Selected work

Personal intelligence system

Clockwork

A private life-operations system that makes deliberate progress visible.

Role
Product, design, and engineering
Status
Private beta
Clockwork’s public Today view, with the next step, quick logging controls, and a read-only guest notice.
Actual interface / Guest preview Captured September 26, 2026 · Read only

A day, with its context intact.

01 / ORIENT

Know what comes next.

Today puts the next required step beside quick logging for food, water, weight, and notes.

02 / CONNECT

Keep the week in view.

Plan, Train, and Insights sit alongside Today, keeping daily execution and longer-term context within reach.

03 / REVIEW

Keep changes deliberate.

The product’s AI workflow prepares suggestions for review before saving. Guest access keeps editing disabled.

The screenshot shows the public guest interface. Editing and private body metrics are unavailable in this view.

A private life-operations system that makes deliberate progress visible.

I designed and engineered the product, data model, Cloudflare platform, and controlled Grok intelligence layer across training, nutrition, recovery, skincare, and adherence.

01 / Problem

Turn scattered routines into one operating system.

Unify daily execution and longitudinal understanding across training, meals, hydration, supplements, skincare, recovery, and body trends without creating competing sources of truth.

02 / Context

Daily decisions only become useful when the history stays coherent.

Personal routines usually fragment across notes, trackers, memory, and disconnected dashboards. The useful product is not another log; it is a coherent operating layer that knows what applies today, preserves the history, and explains the signal without inventing it.

03 / Constraints

Intelligence cannot outrun trust.

  • 01Keep application code authoritative for arithmetic, validation, permissions, and writes while the model interprets bounded evidence.
  • 02Protect private routine and body data with authenticated storage, explicit review, provenance, and a read-only guest boundary.
  • 03Make capture fast enough for daily use without separating Today, Trends, and Intelligence into conflicting systems.

04 / Approach

One source of truth from capture to interpretation.

Build one structured operating model beneath the experience: protocols resolve the correct daily branch, the timeline captures execution, D1 preserves canonical records, deterministic analytics calculate the evidence, and Grok works through bounded tools to interpret or propose changes for review.

Product systemsSvelte + TypeScriptCloudflare Workers + D1Controlled AI orchestration

05 / Outcome

Private beta, with a public guest preview.

Clockwork is in private beta. Its public guest view exposes Today, Plan, Train, and Insights while disabling edits and concealing private body metrics. The preview makes the product explorable while keeping personal records and editing protected.

06 / Lessons

What the system keeps clarifying.

  1. 01

    Trustworthy AI products need application-owned truth, bounded tools, explicit review, and durable audit trails.

  2. 02

    Low-friction daily execution and deep longitudinal analysis only reinforce each other when they share the same data model.