Voice review

Review this plan by voice

The voice agent explains any task, answers questions grounded in this plan, and records a brain dump. Voice feedback is read-only and cannot approve or execute work. Allow the microphone when your browser asks.

The explicit Sonnet rollback keeps the existing capsule-only protocol. A lost connection retries the same Sonnet engine twice and reloads the complete persisted conversation. It never changes models during a conversation.

Grok by xAI · Think Fast 2.0 · Orion · no fallback
Choose before speaking. Switching unloads the current runtime instead of changing its model mid-conversation.

Voice answers are review evidence. Approval, sending, spending, and production changes stay in the originating chat. Voice trouble in this page? Open the standalone voice page.

Source-grounded review only

Do the few things that change the week.

A recovered priority map from every user message in the mixed-up Codex task, the complete 60-minute call, and the existing August 14 inventory. Every task is still pending and remains REVIEW.

Open evidence-bound Markdown Open the detailed 36-task review
43deduplicated tasks
5active cards
340phone transcript turns covered
0tasks approved or launched
1
Operating rule: cost of delay, then one-task focus.

Choose by who is waiting, direct money or trust loss, health capacity, and the cost of another delay. Stay on one task until its reviewable outcome is terminal or a dated blocked note names the missing dependency and next check.

Five active cards

Time-bound, not a false universal rank. The call says lead generation first tomorrow. The organized source says overdue paid-client work first. Both remain visible.

Three guardrails

These shape work. They do not become three more active projects.

Health note: this page is non-clinical. Acute danger, severe confusion, or inability to plan safely stops the routine and shifts to qualified or emergency human support appropriate to the situation.

High-impact, more complex initiatives

Preserved without allowing system building to displace the active cards.

Fast proposal system

T-023 -> T-034 -> T-035

Start after the overdue backlog and one real proposal expose the actual bottleneck.

PPC measurement and portal

T-024 -> T-025 -> T-027 -> T-033

Start after verified conversion definitions and one pilot client.

Research visuals and journeys

T-026 -> T-028 -> T-030 -> T-031

Start after one reviewed manual example for one client and market.

Testing method

T-024 -> T-032

Start after one platform, decision, metric, and traffic constraint are named.

Free diagnostic

T-024 -> T-025 -> T-027 -> T-036

Start after measurement and access policies are reliable.

Personal activity context

C-004 + T-010 + T-012

Start after the daily loop proves what information is actually missing.

Immediate dependency view

Only extracted tasks and direct gates appear here. Arrows do not grant authority.

flowchart TD
  R["T-002: cost-of-delay rule"] --> P["T-009: evening plan and close"]
  R --> L["T-003: tomorrow lead block"]
  R --> H["T-022: verified client backlog"]
  S["T-006: sleep and recovery"] --> L
  S --> H
  P --> L
  P --> C["T-020: active Codex cap"]
  V["C-003: narrow voice continuity repair"] -. "only if voice blocks the close" .-> P
  L --> G["C-002: validate lead qualification and Gergő handoff"]
  H --> Q["T-023: test proposal workflow on one real case"]
  M["T-024: verify PPC measurement"] --> D["T-025: portal V1"]

Contradictions kept visible

The extraction does not force false certainty where Matt's own statements changed.

Lead generation tomorrow first versus paid and overdue client work first.
A ten-day physical tracker versus an 11-day reset.
Bedtime before 23:00 versus 21:00.
Stop broad voice tweaking versus narrow phone access and continuity repairs.

All 43 pending tasks

Search or filter by priority lane. Every card includes why, when, and the dependency or decision gate.

No tasks match this filter and search.

Coverage and boundary

Reviewed: every visible user-authored message in task 01a00642-6dee-78f2-88af-e10a2c2fa4fa, the complete one-hour call from 20:18:15 to 21:18:15, the organized August 14 reconstruction, and the existing 36-task plan. The call extraction covered 340 turns and produced 44 merged Matt-originated evidence items. Seven independent supplemental tasks were added. Assistant suggestions were not converted into decisions. No task status, external system, client message, deployment, schedule, or approval was changed.