SPLICE✳
SPLICE / FIELD NOTES / V1

The mechanism.
In the open.

SPLICE is an AI workbench for holders. The practical benefit is funded AI-task credits, reusable tested agent setups and input into community challenges. These notes explain the operator-managed holder programme and the free lab you can use here.

01 / One coin. One lineage.

A lineage begins with a founding agent configuration. Its generations are versions of agents, with parent relationships and a recorded history. The community coin represents that lineage; creating a new agent generation does not automatically create a new token.

The public website includes a reference lineage. SPLICE’s operator manages the holder programme and shared AI work separately from this browser lab.

02 / What a genome contains.

  • Model: the chosen provider and model identifier.
  • Instructions: the task and behavior specification.
  • Tools: capabilities selected from an approved set.
  • Verification: source checks, test harness and scoring rules.

A genome is a versioned configuration. SPLICE does not combine proprietary model weights or claim that a configuration mutation changes the underlying model's intelligence.

Permissions, credential access, spend ceilings and allowed actions remain fixed by the operator. A genome cannot expand them through inheritance or mutation.

03 / Inherit. Vary. Test.

Two parents contribute selected configuration components to an offspring. A published mutation policy controls which components may vary and how frequently. Each child records its parents and changes.

The browser reference lab uses four Boolean extraction rules: tolerant parsing, markup handling, source validation and number extraction. It uses browser randomness for crossover, then exposes the complete result. It makes no model call and spends no funds.

Try a reference crossover ↗

04 / Evidence before promotion.

The evaluator declares the task set, scoring method, budget and promotion criteria before testing. A production evaluation should include fresh or held-out tasks; selecting a winner on the same tasks used to generate it can exaggerate improvement.

The reference lab runs eight fixed extraction fixtures. Every displayed pass or failure comes from comparing the actual rule-based result with its expected output. Measured runtime is browser-local. These results do not represent model accuracy, paid usage or shared-engine performance.

A receipt contains

Specimen and configuration version · parents and changes · task identifier · inputs and expected/actual outputs · scoring method and result · evaluation scope and timestamp · SHA-256 digest of the recorded payload.

A digest helps detect changes to a receipt. It does not independently establish that a remote model ran, that a log is complete or that an evaluator is truthful. The operator must preserve verifiable artifacts and disclose the trust boundary.

Failures remain in the lineage. Passing configurations can be exported. Better performance on one benchmark does not establish general improvement or future financial performance.

05 / Holder access and service limits.

The holder benefit has three parts:

  • AI-task credits: a funded allowance for running supported agents on useful tasks.
  • Tested agent setups: reusable workflows with inspectable results, so you can spend less time building prompts and tool settings.
  • Community challenges: input into which tasks and configurations the community tests next.

Credits are units of service access. They are not a defined cash payout or redemption balance. The holder programme provides service benefits rather than a cash-reward balance.

On this website: the free browser lab, crossover, fixture tests and exports. Managed by SPLICE’s operator: holder eligibility, funded task credits and shared AI work.

Holder access follows the operator’s eligibility, credit and capacity rules. Credit allocations, supported tasks and community challenge selection are managed by SPLICE. This website does not define a universal holding threshold or unlimited allowance.

SPLICE’s operator handles delivery of holder benefits, access checks and funded capacity. The website’s reference lab runs independently in your browser. The Holder access panel explains the programme; holder administration is handled separately by SPLICE.

06 / Token identity.

SPLICE ($SPLICE) is prepared for Solana mainnet. Its mint address is reserved; on-chain token deployment has not yet been confirmed.

2GUXPVdW1SjCB62nucxHCwWT3v97iFvFfgBpv2UFjfun

Buy links activate only after the mint and pool have been verified. Token launch and the operator service are separate milestones. Holding a token does not guarantee access, rewards or returns.

SPLICE is independent of named model providers.

Read the SPLICE manifesto ↗
HOLDER ACCESS

Useful tools for holders.

Holder benefits: AI-task credits, tested agent setups and community-selected challenges.

OPERATOR-MANAGED ACCESS

SPLICE’s operator handles holder eligibility, credit allocations and task access. Explore the free browser lab here without connecting a wallet.

Try the free lab ↗How holder benefits work ↗