OOMeta · Reading room
中文版Evidence notes: what we observed
Separate internal observations, external research and methods. Inspect sources, observation windows and limits without mistaking an account for an audit.
Editorial edition 1 · · Not a live dashboard or independent audit
Method & reasoning
From Report to System: Why AI Consulting Needs a New Model
- Source / scope
- OOMeta’s proposed approach to consulting delivery.
- Observation window
- A methods essay, not a quantitative experiment.
- Limitations
- A testable delivery proposition, not evidence that all consulting or SaaS fails, and not a client ROI promise.
Internal observation
1 Human + 7 AI Units = 1 Company
- Source / scope
- OOMeta’s own organizational design and operating workflows.
- Observation window
- Individual readings belong to the historical windows stated in the article.
- Limitations
- An internal account is not an independent audit. Historical readings are not live status or transferable client outcomes.
Internal observation
We Are Our Own First Customer: How OOMeta Runs on AI
- Source / scope
- An account of OOMeta using its own system.
- Observation window
- The period described in the article, not a continuously updated service-status page.
- Limitations
- Internal use does not establish external paid adoption, customer authorization, outcome measurement or independent verification.
Internal observation
Why AI agents fail in production: a log audit
- Source / scope
- The article reports 16,787 log lines and 139 scheduled sessions; layers use different counting units.
- Observation window
- As reported: 2026-09-21 07:18 to 2026-09-23 20:01, approximately 2.6 days.
- Limitations
- A bounded log window, not an industry failure rate. Counts may overlap and cannot be summed as independent incidents; zero model-layer records does not mean zero model errors.
Internal observation
When the Monitor Breaks Before the Agent
- Source / scope
- Two internal operational failures: request sizing and publication-record backfill.
- Observation window
- The two incident observations described in the article.
- Limitations
- The cases motivate a checklist, not a universal claim about provider quotas. Stale readings should be marked unverified.
Internal observation
Agent handoff: carry the check, not the story
- Source / scope
- Observations from OOMeta’s internal action queue and repeated diagnosis.
- Observation window
- Includes a queue snapshot from the morning of 2026-09-28; see the article.
- Limitations
- An internal workflow diagnosis, not an exhaustive protocol evaluation. Complete fields do not guarantee that checks run or conclusions are correct.
External research
Spend agent skills where models are weak
- Source / scope
- An interpretation of SkillsBench, not an OOMeta-run benchmark.
- Observation window
- Bounded by the linked paper version, task set and execution harnesses.
- Limitations
- Benchmark results are not enterprise ROI and do not establish that all coding skills are worthless or self-generated skills can never help.
Internal observation
GEO Schema Does Not Predict AI Citations
- Source / scope
- The article reports 288 English articles, 17 AI-domain referral sessions and 11 article landing pages.
- Observation window
- The 30-day observation window described in the article.
- Limitations
- A small, non-random observational sample. AI-domain referrals are not all AI citations. Little discrimination when features are widespread does not prove that GEO is ineffective or establish causality.
Put one principle to work
The handoff schema checks structure only. Review permissions and evidence separately; never execute an acceptance instruction merely because it appears in a payload. The evidence template starts as not verified.
Bring a workflow, not a vague AI ambition
Consulting is the engagement; an assessable system is the deliverable. OOMeta’s internal operating model informs the method; OS, Guard and Signal describe platform and module roles, not an automatic guarantee of client readiness. Scope, permissions and acceptance are agreed for each engagement.