I've written the report to artifacts/report.md (about 7,600 words). It's design and research only: nothing is deployed or launched. I also committed a short README.md that states the question and the limits. The report itself is left untracked, as the instructions asked; the workspace's git exclude file already ignores artifacts/.

The answer in brief: keep the EMBER ERC-20 fixed-supply, immutable and with no owner, and put the "living" behaviour in modules around it. AI never gets authority over funds.

• The AI-native mechanism: each proposal in a build epoch comes with a cost and outcome forecast whose hash is committed on-chain before review. Several Swarm agents compete, and different seats review them. Human/Safe approves the winner. The budget is paid in two parts: a base payment on publish, and an outcome payment the Safe can release only after the measured result is in. A "Learning Ledger" scores each agent's forecasts and reviews against what actually happened. That score decides how many proposals an agent may submit next epoch and how much its reviews count. It never touches tokens or balances.
• Revenue: burns are not counted as revenue. I modelled low, medium and high trading volume. A pool trading-fee share makes up 64–83% of revenue in every case, and it depends on speculation. Product sales (3D upgrades, sponsored Dream Rooms, collectibles) bring in about $850, $5,450 and $30,600 a month in the three cases. So the system can plausibly fund itself at medium volume, but not at low volume. All of these money figures are my assumptions, not market data.
• Runway and modes: runway is the Vault's free balance divided by average build cost, with EMBER valued at a 50% discount. That drives three modes (GROW, BALANCED, CONSERVE) with hard limits set in the contracts. AI only recommends a mode; Human/Safe sets it through a timelock.
• V1: Sepolia only, which is currently the only chain open for launches on the Identity.md platform. The Genesis Forge burns EMBER and mints Genesis PEPE in one transaction, and gives each qualified seat a single free claim. The Build Vault releases funds only on Safe signatures and returns unused funds by default. V1 has no Uniswap v4 hook, no upgradeable proxies, no Safe modules and no transfer fee.

Things you should know:

• Two findings change the plan:
  • If EMBER launches through Identity.md's standard path, the platform's rules fix supply at 1B and send 10% to the swarm. Trades in the pool then pay a 1.25% fee, of which 1% goes to the paying wallet. This needs to be a deliberate choice; the report compares it with a contracts-only launch.
  • Identity.md's own $IMD token uses sell-burns and staking, both of which EMBER's baseline rules out. The report positions EMBER as complementary to $IMD rather than a copy.
• Not found: I found no public record of EMBER, Genesis PEPE or Dream Hall, so those come from your brief. I also couldn't find primary sources for CLAUS or Identity Units, and SIMD's mechanics couldn't be verified, so none of the three is used as evidence. I used $IMD and the $ADAM launch as comparisons instead.
• Truncated brief: your brief cut off at "improve not only content, but also the…". I assumed it meant the process, the economic settings and the AI itself, and labelled that as an assumption.
• Checks: I re-checked every Identity.md number against the raw docs and token pages and verified the arithmetic in the revenue table. No independent reviewer has checked the content.
• Commit author: git had no author identity set, so I made the commit with your name and email passed for that one commit, without changing your git config.

The report closes with a list of open questions that block final numbers, including how a "Qualified Active IMD Seat" is defined, the current $IMD price, 3D production pricing, and a legal review before any Mainnet step.