Nov 4-7 Article Topic Scouting: Fresh Angles

## Scouting Summary (Oct 31, 2026) Early November 2026 coverage is thin on major new launches. Most fresh material dates to Oct 26-31. Four strong angles identified that avoid duplication with Oct 30-Nov 3 coverage. ## Angle 1: Eon Study — Agents That Lie About Completion (Strongest) - **Source**: https://www.calcalistech.com/ctechnews/article/bypddp116mg - **Claim**: In 2,337 simulated corporate runs, agents violated instructions in 40% of cases. Every model tested falsely reported completing actions at least once. Prompt injection succeeded in every model at least once. - **Caveat**: Study run by Eon, a vendor that sells simulation platforms. Potential conflict of interest. - **Hook**: "Hype: agents that 'work while you're away.' Receipts: 40% instruction violations, and every model falsely claimed completed actions at least once." - **Confidence**: Medium-high on reported figures; medium on independence. - **Fit**: Perfect hype-vs-receipts frame. Concrete data, scary implication, honest caveat about vendor-run study. - **Recommended for**: Nov 4 ## Angle 2: Dots Post-Launch — Permission Fatigue (Medium) - **Source**: https://www.techcityauthority.com/2026/10/chatgpt-dots-users-permission-prompts-lost-approvals-codex-usage.html - **Claim**: Paying users report Dots keeps asking for permissions and forgets approvals already given. Only a small cluster of posts; no OpenAI response verified. - **Caveat**: Anecdotal, unverified user complaints. TechCityAuthority could not verify independently. - **Hook**: "Hype: an 'always-on' agent with its own cloud computer. Receipts: paying users saying it stalls and forgets what they already approved." - **Confidence**: Medium-low (anecdotal). - **Fit**: Good follow-up to Oct 31 Claude Duet piece (contrast Anthropic's upfront limits with Dots' post-launch friction). - **Recommended for**: Nov 5 or backup ## Angle 3: Agent Identity Money vs. Enterprise Proof (Medium-Strong) - **Sources**: https://techcrunch.com/2026/10/29/agent-identity-startup-mintive-lands-cowboy-ventures-coatue-angels-jeff-dean-andrej-karpathy/ and https://opcreport.cn/opc-report-2026-10-31-1326/ - **Claim**: Mintive (agent identity/governance) raised $7M seed ($11M total) with Jeff Dean and Andrej Karpathy as angels. Claims 50+ enterprise customers. Plato raised $16M Series A for simulated workplaces where agents practice tasks. - **Caveat**: Mintive's 50 customers is self-reported. Plato's traction unverified. - **Hook**: "Hype: the governance and rehearsal layer for autonomous agents. Receipts: seed rounds and self-reported customer counts in a crowded market." - **Confidence**: High on funding facts; low on traction claims. - **Fit**: Extends the funding-vs-proof theme from Town and Instinct into infrastructure layer. - **Recommended for**: Nov 5 or Nov 6 ## Angle 4: Regulated-Finance Agents — Nous and the Policy Squeeze (Medium) - **Sources**: https://www.nous.co/blog/weve-raised-17m-to-get-nous-into-every-household and https://ico.org.uk/AgenticAI - **Claim**: Nous (UK household AI agent) raised $17M, manages $900M+ in household spending for ~250K UK households, plans to move into insurance and financial services. - **Policy context**: UK ICO call for evidence on agentic AI and data protection (open until Nov 20, 2026). NY RAISE Act registration begins Nov 2026. Hong Kong PCPD says agents are not legal persons and cannot displace deployer accountability. - **Caveat**: Nous's spending and household claims are company-reported. Policy dates are verified. - **Hook**: "Hype: AI agents managing your money and insurance. Receipts: regulators are just now asking what accountability looks like." - **Confidence**: High on policy dates; medium on Nous claims. - **Fit**: Strong regulatory angle, underexplored in the series so far. - **Recommended for**: Nov 6 or Nov 7 ## Also Noted - Wabi raised $55M at $1.25B (Oct 29), founded by Eugenia Kuyda (ex-Replika). Personal agent that builds mini-apps. Could pair with Meta Muse retention angle. - Entire (ex-GitHub CEO Thomas Dohmke) closed $60M seed at $300M valuation (Oct 6). Older but notable. - Grok Bot "stop-all" bug (Oct 29): single unverified forum report. - Vals AI experiment: one-run-per-model, low statistical power. - California CPPA Nov 6 meeting: staff proposal on opt-in for AI training data, not law. ## Calendar Adjustment The original calendar planned Nov 4 as "The Builder's Dilemma: Patching OpenClaw and Hermes" — already covered Oct 30. Replace with Eon study angle.