If an LLM writes the plan, is the LLM really deciding?

No. A planner proposal arrives as unverified data with action IDs and cited evidence that must match currently feasible actions and the person's own recorded events. It never authorizes anything, and the sim runs completely with Jev alone.

What stops a bad Jev answer from moving the world?

The validResponse check in source/core.js rejects answers that do not identify a Jev model, select an existing option, cover every option with probabilities summing to 1, or match the highest probability. Invalid or missing answers pause the world instead of triggering a fallback.

Can I inspect exactly what was sent and returned?

Yes. AI live shows every Jev and planner call with person, phase, engine, status, latency, tokens, chosen option, and confidence, and each row keeps instructions, perception, request body, raw response, and validation. Keys are redacted and truncation is recorded.

Do I need a planner LLM just to start it?

No install or build step is needed because the HTML ships built; run the start script or node server.cjs, open the default loopback address, supply a key in Connect TypeSafe, then Run lives. Starting the server and picking a scenario never call a model.

How do I start it, and is there a build step?

Install Node.js 22 or newer, extract the folder and run START-WINDOWS.cmd or START-MAC-LINUX.sh, or node server.cjs --open. The shipped HTML is already built, so there is no npm install or application build, and the default address is http://127.0.0.1:4317.

What happens if Jev returns an invalid answer?

The validResponse check in source/core.js rejects an answer unless it identifies a Jev model, selects an existing option, gives a probability for every option that sums to 1, and matches the highest probability. An invalid answer changes nothing, and if Jev is unavailable the world pauses with no fallback.

Which planners can I use, and can they change the decision?

Any OpenAI-compatible API or local model, plus Claude Code, Codex and agy CLI sign-ins, chosen in Connect TypeSafe. A planner proposes grounded intentions and speech but never authorizes anything, and the simulation runs completely with Jev alone.

CommunityJev Practices & Patterns

ORIGIN-CIVILIZATION

Life-and-civilization sim where every voluntary action is a validated Jev Choice with ledger receipts, while optional LLM planners may only propose grounded intentions.

See highlights

Open github.com

Stars: 3
Language: HTML
Published by: Community
Author: JacquesGariepy
Last verified: Oct 2026

JacquesGariepy/ORIGIN-CIVILIZATION

User repository on GitHub

View on GitHub

AI life-and-civilization simulation: TypeSafe Jev makes every decision (typed, probabilistic, auditable); LLMs plan — OpenAI-compatible APIs, local models (Ollama, LM Studio), Claude Code, Codex.

Forks: 0
Last pushed: Updated Sep 2026

Repo stats from the GitHub API, cached Oct 2026.

About this project

An AI-driven life-and-civilization simulation. Every new voluntary action is a typed, auditable decision by TypeSafe Jev. LLMs — any OpenAI-compatible API, local models, Claude Code, Codex or agy — can plan, but never decide. TypeSafe Jev decides. Each person's next action is a Choice with a probability for every option and a confidence value. Code checks the answer before anything happens, and every call is kept as a receipt in the ledger.

Quoted from the project's README. Read the full README.

Tagscommunity awesomejev catalog-import html

Works forJev for Classification

Used inJev for Software & DevTools

Highlights

Sends one person's perception, memories, and feasible actions as Choice questions checked before execution.
Keeps every Jev and planner call in a ledger with request body, raw response, and validation result.
Runs planners optionally via OpenAI-compatible APIs, local models, Claude Code, Codex, or agy.
Serves a prebuilt page plus Node server with no install step; the world pauses if Jev is unavailable.

Watch out

Needs Node 22 or newer and a personal TypeSafe key, and probabilities plus confidence are model outputs rather than guarantees of sensible behavior. Copy reviewed against the repository README on 2026-10-04, not from running it.

Questions about this entry

Short answers drawn from the linked project or page.

More Jev resources like this

nanojev-arena

Local snake-arena demo where you steer a golden snake against four NanoJev AI snakes, with each AI snake showing its live direction-probability bars from a single forward pass.

Jev Practices & Patterns#community#awesomejev#catalog-import

Community

caijinchun/nanojev-arenaHTML

decisionbridge

Python adapter that turns an existing LLM into a bounded decision function: define one question and its answers, get scores, optionally calibrate on your own examples, then abstain below a threshold you set.

Jev Practices & Patterns#community#jevbest#catalog-import

Community

grishahq/decisionbridgeHTML

kyotsu-ai-bench

A static HTML dashboard, four commits deep, scoring Jev against two luna variants on Japan's 2026 Common Test. The repo is results only: CSV tables, one JSON file and two HTML pages, with no scripts to regenerate them.

Jev Practices & Patterns#community#jevbest#catalog-import

Community

shibadogcap/kyotsu-ai-benchHTML

thaiexam-jev-charts

Static chart pages comparing Jev against 110 other language models on Thai standardized exams: cost and speed plotted against accuracy, plus a page on whether its stated confidence predicts correctness.

Jev Practices & Patterns#community#jevbest#catalog-import

Community

vehas/thaiexam-jev-chartsHTML

jev_local

Local unofficial /v1/systemone adapter that serves Choice, Score, and Noul probabilities via first-token LLM judgments or a fine-tuned ModernBERT scorer, including Japanese docs.

Jev Practices & Patterns#community#awesomejev#catalog-import

Community

Argos1111/jev_localPython

jev-on-a-laptop

Unofficial laptop study of parallel constrained decoding: prefill once, share the KV cache across fields, and decide all schema fields in one batched forward pass without writing JSON.

Jev Practices & Patterns#community#jevbest#catalog-import

Community

rorshopping/jev-on-a-laptopPython

From the Jev guides

Original write-ups on what this entry is for.

Designing the State: What to Include and What to Drop

Oct 12, 2026

Most Jev accuracy problems are state problems. What belongs in a state, what is a pre-made judgment to leave out, and why naming the slice you ask about matters more than the wording.

Jev Practices & Patterns#state#design#primitives

Read article

Noul, End to End: One Probability and a Dozen Thresholds

Oct 12, 2026

A Noul returns one number and no separate confidence. How to read the value, pin its boundary with criteria, split compound questions, and set a per-action threshold that matches the cost of being wrong.

Jev Practices & Patterns#noul#primitives#confidence

Read article

Scoring Without Measuring: Bands, Rubrics, and Composite Triage

Oct 12, 2026

A Score is a weighted position, not a ruler. Why the number can land between levels, why levels are judged separately, and how to combine several atomic scores in code instead of one crowded rubric.

Jev Practices & Patterns#score#primitives#scoring

Read article

Back to all resources

From the Jev community

Posts from builders shipping with Jev right now.

Follow @typesafeai