constraining

Type: kb/types/tag-readme.md · Status: current

Making semantics more focused by narrowing the space of valid interpretations an artifact admits — from partial narrowing (conventions) to full commitment (deterministic code). The primary mechanism for hardening deployed systems, with relaxing as its deliberate reverse. A child of learning-theory.

Definition and spectrum

• constraining — definition and spectrum: storing an output, writing a convention, adding structured sections, extracting deterministic code
• codification — the far end, where the medium itself changes from natural language to a symbolic artifact with formal semantics
• agentic systems interpret underspecified instructions — the foundation: the spec-to-program projection model, semantic boundaries, and the constrain/relax cycle

Instances and techniques

• storing LLM outputs is constraining — the simplest instance: keeping a specific output resolves underspecification to one interpretation
• constraining during deployment is continuous learning — versioned constraining beats weight updates on inspectability and rollback
• spec mining as codification — observe behavior, extract deterministic rules, grow the calculator surface monotonically
• error messages that teach are a constraining technique — in agent systems the error channel is an instruction channel
• methodology enforcement is constraining — review gates and validation as constraining applied to the KB's own methodology

Deciding and reversing

• codify-versus-LLM decision heuristics — four lenses on the codify-vs-LLM decision, with evidence they come apart at the edges
• specification strategy should follow where understanding lives — when to commit: before execution, during execution, or after repeated observation
• progressive constraining commits only after patterns stabilize — the default discipline: codify after the pattern proves the need
• unified calling conventions enable bidirectional refactoring — how to commit reversibly: neural and symbolic components behind the same callable interface
• codification and relaxing navigate the bitter lesson boundary — why reversibility matters: codification is a bet that may need relaxing
• operational signals that a component is a relaxing candidate — five testable signals for detecting when to reverse codification

Related Tags

• deploy-time-learning — the framework constraining serves; the verifiability gradient locates constrained artifacts
• distillation — the orthogonal mechanism: extraction rather than narrowing

Other tagged notes

• Ad hoc prompts extend the system without schema changes - Any system with an LLM agent layer can absorb new requirements through natural language prompts without changing the deterministic base
• Constraining and distillation both trade generality for reliability, speed, and cost - Constraining narrows interpretation (largest gain at codification, where substrate changes); distillation extracts under a context budget. Same capacity decomposition, different operations
• Current LLM inference removes composition friction's filter and hides its signal - The effort of turning a vague idea into committed prose does double duty — it filters out ideas that cannot survive concretization and it signals where understanding is weakest; current LLM inference removes the filter (the unsound idea ships anyway) and hides the signal (the model's confidence tracks typicality, not soundness, so there is no faithful stall to read off generation). Both losses are intrinsic to the generated output but reconstructable downstream by a separate check; the gap-hiding itself is shared with human writing — the difference is rate and observability, not kind
• Enforcement without structured recovery is incomplete - The enforcement gradient covers detection and blocking but has no recovery column — recovery strategies (corrective → fallback → escalation) are the missing layer, and oracle strength determines which are viable at each level
• Fixed artifacts split into exact specs and proxy theories - Fixed artifacts are safe when their spec fully captures the problem; they are risky when they encode proxy theories whose components may not compose into the larger capability
• Inspectable artifact, not supervision, defeats the blackbox problem - Chollet frames agentic coding as ML producing blackbox codebases — codification counters this not by requiring human review but by choosing readable artifacts (code, prompts, schemas) that any agent can inspect, diff, test, and verify
• Legal drafting solves the same problem as context engineering - Legal drafting parallels context engineering because both write ambiguous natural-language specifications for judgment-based interpreters, but law develops constraining more than codification
• LLM↔code boundaries are natural checkpoints - At each LLM↔code transition both semantic underspecification and execution indeterminism collapse simultaneously, making these boundaries natural places to anchor debugging, testing, and refactoring
• Opacity is a scale threshold, not a class property - Opacity is not a representational form; any representation becomes practically opaque at sufficient scale, though distributed-parametric artifacts cross that threshold earliest.
• Underspecification and indeterminism complicate programming for prompts in distinct ways - Indeterminism doubles test runs (statistical testing over distributions); underspecification doubles test targets (spec analysis for ambiguity). Conflating the two leads to misdiagnosis