Reference
The vocabulary, defined.
30 terms we use across the desks, each defined in one sentence and then explained. Grouped the way the coverage map groups them — by which domains the term sits between.
AI & Data
Models, inference and the hardware under them- Agentic AIAI-02Software that uses a language model to decide a sequence of actions and then carries them out against real systems, rather than only returning text.
- Bandwidth DePINAI-05A DePIN category in which participants share the unused capacity of an ordinary internet connection, and buyers pay for distributed access to the public web.
- ChunkingAI-03Splitting documents into passages before they are embedded and indexed — the step that decides what a retrieval system is able to find at all.
- DePINAI-05Networks that coordinate privately owned physical hardware — GPUs, storage, wireless capacity — into a shared marketplace, with settlement on-chain.
- DistillationAI-04Training a smaller model to reproduce the behaviour of a larger one, so the capability survives at a fraction of the cost to run.
- Fine-tuningAI-04Continuing to train an existing model on a narrow dataset so it performs a specific task better than prompting alone can achieve.
- Speculative decodingAI-04An inference technique where a small draft model proposes several tokens at once and the large model verifies them in a single forward pass, producing identical output in fewer passes.
Cybersecurity
Defence, exploitation and what gets prioritised- Formal verificationSEC-02Proving mathematically that code satisfies a stated property for every possible input, rather than testing that it behaves correctly on the inputs someone thought of.
- KEVSEC-04A catalogue maintained by the United States Cybersecurity and Infrastructure Security Agency listing vulnerabilities confirmed to be exploited in the wild.
- Supply-chain attackSEC-01An attack that compromises a dependency, build tool or update channel, so the victim installs the attacker's code through a route they already trust.
- Threat modelSEC-01A written statement of who the adversary is, what they can do, and what they are trying to reach — the document that makes every other security decision checkable.
- Zero TrustSEC-01A security model that grants no implicit trust from network position, and authenticates and authorises every request as if it came from an untrusted network.
FinTech & Web3
Settlement, market structure and tokenised claims- Basis tradeFIN-04Buying an asset and selling a future against it to capture the gap between the two prices, which converges to zero at expiry.
- CustodyFIN-01Holding the keys that control an asset on someone else's behalf — which in practice means holding the asset, whatever the account statement says.
- ePBSFIN-03A change to Ethereum that writes the existing split between block proposers and block builders into the protocol itself, instead of leaving it to software outside the protocol.
- MEVFIN-03The value a block producer can capture by choosing which transactions to include and in what order, over and above fees and issuance.
- Perpetual futuresFIN-01A derivative that tracks a spot price with no expiry date, held to that price by a periodic payment between long and short holders called funding.
- RWAFIN-04Assets with an off-chain legal claim behind them — treasury bills, money-market funds, receivables — issued and settled as tokens.
- SequencerFIN-03The party that decides the order of transactions on a rollup before they are published to the chain underneath — usually one party, usually the operator.
- StablecoinFIN-05A token designed to hold a constant value against a reference currency, most often by holding reserves in that currency.
- Tokenised depositFIN-04A commercial bank deposit recorded on a ledger rather than only in the bank's own books — legally the same claim, with a different settlement mechanism.
Model integrity
Where AI and security meet- Agent sandboxSEC-01An enforced execution boundary around a program that reads untrusted input, limiting what it can reach rather than what it is allowed to intend.
- Prompt injectionNEX-01An attack that places instructions in content a language model reads, causing the model to follow the attacker's directions instead of the operator's.
- zkMLSEC-03Producing a cryptographic proof that a specific model produced a specific output from a specific input, verifiable without re-running the model or seeing its weights.
Autonomous execution
Where AI and finance meet- CFMMNEX-02An exchange that holds reserves of two or more assets and quotes prices from a formula over those reserves, rather than from an order book.
- Concentrated liquidityNEX-02Liquidity provision where capital is committed to a chosen price range rather than the whole curve, earning more fees while the price stays inside it and nothing when it leaves.
- IntentAI-02A statement of the outcome a user wants, signed and published for others to fulfil, rather than an instruction specifying how to achieve it.
Verifiable settlement
Where security and finance meet- Smart contract auditSEC-02A review of deployed or deployable contract code for defects that let value be taken, frozen or blocked, delivered as findings ranked by severity.
- ZK-SNARKSEC-03A proof that a computation was performed correctly, short enough to check quickly and revealing nothing about the inputs beyond the truth of the statement.