What the Organization Compiler is
The Organization Compiler is the part of JGI that answers one question per mission: what is the smallest organization that can do this job well? It does not build a permanent department. It compiles a temporary organization from existing capabilities, temporary workers, and minimal bridges, runs the mission, and dissolves the organization when the mission ends. Only what proved reusable persists.
The twelve sections below are reproduced in full from JGI V6, in order. The five tools after them implement those sections as working calculators. Nothing in the tools invents a formula the paper gives; where a tool needs a number the paper does not supply, the number is labelled illustrative.
Section 18 Planner of Planners
PRIME should not always plan directly.
For complex missions: generate multiple independent plans.
Possible plan generators: PRIME, specialist worker, skeptical worker, economic optimizer, security critic.
Then evaluate them.
Plan utility: expected success, cost, time, risk, reversibility, resource contention, evidence quality.
Choose a winner.
Section 19 Recursive Decomposition
Mission decomposition continues until tasks are: independently executable, verifiable, resource-bounded, authority-classified, small enough to estimate.
Do not over-decompose.
Stop when coordination overhead exceeds decomposition benefit.
Section 20 Cognitive Router
JGI should select the right cognition for each task.
Options may include: PRIME, Grok specialist, large Grok swarm, Claude, GPT, local model, deterministic code, SQL, search, constraint solver, statistical model, human judgment.
LLMs are not automatically the best tool.
Route using: P(success), cost, latency, privacy, tool availability, historical reliability, verification burden.
Section 21 Model Performance Memory
For each cognitive worker maintain performance by task family.
Example: worker/model, domain, task type, attempts, passes, failures, Sentinel rejects, cost, latency, calibration, quality.
Estimate reliability with a Bayesian posterior.
Use this in future routing.
Section 22 Elastic Organization Creation
JGI may create temporary organizations for missions.
Example: MISSION CEO = PRIME. Temporary team: Research Lead, Repo Archaeologist, Engineer, Data Scientist, Economic Analyst, Adversarial Critic, Verification Specialist.
The team dissolves after the mission.
Permanent infrastructure should NOT be the default.
Section 23 Organizational Compiler
For a mission determine: required knowledge, required tools, required parallelism, required independence, required verification.
Then compile the smallest effective organization.
Section 24 Persistent Bot Birth
A new permanent Bot requires evidence.
Calculate: recurrence, unique context value, distinct ownership, independent schedule need, specialized tool need, long-term learning benefit, minus: coordination cost, role overlap, maintenance cost.
Create only when durable value exceeds cost.
All Bots need: birth reason, scope, authority, metrics, retirement condition.
Section 25 Bot Death
Periodically challenge every persistent agent.
If: low unique value, high overlap, rare usage, weak performance, then: extract useful skills, preserve memory, retire Bot.
JGI should be capable of simplifying itself.
Section 69 Self-Assembling Mission Systems
A mission may require a temporary mini-company.
JGI may assemble: workers, skills, data, tools, workflow, memory, budget.
Execute mission.
Then dissolve organization.
Persist only what proved reusable.
Section 96 Swarm Diversity
When using multiple agents, diversify: roles, prompts, sources, models, assumptions.
Twenty identical agents are not twenty independent opinions.
Section 97 Information Gain Per Agent
Spawn another worker only when: ExpectedInformationGain exceeds coordination + compute cost.
Stop swarm growth when marginal returns fall.
Section 170 Specialization Emergence
If one worker/model repeatedly excels at a task class: increase routing.
If a durable role emerges: consider Skill/Workflow/Bot.
Let specialization emerge from evidence.
Interactive Tools
All token and dollar figures in the tools below use illustrative rates stated next to each tool. The paper gives no token counts, no prices, and no numeric formulas for organization compilation, routing, bot birth, or swarm growth. The formulas here are labelled illustrative implementations of the paper concepts, and every one of them shows its arithmetic so you can check it by hand.
Tool 1. Mission Org Composer Sections 22, 23, 69
Pick an objective. The compiler returns the smallest competent team from the paper Section 22 role set, with a communication topology and a cost estimate.
Tool 2. Recursive Decomposition Planner Section 19
Enter a mission. The planner decomposes it until subtasks are independently executable, verifiable, resource-bounded, authority-classified, and small enough to estimate. It stops when coordination overhead exceeds decomposition benefit.
Tool 3. Cognitive Router Sections 20, 21
Enter a subtask. The router selects the right cognition from the paper Section 20 option list, using P(success), cost, latency, privacy, tool availability, historical reliability, and verification burden. Reliability comes from the built-in Model Performance Memory below, estimated as a Bayesian posterior.
Model Performance Memory (built-in records, illustrative)
| Worker | Task family | Attempts | Passes | Failures | Reliability |
|---|---|---|---|---|---|
| Claude | Research | 20 | 18 | 2 | 0.864 |
| GPT | Creative | 15 | 12 | 3 | 0.765 |
| Local model | Confidential | 10 | 9 | 1 | 0.833 |
| Deterministic code | Code | 25 | 25 | 0 | 0.963 |
| SQL | Data analysis | 21 | 20 | 1 | 0.913 |
| Human judgment | Judgment | 16 | 15 | 1 | 0.889 |
| PRIME | Planning | 16 | 14 | 2 | 0.833 |
| Large Grok swarm | Research | 10 | 8 | 2 | 0.750 |
| Search | Research | 25 | 22 | 3 | 0.852 |
Tool 4. Bot Birth and Death Simulator Sections 24, 25, 170
Score a candidate permanent bot on the paper Section 24 factors. A bot is born only when durable value exceeds cost. Then run the Section 25 death challenge against the roster.
Candidate bot
Persistent roster
Tool 5. Swarm Diversity and Information Gain Scorer Sections 96, 97, 170
Score a team on the five paper Section 96 diversity axes, then test whether adding another worker is worth it under Section 97: spawn only while expected information gain exceeds coordination plus compute cost.
Specialization emergence Section 170
Enter a worker record. If one worker repeatedly excels at a task class, routing increases; if a durable role emerges, consider a Skill, Workflow, or Bot.