Personal AI Delegate (PAID)
Jimmy's self-built "personal AI delegate" — user-authorized, user-accountable, information belongs to the user; the AI version of the principal-agent relationship in the same category as lawyer / accountant / doctor.
1. Core Product / Service
PAID is a personal delegate agent that runs in IM, letting the AI handle asynchronous collaboration with other people or other agents within scopes the user has authorized.
- Three-tier response modes: direct answer / request authorization / decline; initial human-machine ratio 50/50, target reaching 80%+ auto-answer.
- RAG input layer: locally authorized information (owner.json / persona.md / sop.md / counterparties/) + web search tool combination.
- Multi-platform IM: Phase 1 doubles down on Slack, Phase 2 bridges across organizations, Phase 3 fans out to Lark / Telegram / Slack multi-platform.
- Local Dashboard: approval queue, outbound queue, audit log visualization (v0.5 is CLI for now, dashboard is a W2+ TBD).
- Stack: runtime runs on hermes-agent, IM gateway reuses openclaw, peer review references review-agent; model routing via openrouter.
- Deployed form: hermes plugin
paid-v1, MIT open-sourced from 2026-05-02. Install = clone to~/.hermes/plugins/paid-v1. - Four-layer pipeline: L1 input regex / L3 approval loop (slash command version, no card button) / L4a + L4b output regex (observer-only) / L4c LLM post-check and L4d source attribution are listed as W2 TBD.
2. Target Users & Pain Points
Primary persona: Seniors managing Juniors — founders / research leads / training mentors / senior professionals. The pain is being drowned in low-density queries ("how do I do this", "how did you handle that before"), yet unable to outsource directly because the cost of a wrong answer falls on them.
Jimmy himself (dogfood #1): after exiting crypto, he is back in a state of high-frequency IM queries with demos built but never pushed. PAID both screens queries for him and forcibly breaks the "are they worthy of me" filter pattern, as a contract project for May external exposure.
Relationship to ai-human-hybrid / human-in-the-loop-ai: the core mechanism is "auto within scope + escalate when out of bounds", not full autonomy and not pure copilot.
3. Competitive Landscape
| Project | Form | Difference vs PAID |
|---|---|---|
| Rabbit R1 | Standalone hardware + LAM | Device form, OS-level agent; PAID is a persona agent inside IM, no hardware |
| Humane AI Pin | Wearable (already stalled) | Negative case: hardware paths divorced from existing IM/OS are hard to land |
| General ChatGPT / Claude.ai | User-invoked chatbot | Lacks principal-agent authorization semantics, lacks proactive outbound, lacks approval queue |
| Cosmos / various AI assistants | "Concierge" productivity copilot | Heavy on calendar / email management; PAID is heavy on "replying to others on your behalf" outbound delegation |
| review-agent | Peer review / review scenarios | Same-stack sibling product, but review-agent is single-task, PAID is a resident persona |
Differentiation anchors: (1) runs in the user's existing IM rather than a new app; (2) explicit principal-agent legal metaphor rather than "AI assistant"; (3) output layer regex + LLM dual gate — out-of-bounds, prefer placeholder/decline over making things up.
4. Unique Observations
- Design tradeoffs: v0.5 cut the dashboard, card button, auto counterparty discovery, FTS5, and web retrieval, forcing the minimum closed loop on the J3 approval loop (slash command suffices) — prove the 3-state pipeline runs end-to-end before expanding.
- Key technical risk (design/01_review_decisions §3.3): whether relying on
pre_llm_callto inject context so that the LLM proactively says placeholder / decline is reliable; if unreliable, fall back to apre_gateway_dispatchhook that rewrites the message. At that point Layer 4a/4b must upgrade from observer to enforcer. - Anti-consensus success criteria: May success = at least 5 of 7 falsifiable indicators checked + sustained daily reflection ≥ 20 days, not product perfection. Emphasizes "whether it has been pushed in front of users" over "whether it's done well".
- Writing pipeline: daily reflection (
daily/YYYY-MM-DD.md) + Sunday sync (sunday-sync.sh) + Sprint log (SPRINT_LOG.md), forcing audit on own output.
5. Financials
- Budget: $5000 / month (May, self-funded), 40h / week invested; no fundraising, no hires.
- Monetization path: no charging in May, the goal is "be used" and "be written about". Decide end of month whether to enter a paid experiment.
- Infrastructure cost: models routed via openrouter, IM gateway reuses the existing openclaw VPS (sunk cost), so marginal cost is mainly LLM tokens + own time.
6. People & Relationships
- Builder: Jimmy (solo development, dogfooded for self use)
- Connected components: hermes-agent (runtime), review-agent (peer-review reference, same-stack sibling), openrouter (model routing)
- Related modules: claude-code-sessions (development process dependency), deep-research-workflow (info organization), ai-human-hybrid / human-in-the-loop-ai (design philosophy)
- Testers: 5-person invite pool (W0 outputs/testers_w0.md), targeting first-batch feedback in W1
Sources
- local: 2026-05-01-summary.md
- local: memory/project_paid_may (W0/Sprint #1+#2 progress, design tradeoffs, May path)