Codex Glossary

Work requiring attention multiple times per week, tied to immediate shipping deadlines.

HF threads are the foreground of creative operations. They demand regular check-ins, carry deadline pressure, and represent active client or project commitments. Limited to 5 concurrent items to prevent overload.

Also known as: rapid-response work, daily-cycle tasks, active sprint work, hot tasks, urgent work, fast-turnaround tasks

Related: lf , dormant , wip-limit

Used in: multiclock , multi-clock-problem , governance-protocols

Work refreshed every 2-6 weeks for strategic development.

LF threads represent important-but-not-urgent work that benefits from incubation. They receive scheduled attention to prevent neglect while avoiding the pressure of daily deadlines.

Also known as: periodic refresh work, strategic cadence tasks, background development, slow-burn projects, maintenance rhythm work

Related: hf , dormant , incubation

Used in: multiclock , multi-clock-problem

Parked ideas reviewed monthly or quarterly, deliberately not active.

Dormant threads are not abandoned—they are intentionally resting. Monthly reviews surface aging items that may deserve promotion. This prevents idea loss while maintaining focus on active work.

Also known as: parked ideas, incubation queue, idea backlog, sleeping projects, cold storage, someday-maybe list

Related: hf , lf , aging-points

Used in: multiclock , multi-clock-problem

Numerical index balancing exploitation of proven ideas with exploration of new concepts.

Based on multi-armed bandit algorithms from machine learning. Combines base value, confidence adjustments, and aging points to produce a daily priority ranking that prevents both stagnation and chaos.

Also known as: priority index, exploitation score, attention allocation index, priority ranking, task score

Related: exploration-exploitation , aging-points , ucb

Used in: bandit-scoring , multiclock

Points accumulated daily by waiting items to prevent indefinite neglect.

Each day an item waits without attention, it gains age points. This ensures even low-priority items eventually surface for review, preventing good ideas from being permanently buried.

Also known as: priority decay, starvation prevention, waiting time bonus, neglect compensation, queue aging

Related: dormant , bandit-score

Used in: multiclock , governance-protocols , bandit-scoring

The tension between trying new approaches (exploration) and optimizing known successes (exploitation).

Creative studios must balance shipping proven work with experimenting on new directions. Too much exploitation leads to stagnation; too much exploration prevents mastery.

Also known as: explore vs exploit, new vs proven, innovation vs optimization, discovery vs delivery

Related: bandit-score , ucb

Used in: bandit-scoring , multiclock

A bonus added to items with high uncertainty to encourage trying less-tested options.

Items that have not been tested much receive a confidence bonus that decreases as more data is gathered. This mathematically ensures adequate exploration.

Also known as: confidence bonus, uncertainty bonus, exploration bonus

Related: bandit-score , exploration-exploitation

Used in: bandit-scoring

Central database tracking all work threads with clock frequency assignments.

The MIR is the single source of truth for all active, low-frequency, and dormant work. Each entry includes clock assignment, status, owner, and metadata for prioritization.

Also known as: idea register, task database, work registry, master task list, project tracker

Related: hf , lf , dormant , bandit-score

Used in: multiclock , governance-protocols

Maximum number of items allowed in active work state simultaneously.

Derived from Kanban and queuing theory. Limiting work-in-progress reduces context switching costs, improves flow, and forces completion before starting new work. Typical HF limit is 5 items.

Also known as: work-in-progress cap, concurrent task limit, active thread cap, parallel work limit, context switching limit

Related: hf , littles-law

Used in: multiclock

Lead Time = WIP / Throughput. Reducing WIP directly reduces delivery time.

A fundamental law of queuing theory proving that the only way to reduce lead time without increasing throughput is to reduce work-in-progress.

Also known as: queue theory, throughput equation, flow time formula

Related: wip-limit

Used in: multiclock

Quality control framework: Specification (define done), Verification (prove completion), Learning (extract insights).

Every task passes through three gates. Gate 1 defines success criteria before starting. Gate 2 verifies completion with evidence. Gate 3 captures learnings for future improvement.

Also known as: specification-verification-learning, quality gates, closure protocol, completion checklist, done criteria

Related: exemplar-library , agent-brief

Used in: governance-protocols

Collection of approved work samples used as quality references for AI and human output.

Instead of abstract style guides, the Exemplar Library contains concrete examples of excellent work. AI agents and humans reference specific exemplars to match tone, structure, and quality.

Also known as: reference library, quality benchmarks, tone references, template library, gold standards

Related: three-gate-system

Used in: governance-protocols

Structured instructions for AI agents including context, deliverables, and success criteria.

Agent briefs come in three tiers: Micro (single-sentence), Standard (context + deliverables + criteria), and Macro (full strategic context with milestones). Brief complexity matches task complexity.

Also known as: task brief, delegation brief, AI instructions, work order

Related: three-gate-system , autonomy-matrix

Used in: governance-protocols

Four-tier framework defining what AI can decide autonomously vs. what requires human approval.

Tier 1: Full autonomy (task updates, formatting). Tier 2: Guided autonomy (priority changes, drafts). Tier 3: Human authority (strategy, finances). Tier 4: Human only (ethics, legal).

Also known as: decision boundaries, delegation tiers, authority levels, AI permissions

Related: agent-brief

Used in: governance-protocols

Structured note capturing work state before any context switch to enable seamless resumption.

Contains: Context (where we are), Decisions Made (key choices), Next (immediate action), Blockers (obstacles), Open Questions (unresolved issues), Artifacts (relevant links).

Also known as: context handoff, session summary, work state capture, resumption note

Related: three-gate-system

Used in: governance-protocols

The cognitive phenomenon where stepping away from a problem leads to better solutions upon return.

Research shows that breaks from focused work allow subconscious restructuring. The Multi-Clock system deliberately schedules incubation windows for LF and Dormant work.

Also known as: creative pause, subconscious processing, stepping away, percolation time

Related: lf , dormant , dmn

Used in: multiclock , multi-clock-problem

Brain network active during internally-directed thinking, associated with creativity and insight.

Neuroscience research links the DMN to creative insight. Engaging this network requires stepping away from focused external tasks—supporting the case for scheduled incubation.

Also known as: resting state network, mind wandering network, internal thought network

Related: incubation

Used in: multiclock

The cognitive cost of switching tasks before completion—mental residue from Task A harms performance on Task B.

Research by Sophie Leroy shows that unfinished tasks create attention residue that degrades focus on subsequent tasks. This supports the Multi-Clock emphasis on completion before switching.

Also known as: task switching cost, mental carryover, focus fragmentation, context switching penalty

Related: wip-limit

Used in: multiclock , multi-clock-problem

Time spent deciding what to work on rather than doing the work itself.

Studios without clear prioritization systems spend excessive time each day deciding what deserves attention. The Multi-Clock system reduces decision latency by providing clear daily priorities.

Also known as: decision fatigue, priority paralysis, what-to-work-on delay, task selection time

Related: bandit-score , mir

Used in: multi-clock-problem , governance-protocols