SIFT

The difficult moment between planning and doing.

I built SIFT around my own experience of rumination and overthinking. The problem was not simply a lack of goals: overwhelm, avoidance, and familiar patterns could interrupt the move from intention to action.

SIFT explores that execution gap through three lenses: fading motivation, moments of mental paralysis, and reactive thoughts that may point to recurring beliefs or behaviours.

Give an unstructured thought a structure.

Voice dictation or text input captures a thought when friction occurs. LLM-powered serverless endpoints parse the entry into suggested emotional drivers, cognitive polarity, and possible points for change.

The seven-layer model organises the reflection:

These layers are the product’s interpretive framework. The AI output offers hypotheses to reflect on, rather than established facts about a person.

The Neural Hinge.

This feature presents two possible directions from an active loop: Path A explores a change in response; Path B examines the possible cost of repeating the habitual response.

The purpose is to make a decision point easier to notice.

A mirror for recurring themes.

The Stoic Identity Mirror and synthesis features aggregate recurring nodes across the seven layers. Second-person reflections and pattern cluster reports help surface themes across entries.

These are AI-generated reflections, not clinical diagnoses. Their value depends on whether users recognise, question, and find them useful.

Make the next step small enough to start.

Side Quests translate an identified point of friction into a discrete action intended to take about two minutes. The design goal is to reduce initiation friction and create momentum.

Lower friction is a design intention; sustained behavioural change is not claimed as a measured result.

Exploring the interface

Built across interface, state, and intelligence.

• Frontend: Next.js 15 App Router, TypeScript, Tailwind CSS 4, and Framer Motion.
• State and sync: Zustand with optimistic updates and local persistence.
• Database and authentication: Supabase PostgreSQL, row-level security, realtime subscriptions, and OAuth integration.
• Intelligence: Gemini API orchestration through serverless edge routes.

I built SIFT entirely on my own, connecting thought capture, structured reflection, and action suggestions within one experience.

The moment SIFT was designed for was also the moment I least wanted to use SIFT.

I had designed a cognitively expensive interaction for a cognitively overloaded moment. The intervention also lived inside one of the environments already competing for attention.