Language

Choose a language

Topics · Explainer

COMPUTE — options on the network difficulty, settled by the chain header itself

Infographic: COMPUTE — options on the network difficulty, settled by the chain header itself

On TensorCash, the work that secures the chain is real AI inference: the network is designed so that running model forward passes is the mining, not a byproduct of it. As that turns the network into a standing signature — a proof that real AI inference happened — its difficulty takes on a second meaning. Difficulty is simply how much inference capacity is competing to secure the chain, which makes it one of the clearest on-chain readings of inference-market activity and AI-compute growth.

Which raises a question: what if that reading could be tokenised — turned into a plain, transferable asset that lives on the chain itself, whose on-chain settlement tracks how the difficulty index moves? A miner could use it to offset their own exposure to difficulty; anyone could hold a token tied to the same index, without an intermediary and without leaving the chain.

That is now live on mainnet. This post is a technical walk-through of exactly what was issued and how it settles — every number below is on-chain, and you can verify each one yourself.

First, the shape of the thing. The COMPUTE Index — the network’s mining difficulty read as AI-compute throughput, defined in full just below — over the chain’s history so far:

Network difficulty as throughput — target versus implied, log scale over time Chart on a logarithmic tokens-per-second y-axis against wall-clock time, from 2026-06-28 to 2026-07-20, the same view the explorer shows. The solid maroon step line is the target throughput implied by difficulty, climbing in retarget steps from about 197 to about 129,857 tok/s; small amber dots mark the six difficulty retargets on it. The dotted grey line is the throughput implied by observed block times, which runs higher (up to about 943,883 tok/s) because blocks arrived faster than the 600-second target, so difficulty retargets upward to follow. Decade gridlines at 100, 1K, 10K, 100K and 1M. The issuance point on 19 July 2026, target about 32,464 tok/s, is marked. Difficulty as throughput: target vs implied (log scale) Read from the chain — the same view the explorer shows; x-axis is wall-clock time. 100 1K 10K 100K 1M tok/s (log scale) 2026-06-28 07-06 07-11 07-16 2026-07-20 block time (wall-clock) issued 19 Jul 2026 target ≈ 32,464 tok/s target (difficulty) implied (block times)

Network difficulty as throughput (target, solid) against the rate implied by observed block times (implied, dotted), on a log scale — the same view the explorer shows. Blocks ran ahead of the 600-second target, so implied sits above target and difficulty retargets upward to follow. It’s a factual record of the index over time, nothing more.

The underlying: the COMPUTE Index

Every TensorCash block is minted by a real forward pass, so the network’s mining difficulty has an interpretation no hash-based chain’s does: it tracks the AI compute committed to the network. The COMPUTE Index makes that explicit. It reads the difficulty as an expected throughput in consensus tokens per second:

COMPUTE Index  X  =  GetBlockProof(target) · 256 / 600   (tokens/sec)

target decodes from the compact difficulty field (nBits) in the block header; 600 is the consensus target block spacing in seconds. That’s the whole definition, and it’s re-derivable by anyone from a header. These series were issued on mainnet on 19 July 2026; at that point (mainnet height ~13,466, nBits 0x1f00dc7f) the index was ≈ 32,464 tok/s. That is a historical reading at issuance — the index has moved since, as the path below shows.

The COMPUTE Index over recent blocks

The index is not a smooth line: difficulty retargets in steps, so the reading holds flat between retargets and then jumps. The chart below is the actual on-chain series across the blocks around issuance, read straight from each header — a factual record of where the index has been, nothing more. (It is distinct from the strike-relative settlement diagrams further down: this is the index’s own path over time; those describe how a vault is split at maturity.)

The COMPUTE Index over recent blocks Line chart of the COMPUTE Index in tokens per second, read from the chain, against block height from 11,740 to 14,140. The index is a step function set by difficulty retargets: about 8,116 up to block 12,080, about 32,464 from block 12,100, and about 129,857 from block 14,120. The issuance point at block 13,466 sits on the 32,464 plateau. The COMPUTE Index over recent blocks Read from the chain — a step function set by difficulty retargets. 0 50K 100K COMPUTE Index (tok/s) 12,000 12,500 13,000 13,500 14,000 block height ≈ 8,116 ≈ 129,857 tok/s issued 19 Jul 2026 index ≈ 32,464 tok/s

The anchor: COMPUTE is a rulebook, not money

The series are sponsored by a root asset with the ticker COMPUTE — a difficulty anchor. It is deliberately a zero-economic-claim namespace asset: 100 indivisible governance units, a 51% quorum, and a public charter committed on-chain (icu_plain_commit, readable by anyone with geticupayload). COMPUTE is not the money in these contracts; it’s the namespace and the amendable rulebook under which difficulty series are issued. Holding a governance unit is a vote on the charter, nothing more.

  • Ticker COMPUTE, registry id f9b30f21fd6b7a13dff6497ba7ac3f62a3270c09a5adbed274aad4c85161a1a8view on the explorer.
  • Registered in tx baa2c77f25ede0a7689558db9f156813c496726d38bdfcf6b3c7a1d7a695248d; the 100 governance units minted in 559554e098e3822484ecd61c3632802c937940411bb518fe4e7ed3a534fee9f4.

The term sheet lives on-chain

Every asset registered on TensorCash can bind a canonical term sheet to itself, persisted on-chain and readable by anyone. The asset carries a hash-committed ICU payload — you read it with geticupayload or straight from the explorer — recorded in the registry as a commitment (icu_ctxt_commit) and immutable except by a governed rotation. For these option series the binding is even tighter: the series’ canonical id is the tagged hash of its descriptor, so a cold verifier recomputes the id straight from the terms and confirms nothing was swapped underneath it. (The COMPUTE anchor’s own id is a fixed tagged hash, SHA256("TSC-COMPUTE-ANCHOR/v1"), and its charter is committed separately in the same readable way.)

For these series the payload holds a machine-readable descriptor (the TSC-ICU-OPTSERIES-1 band: strike, direction, fixing height, leverage) alongside a human-readable summary. The contract terms therefore live in the chain, not in an off-chain PDF that someone has to host and honour.

This is not special to options. Any registered asset can commit a canonical term sheet or charter to itself this way, immutable except by a governed rotation — the COMPUTE anchor does exactly that with its public charter. Options are just one shape of term sheet the mechanism can carry.

What “intrinsic” means

These series settle with no oracle, no issuer publication, and no off-chain data — an intrinsic settlement. At the fixing block height, consensus does the work itself:

  1. Read the block header’s own nBits.
  2. Decode the target and compute the realized COMPUTE Index X.
  3. Enforce the split of each contract by the rule committed in the series descriptor.

Because the only input is the header, the result is not a claim you trust — it’s a computation every full node repeats and agrees on. Re-run it a year later on the same block and you get the same number. That is the whole point of an intrinsic series: the contract’s terms are consensus rules.

The covenant-vault lifecycle

Each contract is issued as a set of lots, and every lot is a covenant vault holding 1 TSC. The token that represents a lot is an ordinary fungible, transferable, consensus-native asset — one token is one claim on one vault’s pot.

The lifecycle has five movements, all enforced by the protocol:

  1. Register. The series descriptor — strike, direction, fixing height, leverage — is committed on-chain first, in the registration transaction. That commitment is what defines the series: the asset’s canonical id is derived from the descriptor, so the terms and the identity are one and the same.
  2. Issue. The writer self-mints all the lot tokens and funds one 1-TSC vault per lot against the already-registered series.
  3. Transfer. The tokens change hands like any other consensus asset. Whoever holds a token holds that lot’s pot claim.
  4. Settle. At maturity (the fixing height, plus a short burial delay), anyone can submit the settlement transaction. Consensus computes the realized index from the header and splits each vault between the holder pot and the writer’s reclaim according to the payoff rule below.
  5. Redeem / reclaim. A holder sweeps their pot by redeeming a token; the writer can reclaim unsold lots early via a buyback path.

No step depends on the issuer being online, honest, or solvent at maturity — the vault already holds the collateral, and the split is a rule, not a promise.

The two live series

Both series fix at the same height and share the same shape (leverage λ = 4, 100 lots × 1 TSC = 100 TSC collateral, descriptor version 2). They differ only in direction and strike.

COMPUTE.V26C2MCOMPUTE.V26P100K
Directioncallput
Strike K2,000,000 tok/s100,000 tok/s
Fixing height28,41628,416
Settle lock28,51628,516
Leverage λ44
Collateral100 lots × 1 TSC100 lots × 1 TSC
Registry id8a31784f9ebe8a94250d3c6c7702196e0f9814c413bd317fc892f18a5ccbce2459080a03b813589fa2f0e6cc2f17e5eb386dd0abc15f89b7763c1e17218a7cbb
Canonical series id24cecb5c8af192c87f31bd13c414980f6e1902776c3c0d25948abe9e4f78318abb7c8a21171e3c76b7895fc1abd06d38ebe5172fcce6f0a29f5813b8030a0859
Registration txe220a481cee7e9335917ac84824f7383421e6c34a1d730dcf0c4d5d6e820e4e19f8c3b501be95b89b9c78b6dc477f5e15068a620d1c8b30e35a2f6a3babdfcfa
Issuance txe0a309feaad7ccaef11e7bedc3e4a5841a88493aec93e41558f47493ccaddcf59148173679d477f60ac1e55c78a4c30d5c51865ab77bdab1d3ec21709d0c2d45

The registry id is the one to query by — it’s what wallet RPCs and the explorer show. The canonical series id is the same value byte-reversed (the internal, descriptor-derived form); the two are just two orderings of the same 32 bytes. You can also just look either series up by its ticker.

View on the explorer: COMPUTE.V26C2M (call) · COMPUTE.V26P100K (put).

Fixing height 28,416 lands around 31 October 2026, 12:00 UTC at 600-second spacing — but the height is authoritative, not the date.

How the covenant computes the split

At maturity, each 1-TSC vault is divided between the holder pot and the writer’s reclaim as a function of the realized index X. Per 1-TSC token, with strike K and leverage λ = 4:

  • Call (V26C2M): routes to the holder pot only when X > Kmin(1, 4·(X − K)/K) TSC; 0 when X ≤ K.
  • Put (V26P100K): routes to the holder pot only when X < Kmin(1, 4·(K − X)/K) TSC; 0 when X ≥ K.

Whatever the rule does not route to the holder pot returns to the writer, so the two shares always sum to the vault’s 1 TSC. The min(1, …) cap means the holder pot fills the whole vault once the index is 25% past the strike (because 4 × 0.25 = 1).

The two diagrams below are pedagogy — they show how the covenant computes that split, plotting the TSC routed to the holder pot against the realized index. They are not a forecast and not an offer; they describe a settlement rule. For orientation, each diagram marks the index’s value at issuance (≈ 32,464 tok/s, on 19 July 2026) — a historical reading, shown only to locate that day’s value on the rule.

Call settlement split for COMPUTE.V26C2M Step curve of TSC routed to the holder pot versus the realized COMPUTE Index. It is 0 for an index at or below the strike of 2,000,000 tokens per second, rises linearly to a full 1 TSC across 2,000,000 to 2,500,000, and stays at 1 above 2,500,000. The index at issuance, about 32,464 tokens per second, sits far left in the flat 0 region. Settlement split — COMPUTE.V26C2M (call, K = 2,000,000 tok/s) y = TSC the covenant routes to the holder pot at realized index X; the remainder returns to the writer. 0 0.5 1 TSC to holder pot / vault 0 1.0M 2.5M COMPUTE Index X (tokens / sec) K = 2,000,000 above the curve → writer reclaim curve reaches full 1 TSC at X = 2,500,000 (K + 25%) holder pot index at issuance ≈ 32,464 tok/s (flat region: split routes 0 to the holder pot)
Put settlement split for COMPUTE.V26P100K Step curve of TSC routed to the holder pot versus the realized COMPUTE Index. It is a full 1 TSC for an index at or below 75,000 tokens per second, falls linearly to 0 across 75,000 to the strike of 100,000, and stays at 0 above 100,000. The index at issuance, about 32,464 tokens per second, sits in the flat full-1-TSC region. Settlement split — COMPUTE.V26P100K (put, K = 100,000 tok/s) y = TSC the covenant routes to the holder pot at realized index X; the remainder returns to the writer. 0 0.5 1 TSC to holder pot / vault 0 50K 125K COMPUTE Index X (tokens / sec) K = 100,000 full 1 TSC for X ≤ 75,000 (−25%) below the curve → holder pot writer reclaim index at issuance ≈ 32,464 tok/s (flat region: routes 1 TSC to holder pot)

On the day of issuance (19 July 2026) the index read ≈ 32,464 tok/s. Where that value falls on each curve is simply arithmetic on the rule above; the diagrams place the issuance tick for reference. What the index does between now and the fixing height, and therefore how any given vault ends up split, is not something this post predicts — the settlement is whatever the header says at height 28,416, computed by every node.

Reading the name

The ticker encodes the descriptor in a fixed convention:

COMPUTE.<MonthCode><YY><C|P><Strike>
month codes:  F G H J K M N Q U V X Z   (V = October)

So V26 = October 2026, C2M = call at strike 2,000,000, P100K = put at strike 100,000. The month label and the display strike are human conveniences; the authoritative terms are the strike_nbits and fixing_height committed in each series’ descriptor. (For the two series above, strike_nbits is 503551042 for the call and 508007729 for the put.)

Check it yourself

Nothing here asks for trust. The charter and each series descriptor are readable on-chain with geticupayload, and the realized index is just arithmetic on a block header you already have — GetBlockProof(target) · 256 / 600. Pull the descriptors, recompute the index at the fixing block when it arrives, and confirm the split the covenant enforces is exactly the one the rule specifies.

Don’t trust — verify. The number that settles the contract was never something a third party had to report. It was in the header the whole time.


This is an informational and explanatory post only. Nothing on this page constitutes an offer, solicitation, or invitation to buy or sell any token, security, or other financial instrument. Neither this website nor any TensorCash social channel is a venue for arranging, negotiating, or discussing over-the-counter (OTC) transactions or sales.

Authored pseudonymously by Imosuke Takakuni.

Our mission

TensorCash turns useful AI work into open money.

Out of the potato age, as our whitepaper says…

We believe people deserve a cheaper, more efficient financial system, and fairer AI that works for everyone. TensorCash makes AI work verified and verifiable. Verification gives AI a face: proof of which model did the work, what it saw, and the rules it followed. That lets anyone confidently buy or sell AI work at the most efficient price. The result is more accessible, more sustainable AI, powering a new generation of financial systems. Today's currencies are the potatoes: antiquated, expensive to move, and trapped behind fee-takers. TensorCash is a more efficient way to move and store value — one that harnesses AI's computational power for everyone while pushing the control outward instead of concentrating it.

— Imosuke Takakuni

About us

Imosuke Takakuni is a pseudonym. The Japanese name is both a tribute to Satoshi Nakamoto and a nod to Potato Land — the parable from our whitepaper. The mission is bigger than any one contributor; it should outlast personalities and charisma. Decentralisation works for everyone, or it doesn't work at all. We want everyone to participate in TensorCash as equals.

Open the mission page →

Get involved

How to get TSC

TensorCash is not selling TSC. The project is not running a token sale, pre-sale, ICO, IDO, or official investment round. New TSC enters circulation through active mining. You can mine it, receive it peer-to-peer from someone who already has it, or run the wallet and be ready for mainnet.

TensorCash is not running an official sale. Do not send money to anyone claiming to sell official allocations.

Get involved

Run the Core wallet

The practical first step is to run TensorCash Core, create a wallet, and learn the RPC surface. Today the public guide starts with regtest so you can create addresses and move coins locally before touching mainnet funds.

Get involved

Donate

No mainnet donation address is published yet. For testing only, the TensorCash testnet address below was generated from the running Core wallet; do not send mainnet funds to it.

Get involved

Spread the word

The shortest useful explanation is: TensorCash turns useful AI work into open money. Share the mission page, the flagship whitepaper, or the Get involved page with one person who cares about cheaper financial rails, fairer AI, or open infrastructure.

TensorCash turns useful AI work into open money.

Get involved

Emission schedule

Bitcoin set the baseline: block rewards only, no discretionary minting, and an exact integer subsidy total of 20,999,999.97690000 BTC. TensorCash keeps the fixed-supply discipline and changes the release curve for a compute-mined network; the implemented recurrence ends at 21,184,153.03530240 TSC.

Supply over blocks

Total subsidy issued

Exact integer subsidy rules from Core: Bitcoin halvings against the TensorCash epoch-decay schedule, shown through the first 6,000,000 blocks.

Horizon
...
BTC @ 6M
...
TSC @ 6M
...
BTC and TSC total subsidy over block count At 6,000,000 blocks, Bitcoin has issued 20,999,999.92710000 BTC and TensorCash has issued 20,979,987.36365355 TSC under the implemented epoch-decay schedule.
Block 0
BTC supply 0 BTC
TSC supply 0 TSC
BTC: 50 BTC, 210,000-block halvings TSC: 715 TSC, 715-block epoch, reward x 3/5, capped epoch length