Definitions of every term ØØT uses in a non-standard or non-obvious way. If a term means roughly what English suggests it means, it is not here.
ØØT. The framework. Pronounced "out". The two crossed-out circles represent legacy headcount and empty hours — the things being negated; the T is Tomorrow / Threshold / the new vector. In ASCII contexts (URLs, repository names, shell commands) we use oot.
Generation 1 (Gen 1). The current release (v1.0). Cloud-track full + privacy-track parity, FIAT compensation, Excel-bridged long-tail tracking, Klarna Test, EU AI Act mapping. What works today.
Generation 2 (Gen 2). v2.0, expected 6–12 months. Stablecoin rails, smart-contract long-tail, Internal Unit Fund, Curator local-LLM ingest, Tier-2 Skill Pack hardening.
Generation 3 (Gen 3). v3.0, expected 12–24 months, research-stage. Cotrugli Ledger anchoring, autonomous agent business units (Kelly / OpenClaw pattern), per-agent compensation.
Lego. A composable unit of the framework — a partner concept, a tool, a Skill Pack, an Excel template, a Routine, a governance document. The framework is a Lego inventory; v1.0 is the operational subset.
Partner. The unit of work in ØØT. Not an employee. Contracted for output, not hours. Receives base + variable + long-tail compensation. Declares a reward species. May (Gen 2+) hold Unit Fund credits and units.
Reward species. From the MHPR Advisors typology of seven partnership reward models. The three primary ØØT options: pure eat-what-you-kill, pure lockstep, declared hybrid with explicit weights. Recorded in reward-species-declaration.xlsx, version-controlled, renegotiable on a defined cadence.
Cohort. A partner's engagement type: full-time partner (full output participation, monthly variable, long-tail, unit holdings), project specialist (defined project, defined deliverable, defined amount, exit), or advisor (continuous engagement, narrow scope, smaller stake).
Output. A shipped artefact: a merged commit, a signed contract, a launched product, a closed deal, a published article, a delivered design. The unit the partner is paid against.
Outcome. The downstream value an output produced. Output is "the contract was signed"; outcome is "the contract generated €2M of revenue over three years." The long-tail layer pays against outcome.
Output spec. The pre-shipped definition of what "done" looks like for a committed piece of work. Authored before work begins; referenced at variable-pay calculation time.
Value tier / value envelope. The size category of an output's expected outcome. A flagship game spec has a larger envelope than a maintenance patch spec. Set at planning, refined as outcomes come in. Drives the variable-pay multiplier.
Attribution agent. The (Gen 1: Routine + Skill + Excel; Gen 2: dedicated agent) that reads commits, specs, reviews, contracts, milestones; produces a daily per-partner output ledger; trains a value-creation model on the firm's own ground truth.
Variable pay. The monthly payment computed from output × value tier × multiplier. Lands the month after output ships.
Long-tail entitlement. A percentage entitlement against an output's outcome. Pays quarterly for as long as the output generates value. Tracked manually in Excel in Gen 1; smart-contract in Gen 2.
Unit Fund. The internal open-ended fund of the company's units. Continuously priced. Dividend-paying. Bid/ask liquidity. The right to subscribe is earned through contribution (subscription credits issued alongside variable pay). Cash without credits cannot buy in. Generation 2.
Subscription credits. Non-transferable, time-bounded entitlements to subscribe to the Unit Fund. Issued alongside variable pay in proportion to contribution. Generation 2.
Two Worlds of Code. Karpathy's distinction between vibe coders (legacy, pre-Software-3.0) and agentic engineers (Software 3.0 native). Self-identification in onboarding helps the partner understand which world they are entering.
Second Brain. A personal knowledge graph — one per individual. Markdown pages with YAML frontmatter and wikilinks, stored locally by the Curator app on the individual's own machine. Captures their own conversations, decisions, research, contacts, project notes. Can be optionally synced two-way to a private GitHub repo (the Second Brain repo, see below). Standard PKM term (Tiago Forte). Distinct from the Firm Brain (which is the shared collective wiki of a firm) and distinct from the Ledger (which holds operational state). Every ØØT partner has their own Second Brain; one of its domains is opted-in to contribute to the Firm Brain.
Firm Brain. The firm's shared collective knowledge graph — the synthesized organisational memory ("the firm IP"). Lives as a Curator Shared Brain instance (v3.0.0-beta+) in a dedicated GitHub repository (<firm-slug>-brain). Each partner contributes from one opted-in domain of their personal Second Brain via Curator's Push protocol; the admin runs weekly Synthesize, which merges contributions into the collective/<domain>/wiki/ tree with UUID-pseudonymous Provenance attribution. Partners then Pull a read-only synthesized mirror (shared-<slug>/) into their personal Curator. Distinct from the Ledger (which is Excel + audit logs operated by Routines) and distinct from each partner's Second Brain (which is personal). Recorded in ADR-002.
Ledger. The firm's GitHub repository holding operational state — Excel files (X1–X9), Routine-written markdown (firm/output-logs/, firm/audit-logs/, etc.), signed-commit history. Mutated by Routines via openpyxl + signed commits per ADR-001. Distinct from the Firm Brain (the firm's synthesized shared knowledge graph) and distinct from the Second Brain (each partner's personal Curator vault). Also distinct from the Cotrugli Ledger (Gen-3 accounting layer — see below). The wizard creates this as <firm-slug>-ledger (defaulted from <firm-slug>-brain in pre-v1.0.1 installs — both names are valid).
Shared Brain (Curator primitive). The Curator v3.0.0-beta+ feature that powers the Firm Brain. A multi-contributor collective wiki stored in organization-controlled storage (GitHub today; Cloudflare R2 with EU residency in Curator v3.1). Three operations: Push (a partner uploads DeltaSummary JSON of their opted-in domain to contributions/<fellow_id>/), Synthesize (admin-run, merges contributions into collective/<domain>/wiki/ with Jaccard + selective-LLM conflict resolution), Pull (each partner downloads the synthesized mirror into a sibling read-only shared-<slug>/ domain). Includes built-in GDPR Article 17 admin revoke endpoint with typed-confirmation gate. Spec: shared-brain-design.md.
Opted-in domain. The single domain of a partner's personal Second Brain that participates in the firm's Shared Brain. The partner selects it during Curator's six-step contributor wizard at onboarding. The partner's other domains stay private. Writes flow: partner edits page in their opted-in domain → Push uploads a preprocessed delta → Synthesize merges into Firm Brain → Pull mirrors the synthesized result back as shared-<slug>/ (read-only).
Synthesized mirror. The local read-only domain (shared-<slug>/ with readonly: true frontmatter) that each partner sees after Pull. Mirrors the firm's collective/<domain>/wiki/ tree. MCP write tools (compile_to_wiki, fix_wiki_issue) refuse mirror-domain writes — to contribute to the Firm Brain, partners edit their personal opted-in domain instead, then Push.
Second Brain repo (deprecated as a framework primitive in v1.0.1). Personal GitHub backup of a partner's entire Curator vault (all domains together) via the Curator's built-in two-way sync. Retired as a framework concept in v1.0.1 per ADR-002 — cloud Routines that need firm context read the Firm Brain repo's collective/<firm-domain>/wiki/ instead, which is synthesized across all partners with UUID Provenance and conflict markers. Personal Curator backup-to-GitHub remains available as a per-partner tooling choice (the partner configures Curator's personal Sync feature on their own; the framework does not provision or coordinate it). Existing v1.0 <firm>-secondbrain repos may be kept as personal backup, archived, or deleted at firm discretion.
The Curator. The reference Second Brain implementation. Open source. Includes the MyCuratorMCP server (17 tools, plus Shared Brain–aware mirror guards in v3.0.0-beta+), the Curator desktop app (knowledge ingest + two-way GitHub sync + Shared Brain wizards), and the canonical SKILL.md. Repo: https://github.com/talirezun/the-curator.
Skill / Skill Pack. A markdown file (or folder of files) following the SKILL.md format from Anthropic Agent Skills. Encodes operational discipline for a domain. Loaded into Claude Desktop, Claude Code, Cursor, LM Studio, or any MCP-compatible client. ØØT ships 12 packs in v1.0.
MCP (Model Context Protocol). The open standard for agent ↔ tool communication. Now governed by the Linux Foundation Agentic AI Foundation. ØØT depends on MCP as a primitive.
AGENTS.md. The cross-vendor agent orientation file at the root of an agent-readable repository. Tells any AI agent how to navigate the repo, what conventions are in force, what tools are available.
CLAUDE.md. The Claude Code-specific orientation file. Same purpose as AGENTS.md but Claude Code reads it preferentially. ØØT ships a CLAUDE.md at the root.
Routine. A scheduled prompt that runs on a trigger (time, GitHub event, API call). Cloud track = Claude Code Routines. Privacy track = OS-native scheduling hitting headless LM Studio. ØØT ships 8 Routines in v1.0. Operational .xlsx state lives in the firm's Ledger repo and is mutated by Routines via openpyxl + signed commits per ADR-001. Cloud Routines that need company context (R5, R2, R8) additionally read the Second Brain repo (scoped to the firm's Curator domain).
Cloud track. The canonical, fastest path. Claude Desktop, Claude Code, Slack, GitHub (Ledger holds markdown + .xlsx state; Second Brain repo holds the Curator-synced semantic graph), the Curator, MyCuratorMCP, Claude Code Routines. Anthropic infrastructure where applicable. Spreadsheet viewer is user choice (Excel / LibreOffice / Numbers / Excel-for-Web).
Community track. The free-to-start, no-subscription, no-dedicated-hardware operating configuration (ADR-003). Install harness and daily-ops agent is OpenCode on free built-in / own-key / local models; the Ledger and Firm Brain are GitHub (unchanged); Brain ingest is Curator + Gemini Flash Lite pay-as-you-go; scheduled Routines run on a three-rung automation ladder (manual playbook runs → laptop cron → GitHub Actions). Makes no sovereignty claims — budget-motivated founders use this, sovereignty-motivated founders use the privacy track. Everything but the harness and scheduler is byte-identical to the cloud track. Harness setup: installer/agent-assisted/OPENCODE-SETUP.md.
Privacy track. The full-Gen-1-parity sovereignty path. LM Studio with local Qwen / Llama / DeepSeek, Desktop Commander MCP, Excel MCP, 4thtech for on-chain communication, PollinationX for decentralised storage, GitHub MCP, OS-native scheduling.
4thtech. The on-chain end-to-end-encrypted communication and file-transfer framework. Wallet-as-identity, fully self-custodial, permissionless. Privacy-track comms layer. https://github.com/4thtech.
PollinationX. The decentralised storage protocol. NFT-based capacity. Encrypted, distributed across nodes. Privacy-track Drive equivalent. https://wiki.pollinationx.io/.
Lumina AI. The RAG chatbot widget. External-facing front door for sales, support, lead-capture. The "Lumina-as-front-door" pattern is referenced in the Sales & BD Skill Pack.
Klarna Test. The signature epistemic check. Any ØØT recommendation that would have produced the Klarna outcome (publicly cut humans, quietly rehired) must be flagged. Ten-question scoring rubric in governance/KLARNA-TEST.md. Each question scored 0, 1, or 2 (max 20). Score <14/20 blocks proceeding. Wired into Code & QA as a pre-merge gate on any PR labelled ai-replaces-human.
METR baseline. Pre-rollout productivity baseline using DORA + SPACE + DX Core 4 metrics. Mandatory before any major Skill rollout. The METR study showed senior engineers were 19% slower while feeling 20% faster — without a baseline, organisations cannot detect their own regressions.
ROWE. Results-Only Work Environment (Cali Ressler & Jody Thompson, Best Buy 2003). The intellectual ancestor of ØØT's pay-for-output principle. Reports ~35% productivity gains and ~90% reduction in voluntary turnover.
SECI model. Socialisation → Externalisation → Combination → Internalisation (Nonaka & Takeuchi). The knowledge-creation cycle the Brain operationalises.
Beyond Budgeting. Hope & Fraser's 1998 management framework. ØØT borrows the rolling-forecast and continuous-allocation principles for the Reporting & Business Review Skill Pack.
AI Champion. A partner who has demonstrably increased their throughput and quality via AI tooling, then mentors others. Earned, not appointed. Tecknoworks/Caplaz case studies show artificial champions backfire.
Cotrugli Ledger. The Generation 3 governance backbone — distinct from the operational Ledger (see Ledger above). Triple-entry-style accounting via PAC-RO receipts, with Cotrugli Score / Vanguard Score reputation, IAAF agent autonomy levels. Theoretical / research-stage in v1.0. Co-authored by Dražen Kapusta.
PAC-RO. Policy-anchored, Co-signed Receipt Object — the unit of accounting in the Cotrugli Ledger. Composed of Facts, Evidence, Policy, and Co-signature. Generation 3.
The Five Theses. Resistance, Centaur work, Partner compensation, Collecting Brain, Composable Lego. The intellectual core of the framework. See MANIFESTO.md.
Bitwarden. Open-source password manager. ØØT secrets policy uses Bitwarden (or 1Password as commercial alternative) for software credentials — API keys, OAuth tokens, recovery codes.
Trezor. Hardware wallet. ØØT secrets policy uses Trezor (or Ledger as alternative) for crypto signing — payroll wallets, treasury wallets, 4thtech identity wallets.
Yubikey. Hardware security key. Used for org-level admin accounts (GitHub, Anthropic, Google admin). Recommended in the secrets policy.
Excel MCP. The MCP server (haris-musa/excel-mcp-server, MIT-licensed) that lets local LLMs read and write Excel files. The piece that gives the privacy track full Gen 1 automation parity.
llmster. LM Studio's standalone headless daemon for the Linux / headless-server path — it hosts local models on LM Studio's OpenAI-compatible server (http://127.0.0.1:1234/v1) on a box with no GUI app. On macOS with the LM Studio desktop app installed there is no separate llmster binary — the app's own server (started with lms server start) fills the daemon role. Either way it is a model server, not an agent: it does not load skills, clone repos, or call MCP tools. On the privacy track the agent harness that does all that is OpenCode (opencode run), which uses the locally hosted model.
lms. LM Studio's command-line tool (ships with LM Studio), used on both the macOS-app and Linux-daemon paths. Manages the model server and loaded models: lms server start, lms load <model> --context-length 32768 --ttl <seconds> (the --context-length is required — LM Studio's default 4096-token context can't hold an agent's system prompt, so a default load fails with n_keep >= n_ctx; the --ttl keeps a model warm so scheduled Routines skip the cold-load), lms unload, lms ls (lists loaded models and their exact identifiers).
Desktop Commander MCP. Local filesystem MCP server. Privacy-track replacement for the Google Drive Claude connector.
Augment Code, Codex CLI, OpenCode. Specialised coding assistants. ØØT references them in the Code & QA Skill Pack: Augment for big multi-file builds, Codex for review/audit, OpenCode for the open-source / privacy-track path.