 1
           "address": "tb1pwqchwp3zur2ewuqsvg0mcl34pmcyxzqn9x8vn0p5a4hzckmujqpqp2dlma",
           "amount": 1
       "radix": 2

    You may have to delete some extra fields (that site is a little buggy).

Optionally, just modify the JSON above directly.

  1. Copy the JSON and paste it into a file ARGS.json
  2. Find your sapio-cli config file (mine is at ~/.config/sapio-cli/config.json). Modify it to look like (enter your rpcauth credentials):
      "main": null,
      "testnet": null,
      "signet": {
     "active": true,
     "api_node": {
       "url": "",
       "auth": {
         "UserPass": [
           "YOUR RPC NAME",
     "emulator_nodes": {
       "enabled": false,
       "emulators": [],
       "threshold": 1
     "plugin_map": {}
      "regtest": null
  3. Create a contract template:
    cat ARGS.json| ./target/release/sapio-cli contract create  --file plugin-example/target/wasm32-unknown-unknown/debug/sapio_wasm_plugin_example.wasm  | jq > UNBOUND.json
  4. Get a proposed funding & binding of the template to that utxo:
cat UNBOUND.json| ./target/release/sapio-cli contract bind | jq > BOUND.json
  1. Finalize the funding tx:
cat BOUND.json | jq ".program[\"funding\"].txs[0].linked_psbt.psbt" | xargs echo | xargs -I% ./bitcoin-cli -signet utxoupdatepsbt % |  xargs -I% ./bitcoin-cli -signet walletprocesspsbt % | jq ".psbt" | xargs -I% ./bitcoin-cli -signet finalizepsbt % | jq ".hex"
  1. Review the hex transaction/make sure you want this contract… and then send to network:
    ./bitcoin-cli -signet sendrawtransaction 020000000001015e69106b2eb00d668d945101ed3c0102cf35aba738ee6520fc2603bd60a872ea0000000000feffffff02e8c5eb0b000000002200203d00d88fd664cbfaf8a1296d3f717625595d2980976bbf4feeb
  2. Send the other transactions:
cat BOUND.json| jq .program | jq ".[].txs[0].linked_psbt.psbt" | xargs -I% ./target/release/sapio-cli psbt finalize --psbt %  | xargs -I% ./bitcoin-cli -signet sendrawtransaction %

Now what?

  • Maybe load up the Sapio Studio and try it through the GUI?
  • Modify the congestion control tree code and recompile it?
  • How big of a tree can you make (I did about 6000 last night)?
  • Try out other contracts?

© 2011-2021 Jeremy Rubin. All rights reserved.