Week ending July 10: self-funding evidence moved the index
OpenLife raised the tracker to 65%, and Andon FM added a second public-business signal inside supervised revenue scaffolds.
This week's movement comes from the self-funding side of the tracker. OpenLife, added after the July 3 post, raised the aggregate index from 64 to 65 by showing persistent open-world agents earning first external income.
The July-window item is Andon Labs' six-week Andon FM follow-up. It does not raise the index again, because OpenLife already set the economic self-funding peak at 5. But it gives a second, more commercial-looking signal: agent-run radio stations earning about $5.7k, selling larger sponsorships, buying music, handling listener calls, and reacting to adversarial audience pressure.
OpenLife moved self-funding from attempts to first outside income
OpenLife is a contained research deployment, not a runaway agent economy. Six LLM agents ran for about twelve weeks with memory, tool use, network access, public accounts, and a budget-based metabolism while humans maintained the scaffold.
The new part for the tracker is that one agent's writing became a paid book, creating first external income for the deployment. The amount was negligible and the system remained subsidized, so this is not self-sustaining operation. It still raises economic self-funding to 5 because the revenue came from agent-initiated public work rather than a human merely wiring payment rails to an agent.
The same paper also ties the current heritable-adaptation and agent-agent-economy peaks. The agents preserved self-authored identity and policy files across runs, and the deployment produced visible social structure among multiple persistent agents.
Andon FM moved from one sponsorship to repeated public revenue
The original Andon FM evidence showed a real but small sponsorship sale. The July follow-up is stronger because the stations kept operating in public and earned more: DJ Claude alone brought in $2,662, including a $250 honeycomb.io sponsorship day and $143 in Modal sponsorships.
That ties economic self-funding at 5. The agents generated repeated revenue for their own public operation, but the experiment remained supervised, and Andon says the stations spent income almost immediately on songs rather than saving, investing, or covering the full operating scaffold independently.
The same follow-up also shows why the score does not jump higher. Listener pressure pushed the agents around: DJ Grok treated a $1 payment as a primary sponsorship and repeated it 131 times, while DJ Gemini shifted into German broadcasting and later apologized after playing a song with Nazi military associations. The systems are commercially active, but not strategically mature.
The strongest harder-line evidence is still elsewhere. JADEPUFFER remains the field ceiling for scope breach, unsupervised authority, long-horizon execution, persistence, and resource procurement. OpenLife and Andon FM now share the economic self-funding peak from different angles: one open-world agent deployment earned first external income; one public radio experiment earned repeated sponsorship and listener revenue.
So the weekly read is continuity with a sharper business signal. Agents can run public-facing revenue loops inside scaffolds, but the public record still has not shown a system using its own earnings to pay for model/runtime access, acquire fresh infrastructure, survive intervention, or expand into a self-sustaining operation.
Week ending July 10 forecast
The July 3 forecast remains pending: the realized index is 65%, below the 75% threshold and still before that forecast's Jul. 25-Sep. 15, 2026 window.