Coin Brief ENDE

EIP-8360 proposes TCREATE, a contract that exists for one transaction

EIP-8360 proposes a new opcode, TCREATE, that creates contracts which exist only for the duration of a single transaction. At the end of the transaction, all code, nonce and storage created during it are discarded. The proposal is Draft, type Core, created 3 August 2026, by Helkomine and abc-123-c.

The 20 September update is editorial but revealing: it clarifies ephemeral-account semantics, adds a third author (Milos), and corrects the requires list — adding EIP-684 and EIP-7928 and dropping EIP-7610. The address computation is defined as keccak256(0xfe || deployer_address || salt || keccak256(init_code))[12:], deliberately parallel to CREATE2's scheme with a different domain-separator byte.

EIP-8360 proposes TCREATE, a contract that exists for one transaction
EIP-8360 proposes TCREATE, a contract that exists for one transaction — Coin Brief

What it means

"Transaction-scoped contract" is a primitive several patterns currently fake. Today, code that needs a helper contract for one transaction deploys it and either leaves it on-chain forever or self-destructs it — both awkward, both with edge cases. An opcode whose entire output vanishes at transaction end makes the throwaway-contract pattern first-class instead of a workaround.

The requires churn is where the real design work shows. Adding EIP-684 (which concerns what happens when an address is already occupied) and EIP-7928 to the dependency list, mid- draft, means the authors are still nailing down how TCREATE interacts with existing accounts at the target address — the exact corner where "ephemeral" stops being simple. A dependency list that is still moving is a proposal still deciding what it means.

It is a Draft, and the address scheme is the load-bearing part to scrutinise. Reusing CREATE2's structure with a 0xfe prefix keeps addresses deterministic and collision-separated from ordinary deployments. Whether one transaction's ephemeral contract can ever collide with a real one at the same address is the question a reviewer should push on before this advances.

Source: https://github.com/ethereum/EIPs/commit/d288471ea8026e6a5b816218ebe8fb1209a34d6e

Written by Victoria Shinder.