SPEAR / 0.1 · open draft
Structure without false precision.
SPEAR combines ordinary language with just enough formal structure to make objectives, abstractions, uncertainty, and repair visible.
natural languagegoals · context · exceptions
SPEARshared specification pidgin
formal languagetypes · invariants · constraints
01TASK02OBJECTS & TYPES03ABSTRACTION04OBJECTIVE05CONSTRAINTS06UNCERTAINTY07OUTPUT08EVALUATION09INTERACTION10EXAMPLES
Build a portable specification
Start with the task, name the abstraction boundary, and define what success means. Empty fields remain explicit so collaborators can see what is still unresolved.
Generated SPEAR
SPEAR/0.1 TASK Optimize the anomaly detector for production traffic. OBJECTS & TYPES event: Event; detector: Model; workload: Distribution ABSTRACTION PRESERVE: recall, latency_p99, throughput IGNORE: implementation language, internal model family OBJECTIVE minimize latency_p99 CONSTRAINTS HARD throughput >= 1e6 events/s HARD delta_recall >= -0.01 SOFT memory <= 64 GB UNCERTAINTY baseline workload may differ from production OUTPUT benchmark report + deployable configuration EVALUATION all hard constraints pass on held-out workload INTERACTION ask if the baseline workload or recall metric is missing EXAMPLES near-miss: lower latency that loses 2% recall
A proposal, not a standard.
SPEAR/0.1 is an experimental schema. Its syntax and semantics should change through documented RFCs, real tasks, adversarial tests, accessibility research, and implementation experience.
Do not mistake formal-looking notation for truth. Types and metrics reduce syntactic ambiguity, but assumptions, loss functions, power, and domain validity still require scrutiny.