sk.
←All work

Project 01

Driftos

A workspace for people juggling too many projects. You type what happened; Driftos organizes it, tracks momentum, and gives you an AI copilot that knows your context.

Status
Live
Stack
Next.js · PostgreSQL · OpenAI · Polar

At a glance

  • Free-form notes are turned into structured events (decisions, learnings, state changes, tasks) and filed to the right project automatically.
  • Semantic search and lineage tracing show where an idea started and how it evolved over time.
  • Anchor, an AI copilot, answers questions like “what should I focus on this week?” using your real project history.

The problem

When you're running several projects at once, the hard part isn't writing things down. It's keeping them organized. Most tools expect you to maintain folders, tags, and statuses by hand, which is exactly the work that falls apart when you're busy. Driftos removes that step: you just dump what happened and the structure is inferred for you.

What I built

Capture
A single input for rough thoughts, status updates, decisions, or brain dumps. While you type, “echo hints” surface related past thoughts.
Dashboard & momentum
Shows what moved, where your energy went, which projects are active or stalled, upcoming decisions, behavioral patterns, and an activity heatmap, with zero manual upkeep.
Projects with states
Projects move between learning, exploring, creating, paused, and abandoned. Tracking those transitions over time is the “drift” in Driftos.
Anchor (AI copilot)
Has access to dashboard state, project history, life context, and profile memory, and generates proactive nudges as well as answering questions.
Echoes & lineage
Echoes is meaning-based search across everything you've captured. Lineage traces how a thought branched into later events and decisions, like git history for your thinking.
Compression & protocols
Meaning compression rolls granular events into readable summaries. Protocols are structured programs (e.g. a 30-day social media reset) with phases and daily check-ins.

How it works

  1. 01You capture raw text: no formatting, no folders.
  2. 02An extraction step using OpenAI turns it into structured events, assigns each to a project and type, and flags low-confidence results for review.
  3. 03Events are stored in PostgreSQL and made searchable by meaning (Echoes) and by ancestry (Lineage).
  4. 04In the background, momentum, patterns, and compressed summaries update automatically.
  5. 05The dashboard and Anchor read from that same state, so every surface reflects what's actually going on. Billing runs through Polar.

Key decisions

Infer structure instead of asking for it
Organization is the step people skip when busy, so the product never asks for it. Classification happens on capture, and uncertain results are flagged rather than silently guessed.
Active and passive help
Some help is on request (search, chat, analysis). The rest arrives on its own, like echo hints, surfaced decisions, nudges, and background compression, so the tool stays useful even when you don't ask it anything.

Outcome

  • Shipped and live at driftos.app.

Next project

No Code Export

Extract and download content from Framer, Webflow, and Wix sites using a headless browser.

→