An independent Bitcoin node implementation

Block space has a price.

Since 2023, files stored in Bitcoin's pay a quarter of the price of the space they take. Bitcoin Tolerant charges them their real weight, at a rate read from every confirmed block. Payments never pay more. Nothing is filtered. Every number is public.

Phase 1 · Observe. It measures and publishes the price. It rejects no transaction.

One block · 4,000,000 · each cell is stored bytes, its fill is the share of full price it pays

72% of the stored bytes in this block are data. In this view, data pays 50% of the block's price.

Transaction body4 WU per byte · always full price
Signatures1 WU per byte · discount kept, by design
Data in the witness1 WU today → 4 WU in Tolerant
Data share of stored bytes72%
Data share of what is paid50%

The same bytes sit on every node's disk in both views. With the witness discount, data fills most of the block and pays a fraction of it.

Witness space

A discount built for signatures.

(2017) counts a witness byte as one weight unit instead of four. That made signatures cheap, as intended. Since 2023 the same discount also covers files placed in the witness, and every full node stores them at a quarter of the price.

Bitcoin Tolerant removes the discount for data, and only for data. It reads script structure, never meaning, so it never needs to judge content.

KeptSignatures keep their discount. Payments are never multiplied.
ChargedData in the witness pays +3 WU per byte, the exact gap to full weight.
Untouched already pays 4 WU. It is never charged twice.
Weight per byteview: witness discount
Body byteinputs, outputs, amounts
4 WU
Signature bytewitness, as designed
1 WU
Data byte in witnessinscription envelope
1 WU
OP_RETURN byteoutput, provably unspendable
4 WU
economic_vbytes = vbytes + witness_data × 3/4
Dynamism

A price that moves with the chain.

The reference rate is the fee actually paid per vbyte in recently confirmed blocks. It is read from each block's , never from this node's , so it describes the network rather than one node's view of it.

Each block counts half as much every 6 blocks, about an hour. One odd block cannot move the price much, and a fee spike shows up within the hour.

Pruning: the price forgets on schedule.

There is no stored history. A running total absorbs each new block and lets older ones fade. After 2,016 blocks, one difficulty period, a block weighs 0.5336, which is nothing. The price is read as each block connects, so it needs no block archive.

Last 36 blocks · fee rate and weightsnapshot

Newest block 970,292 paid 1.19 sat/vB. The reference rate is 1.07 sat/vB.

Newest100%
6 blocks50%
12 blocks25%
36 blocks1.6%
144 · 1 day6·10⁻⁸
2,016 · pruned0.5³³⁶
Fairness and balance

One rate for every byte.

Every transaction is measured in and priced at the same reference rate. A payment's economic size is its real size, so sending money never costs extra. Data that pays its weight is treated like anything else.

BalanceWhen nobody competes for space, the price falls to the floor for everyone, data included. Scarcity is never invented.
AsymmetryUnder-pricing leaks a subsidy. Over-pricing during a quiet market would be just as dishonest.
BatchingOutput count is never priced. Batched payments, the most efficient use of space, pay nothing extra.
Price a transactionillustrative sizes
Type
Market
Data bytes
50,000 B
Real vsize12,620
Economic vbytes50,120
Rate4.1 sat/vB
Pays with discount51,742 sat
Real cost of space205,492 sat
hidden-subsidy

Hidden subsidy of 153,750 sat. The data takes the space and pays a quarter of it.

Video

Governance by savings, in 2 minutes.

2:41

Why Bitcoin Core closes the door to AI agents — and how locking savings opens an open proposal queue without free spam.

Transcript

Hoy, cualquiera puede programar con inteligencia artificial. Entonces, ¿quién decide qué código entra a Bitcoin? En Bitcoin Core, lo decide un grupo pequeño de mantenedores. No hay votación. Hay confianza: en quién eres y en lo que hiciste antes. Ese modelo funcionó cuando escribir código costaba semanas. Pero hoy la confianza en Core está en duda, y el software se mueve más rápido que ese modelo. El código humano también trae errores y relleno. El slop no es exclusivo de la IA. Y una buena reputación pasada no garantiza un buen comportamiento futuro. Ante la inteligencia artificial, Core respondió cerrando la puerta. Su política dice que los agentes no deben abrir propuestas de código. Bitcoin Tolerant propone lo contrario: abrir la puerta. Cualquiera, persona o agente, puede proponer mejoras a Bitcoin. Queremos que hacer vibe coding sobre Bitcoin sea fácil. Claro, con la puerta abierta llega el spam. Miles de propuestas automáticas que nadie alcanza a revisar. Por eso, la entrada se paga con ahorro. Para proponer, bloqueas bitcoin en Safe Hodling, nuestro sistema de ahorro programado. No lo gastas ni se lo das a nadie. Solo lo inmovilizas un tiempo. Mil propuestas exigen mil veces el ahorro. El spam deja de ser gratis. El compromiso se mide en bitcoin por día: cuánto bloqueas y por cuánto tiempo. Un bitcoin durante un año pesa lo mismo que doce bitcoin durante un mes. Pero el ahorro solo te da la entrada. Para avanzar necesitas tres cosas. Ahorro, para que te escuchen. Reputación social: otras personas con ahorro bloqueado respaldan tu propuesta. Y una buena propuesta, bien explicada, que convenza a la comunidad. El dinero compra atención, no decisiones. El consenso de Bitcoin no se toca. Y cada operador elige qué versión corre. Si sabes construir, ahora puedes proponer. Ahorra, construye y convence. Bitcoin Tolerant. Propón con ahorro. Conoce más en bitcointolerant.com.

Today, anyone can code with artificial intelligence. So who decides what code goes into Bitcoin? In Bitcoin Core, a small group of developers decides. There's no vote. There's trust: in who you are and what you've done before. That model worked when writing code took weeks. But today, trust in Core is in doubt, and software moves faster than that model. Human code also brings bugs and filler. Slop isn't exclusive to artificial intelligence. And a good past reputation doesn't guarantee good future behavior. Faced with AI, Core responded by closing the door. Its policy says agents must not open code proposals. Bitcoin Tolerant proposes the opposite: open the door. Anyone, person or agent, can propose improvements to Bitcoin. We want vibe coding on Bitcoin to be easier. And of course, with the door open, spam arrives. Thousands of automated proposals nobody can review. That's why entry is paid with savings. To propose, you lock bitcoin in Safe Hodling, our programmed savings system. You don't spend it or give it to anyone. You just lock it up for a while. A thousand proposals require a thousand times the savings. Spam stops being free. Commitment is measured in bitcoin-days: how much you lock, and for how long. One bitcoin for a year weighs the same as 12 bitcoin for roughly a month. But savings alone get you in. To move forward, you need three things. Savings, so you get heard. Social reputation: other people with locked savings backing your proposal. And a good proposal, well explained, that convinces the community. Money buys attention, not decisions. Bitcoin's consensus stays untouched, and each operator chooses which version to run. If you know how to build, now you can propose. Save. Build. Convince. Bitcoin Tolerant: propose with savings. Learn more at bitcointolerant.com.

Governance

Anyone can propose. Savings pays the entry.

Draft · not live

Design draft for technical review (October 2026). Not live and not in any release.

The problem

Bitcoin Core decides through a small group of maintainers and trust in past reputation. That worked when writing code took weeks.

With AI, code is nearly free. Review is now the scarce resource. Core's doc/AI_POLICY.md responds by closing the door to autonomous agents.

Human code produces slop too, and a good past reputation does not guarantee good future behavior.

An open door with a toll

People and AI agents can open proposals. To enter the review queue, the author locks BTC in Safe Hodling (non-custodial: nothing is spent or given to anyone; it is only locked for a while).

One thousand proposals require one thousand times the savings. Spam stops being free.

Money buys attention, not decisions.

The unit: BTC·day

E = BTC × min(days_remaining, 365). Days are capped at 365.

1 BTC × 365 days ≈ 12 BTC × 30 days — equal area, equal weight.

365.00 BTC·days

Tier

Required R: 0.90 BTC·days

Weight is linear on purpose: splitting BTC into many identities gives no advantage.

Three keys

Savings

So you get heard — BTC locked in Safe Hodling.

Social reputation

Other people with locked savings back your proposal. Reputation multiplies weight (0.5× to 2.0×). It is earned only through a verifiable history in git and on-chain, and it cannot be bought or transferred.

A good proposal

Well explained, so it convinces the community.

Tiers

Infrastructure · tier 2

wallet, RPC, P2P, build, GUI

Base R
0.9 BTC·day
Equivalent to
0.01 BTC × 90 d
Merge
2 of N

Block space policy · tier 3

tolerant_defaults.cpp, pricing, mempool

Base R
18 BTC·day
Equivalent to
0.1 BTC × 180 d
Merge
3 of N + weighted ACKs

Consensus and security · tier 1

validation, script interpreter, CVEs

Base R
Not applicable
Equivalent to
—
Merge
Upstream only

Base values are initial, for discussion. Not calibrated against real data.

Consensus is never touched: no amount of BTC opens the door to consensus. Each operator chooses which version to run.

Three layers

  1. 1AdmissionBTC locked
  2. 2PriorityBTC·days × reputation
  3. 3MergeGreen CI + reproducible builds + tier multisig of maintainers

Savings never merge code by themselves.

How it is verified
  • Lock (today). Safe Hodling savings vault: P2WSH / miniscript with a CLTV unlock height. Non-custodial for the user’s branch.
  • Credential (Phase 1). PoSD — Proof of Sat-Days — issued by Safe Hodling and bound to a proposalRef.
  • Node check (Phase 2 goal). A Tolerant node will verify the lock itself: gettxout, revealed descriptor, and a BIP-322 control proof. Proposed RPC: verifygovlock. Not live yet.
  • Penalty. Spam lowers reputation; the next proposal needs more BTC·days. Early cosign and loans must not count toward a governance toll (Phase 1+ vaults).

If you know how to build, now you can propose. Save. Build. Convince.

Transparency

Every number is public.

The deliverable is a price, not a filter. The node logs every block's rate and every data transaction's verdict, and serves both over . The price stays visible even when it condemns nothing.

All Tolerant logic lives in its own files. Read them, test them, run them.

$ bitcoin-cli gettolerantpricing
{ "reference_rate": 4.107, "floor": 0.1,
  "reliable": true, "blocks_observed": 111,
  "half_life_blocks": 6 }

$ bitcoin-cli gettolerantpricing <txid>
{ ..., "tx": {
  "classification": "WITNESS_DATA",
  "vsize": 210, "economic_vsize": 588,
  "fee": 40, "required_fee": 59,
  "verdict": "hidden-subsidy" } }
Live

Phase 1 · Observe

Measure, classify, publish the price. Rejects nothing.

Testing · v32 RC

Phase 2 · Mining

Local block templates only. Default off.

Planned

Phase 3 · Relay

Optional relay policy. Deliberately last.

Draft · not liveGovernance

Future · Governance by savings

Design draft for technical review (October 2026). Not live and not in any release. Anyone can propose; savings pays the entry.

Identity

Born from Bitcoin Core. Its own implementation.

Like other Bitcoin implementations, Bitcoin Tolerant started from Bitcoin Core's codebase. It has its own name, its own release line and its own pricing layer. It follows upstream releases for validation and security fixes, and speaks the same consensus as every other node.

Bitcoin Core · codebasev29v30v30.3Bitcoin Tolerantown name · own releases · tolerant_* filesupstream fixes merged
Bitcoin Core v30Bitcoin KnotsBitcoin Tolerant
Arbitrary dataRelayed, discount intactFilteredPriced at real weight
BasisPermissiveContent judgmentEconomic cost only
Quiet marketCheapStill filteredCheap, no penalty
Price visibleNoNoYes, over RPC and logs
Run it

Five commands to a node that prices space.

It measures and publishes the price. It rejects nothing. With -tolerantv2=0 the pricing layer is never built.

Needs the same build tools as Bitcoin Core: a C++ compiler, CMake, Boost and libevent. macOS · Linux · Windows

Outside its reach

  • Consensus rules
  • Chain selection
  • Valid blocks
  • What data means
git clone -b tolerant-core https://github.com/HansLove/bitcoin-tolerant.git && cd bitcoin-tolerant
cmake -B build -DENABLE_IPC=OFF
cmake --build build
build/bin/bitcoind -tolerantv2=1 -debug=tolerant
build/bin/bitcoin-cli gettolerantpricing
-tolerantv20

Turns Bitcoin Tolerant pricing on.

-tolerantreferenceblocks6

Half-life of the reference rate, in blocks (1–2016).

-debug=tolerantoff

Logs each block's rate and each data transaction's verdict.

Independent and unfunded. If a public price for block space matters to you, support the work.

To collaborate, share an idea, or discuss a partnership, contact Alicia Lombardi on Telegram.

bc1q6n0h05956f4hlnwdppptjh6aj8wemymyns7r5c
QR code with the Bitcoin donation address