Selected operating system / 01

Personal Chief of Staff / Daily Capture

A daily capture system for turning loose updates into structured tasks, decisions, blockers, reminders and confirmed operating records.

Status

Working prototype and active personal operating system

My role

Designed the capture loop, routing logic, Notion writeback flow, memory layer and confirmation model.

Commercial lens

Execution discipline, project hygiene and stronger follow-through.

Operating problem

Important context often starts as scattered input.

A note after a call, a thought between meetings, a project blocker or a follow-up can all contain operational value. Most tools ask the user to structure that information manually, so important context either gets lost or stays trapped in the wrong place.

The design principle is simple: capture naturally, structure intelligently and only write back after review.

Interface proof

Working capture surface

A live capture and review interface where loose updates are staged, separated and reviewed before anything is written back.

  • Protected writes
  • Review before save
  • Memory-aware
  • Review tray
Personal Chief of Staff capture and review interface showing staged updates, review tray, protected writes and memory status.

Watch walkthrough

Workflow

From loose signal to reviewed operating record.

The workflow makes each proposed destination visible before anything is written back.

Operating artefact

Messy input becomes reviewed writeback.

The system separates meaning, proposes destinations and waits for approval before anything moves into the workspace.

Messy input
Voice-note text

"Blocker on the project. Follow up next week. Status may need to change."

Project update Follow-up
Extracted meaning
  • Blocker Make visible
  • Task Follow up next week
  • Status Hold for review
Preview
Writeback proposal Needs review
Destination
Project record
Objects
Task, blocker, memory
Approve Edit Reject
Confirmed writeback
Outcome

Approved records move into Notion and memory. Nothing writes silently.

Notion writeback Task Blocker Memory

Architecture / stack

The model is useful because the operating layer is controlled.

The AI layer structures and routes the signal, but the system around it controls preview, confirmation, memory and delivery.

01

Capture surface

Typed updates, voice-note text and quick capture moments.

Input where the work appears
02

Reasoning and routing

The AI layer separates tasks, blockers, decisions and project movement.

Model-supported structure
03

Preview and confirmation

Proposed records pause for review, edit or rejection before writeback.

Human approval gate
04

Workspace and memory

Notion receives confirmed records; Supabase stores useful durable context.

Operating record
05

Build and deployment

GitHub tracks the system and Vercel hosts preview and production surfaces.

Delivery layer

Stack surfaces

A compact build across the places work already lives.

Interface, workspace, data layer, repository and deployment surfaces support the operating story while keeping the public view focused on the workflow.

Local development

Build and test application behaviour before deployment.

GitHub repository

Track implementation history and development direction.

Preview environment

Review changes safely before promoting them.

Vercel production interface

Live version of the capture and confirmation surface.

Supabase project

Persistent memory and structured data for durable context.

Notion operating workspace

Receives confirmed records and supports project follow-through.

Governance

Human confirmation before writeback.

The system prepares the operating consequences of an update. The operator decides.

System prepares

  • Captures loose updates
  • Separates tasks, blockers and decisions
  • Prepares a reviewed writeback

Operator decides

  • Approves what is saved
  • Edits context and priority
  • Owns decisions and follow-through

A note-taking app stores the sentence. This system prepares the operating consequences of the sentence.

Where the value shows up

Less execution drag, clearer follow-through.

The same pattern appears in team settings: updates are fragmented, decisions are easy to lose and follow-through often depends on manual discipline.

This system reduces that drag by turning unstructured input into usable records, visible next actions and reviewable operating context.