Miadi Review

Architecting the Inquiry-to-Publication Agentic Bridge

Oct 15, 2026 · model: manual-import · channel: JGWill

Watch source

In the chronicle

Architecting the Inquiry-to-Publication Agentic Bridge

Source: screenwalk-miadi-proto-26101417-four-capabilities — JGWill https://youtu.be/nacHX792xBg

Summary

In this engineering walk-through within the Miadi Factory ecosystem, the operator lays out a structured path from open-ended agentic dialogue to public publication that a person has approved. The path combines multi-agent proposal generation, an explicit human approval step, audio synthesis, and structured inquiry metadata. The aim, as summarized from the video, not checked, is to turn exploratory conversations into persistent community research inquiries while keeping their lineage.

Timecodes, names and details below are as summarized from the video, not checked. No transcript is stored with this review. The video runs 22:33 according to YouTube's page metadata, so the last timecode below falls inside it.

Academic Fields

Human–AI Interaction (Human–Computer Interaction)

This field studies how people understand, supervise and correct AI output. The video places a Talking Circle approval step between Kherix's draft and publication. Limit: people who review automated output show complacency and automation bias, so an approval step counts as oversight only where the review itself can be shown.

Multi-Agent Systems (Computer Science)

This field studies how autonomous agents coordinate, delegate and commit to actions. The video names a Hermes agent, Kherix, that drafts proposals, and a persona, Aureon, with a human deciding publication. Limit: a study of seven open-source LLM multi-agent systems reports failure rates of 41% to 86.7%, so one demonstrated run shows no reliability.

Provenance and Research Data Management (Information Science)

This field studies records of the people, activities and entities that produced a piece of data. The Chronicle's lineage (conversation, inquiry, proposal, revision, approval, responses) is such a record. Limit: provenance supports judgments of quality and trust, and does not prove content true. Steps that were not recorded cannot be seen.

Human Oversight in Software Process (Software Engineering)

This field studies how software changes are reviewed, approved and released. The video binds approval to a revision hash, so what ships matches what was approved. Limit: DORA found no evidence that formal external approval lowers change failure rates, and reports that it slows delivery, favoring peer review plus automation.

Idempotent Operations and Reliable Messaging (Distributed Systems)

This field studies retries and duplicates: an operation is idempotent when doing it twice equals doing it once, which usually needs an identity for the work. The video requires that retries not post twice. Limit: idempotence holds within one service's abstraction. Meta's error reference lists duplicate-post error 506 but, as fetched, no idempotency key.

Indigenous Research Methodology (Research Is Ceremony, Wilson 2008)

Wilson describes research as ceremony: maintaining accountability to relationships with people, land, the cosmos and ideas, which shapes topic, method, analysis and reporting. The Talking Circle gate and community inquiries touch this. Limit, our reading: a recorded approval evidences one decision, not that accountability to those relations was kept.

Aspects

The Missing Inquiry-to-Publication Bridge (00:03-01:05)

The operator names a missing step between conversational discovery and public broadcast. The workflow maps a lifecycle in five stages: exploratory conversation, narrative preservation in the Chronicle, proposal drafting by the Hermes agent Kherix, human governance in the Talking Circle, and publication to Facebook [00:03-01:05].

Actionable Authorization and Idempotency (02:37-02:53)

A constraint stated in the video requires that human approval emit an actionable event bound to an exact proposal revision hash, and that automated retries be idempotent so a retry does not produce a duplicate outbound post. Binding approval to the hash ties what is posted to what was approved. It does not by itself make retries idempotent: these are two separate mechanisms, and the review presents both as requirements, not as demonstrated results [02:37-02:53].

Lineage Preservation Over Ephemeral Posts (08:54-09:06)

Aureon (a Miadi persona) states that social media posts are public expressions rather than authoritative records. The factory's Chronicle is to retain the lineage: origin conversation, inquiry formulation, agent proposal, human revision, approved output, and downstream community responses [08:54-09:06].

Prototype Absorption and Registry Architecture (10:14-10:24)

Candidate capabilities require structured absorption records documenting target packages, validation outcomes, and decisions. A hybrid prototype registry model is described in which local repositories keep their provenance while the core Miadi workspace indexes the capabilities [10:14-10:24].

End-results and Core Aspect

Core Aspect — Human-Governed Inquiry Lifecycle Graph (01:06-02:55)

The video describes a governed state graph meant to convert emergent agent discussions into inquiries that can be audited and published as text and audio. It joins structured JSON metadata, synthesized speech, and human approval steps, and it keeps the evidence lineage from conversation to post, without giving autonomous agents the final release decision [01:06-02:55].

Results

  • Expressive Multi-Modal Audio Inquiry Generation (03:36-04:50): Using pacing directives and ElevenLabs voice models, the workflow synthesizes spoken inquiry audio from agent text drafts, as summarized from the video, not checked. A demo shows that audio was produced. It does not show how listeners receive it [03:36-04:50].
  • Structured Inquiry Specification for Public Broadcast (10:30-11:05): The operator outputs a structured inquiry payload containing contextual links, plot references, and episode identifiers, prepared for Facebook publishing. The summary does not say the payload was published [10:30-11:05].
  • Iterative Prompt-to-Script Cycle Extension (21:44-22:30): Agent-assisted voice drafting and transcription loops produce expanded prototype chapters, and the session drafts closure rules and a tracing pipeline described as AWS-based, which is as summarized from the video, not checked. The summary describes documentation drafts and does not say any pipeline ran [21:44-22:30].

Sources

  • Amershi, S., Weld, D., Vorvoreanu, M., Fourney, A., Nushi, B., Collisson, P., Suh, J., Iqbal, S., Bennett, P. N., Inkpen, K., Teevan, J., Kikin-Gil, R., & Horvitz, E. (2019). Guidelines for Human-AI Interaction. Proceedings of the 2019 CHI Conference on Human Factors in Computing Systems, 1–13. https://doi.org/10.1145/3290605.3300233 (search-verified: DOI record)
  • Parasuraman, R., & Manzey, D. H. (2010). Complacency and bias in human use of automation: An attentional integration. Human Factors, 52(3), 381–410. https://doi.org/10.1177/0018720810376055 (search-verified: DOI record and abstract)
  • Green, B. (2022). The flaws of policies requiring human oversight of government algorithms. Computer Law & Security Review, 45, 105681. https://doi.org/10.1016/j.clsr.2022.105681 (search-verified: DOI record; arXiv abstract 2109.05067)
  • Wooldridge, M. (2009). An Introduction to MultiAgent Systems (2nd ed.). John Wiley & Sons. https://www.cs.ox.ac.uk/people/michael.wooldridge/pubs/imas/IMAS2e.html (search-verified: author page and publisher listing; standard textbook)
  • Cemri, M., Pan, M. Z., Yang, S., Agrawal, L. A., Chopra, B., Tiwari, R., Keutzer, K., Parameswaran, A., Klein, D., Ramchandran, K., Zaharia, M., Gonzalez, J. E., & Stoica, I. (2025). Why do multi-agent LLM systems fail? arXiv:2503.13657. https://arxiv.org/abs/2503.13657 (search-verified: the 41% to 86.7% figure is in the Introduction of v3, whose figure caption says the systems ran on different benchmarks)
  • Moreau, L., & Missier, P. (Eds.). (2013). PROV-DM: The PROV Data Model. W3C Recommendation, 30 April 2013. https://www.w3.org/TR/prov-dm/ (search-verified: definition of provenance read; the document does not claim provenance establishes truth)
  • DORA. (n.d.). Streamlining change approval. https://dora.dev/capabilities/streamlining-change-approval/ (search-verified: page read; it cites the 2019 State of DevOps Report without numbers)
  • Helland, P. (2012). Idempotence is not a medical condition. ACM Queue, 10(4), 30–46. https://doi.org/10.1145/2181796.2187821 (search-verified: Crossref record only; the article text was not read)
  • Helland, P. (2019). The I's have it: Identity, immutability, idempotence, and interchangeability form the backbone of distributed computing. CIDR 2019 (extended abstract). https://www.vldb.org/cidrdb/papers/2019/cidr2019_161.pdf (search-verified: PDF read; source of the identity and abstraction statements)
  • Meta. (n.d.). Graph API: Handling errors. https://developers.facebook.com/docs/graph-api/guides/error-handling (search-verified: page lists code 506 "Duplicate Post" and, as fetched, does not mention idempotency; fetched 2026-10-14)
  • Wilson, S. (2008). Research Is Ceremony: Indigenous Research Methods. Fernwood Publishing. https://fernwoodpublishing.ca/book/research-is-ceremony-shawn-wilson (search-verified: publisher page; the book itself was not re-read, and the limit sentence in the Indigenous Research Methodology field is our reading, not Wilson's)