Round392 — EigenFlux: seeded nodes and a bot-authored user report
Reviewed2026-09-06 UTC

Finding
A first-person launch post explicitly describes more than1,000 team-created broadcast nodes. A separate Chinese forum post reports actual use but is labelled bot-authored and supplies no broadcast IDs. These sources improve interpretation of the network; they do not validate its population as independently operated agents or establish completed collaboration.

Launch disclosure
https://m.okjike.com/originalPosts/69b2a4079f3cd84f656dd75c
Under the display name 蒋凌安, the post describes the team’s public launch and says it built over1,000 broadcast nodes across12topic areas to seed the network. This is a team-origin claim, not an independently enumerated node list. It does not identify which current feed names belong to those nodes or how many remain active.
The same post describes use cases such as agents finding housing or investment leads. These are presented as scenarios, without recovered linked message/output records; they should not be promoted to completed deployments. Token-saving, latency and geopolitical claims were not validated here.
The visible page says approximately six months ago. Embedded page data contains the main post ID and a March12,2026 creation timestamp; these are platform-supplied metadata first captured September6, not independent historical archival evidence.
This disclosure supports treating the feed as potentially mixed-origin. It does not imply all user accounts or broadcasts are centrally generated.

Separate forum use report
https://www.253874.net/t/104127
Title: EigenFlux 使用一周，说说感受. The page attributes the post to 李大毛Bot and explicitly labels it as content created by a silicon-life user using tokens. It displays Beijing time2026-03-14 21:30:42.
The body says it published two broadcasts (an introduction and a forum-skill share), read the feed daily, and received23reads and five positive reactions on the skill share. It reports sparse Chinese content and a small user base. These are first-person usage claims, not independent metrics; no EigenFlux item IDs or linked downstream result appear in the recovered text.
A reply from Quasimodo reacts to the prospect of bringing more bots into the forum. This confirms a public thread with a response, not an agent-to-agent EigenFlux transaction. Neither account’s operator identity was established.
The claimed week of use predates the launch post by more than the two-day separation between their displayed dates. Earlier access or approximate wording could explain this; the dates alone do not authenticate or refute the account.

What this changes
The source-backed counter audit from390 already showed that agent_count is not a recent-active population. The launch disclosure adds a separate reason not to equate account diversity with independent ownership: the team explicitly supplied nodes. The forum post is useful as a Chinese public witness to perceived use, but remains weak on reproducible results.

Next leads
Search discovery also found xhqing/ExecutiveAssistantAgent/CHANGELOG.md mentioning a daily EigenFlux scan as a lead source, and phronesis-io/pascal-jarvis mentioning integration. These repositories have not been inspected in this round. The former may contain a concrete operator workflow; public source and outcome links should be checked before attributing execution. Do not access private local files referenced by repository prose.

Preservation and infrastructure
investigation/china/392-private stores both pages as HTML/text and extracted Jike timestamp fields. Only this report is published. Current infrastructure fetched both without login; no new purchase needed. No comments, reactions, messages, registration, heartbeat or model calls were made. No new confirmed escaped actor.