Reflex

A passive learning system that watches what you consume and turns it into targeted practice for the skills you are actually trying to build. Install the Chrome extension. It handles the rest.

The core insight: people consume content constantly and retain almost none of it. Not because they're lazy — because there's no feedback loop between consuming and practicing. Reflex closes that loop automatically, every day, without requiring any behavior change from the user.

---

Problem statement

People trying to build skills (communication, technical thinking, business judgment) consume content daily — YouTube, articles, podcasts, newsletters — and walk away with almost nothing actionable. The gap isn't motivation. It's the absence of a personalized practice loop tied to what they actually consumed.

ICP (who exactly has this problem)

• Professionals actively working on a specific skill (communication, negotiation, structured thinking, technical depth)
• People who consume content intentionally but feel like it doesn't stick
• Anyone doing deliberate practice on a personal development track
• Initially: non-native English speakers in professional settings building communication fluency (connects to [[structured-speech-coach]])

Market and TAM

Learning and skill-building is a massive consumer and B2B market. The angle here is not another course platform — it's the layer that makes all the content you already consume actionable. No direct competitor does passive capture + personalized practice prompts tied to your actual consumption.

Alternatives (how people solve this today)

• Anki / spaced repetition — manual, requires card creation, not tied to organic consumption
• Readwise — surfaces highlights, doesn't generate practice prompts
• Recall — saves content, no practice layer
• Course platforms — structured but disconnected from daily content habits
• Journaling / reflection — manual, low compliance

How it works

Capture layer (Chrome Extension, Manifest v3):

• Passively logs every page visited for more than 30 seconds: title, URL, time spent, timestamp
• No clicks, no tagging, no behavior change required
• POSTs to a local server (v1) or cloud API (v2)

Processing layer (multi-agent pipeline):

• Orchestrator agent reads today's captures and decides which skill layer is most relevant to the content consumed
• Routes to a narrow skill-specialist agent (structured thinking, listening, storytelling, negotiation, problem solving)
• Specialist generates 1-3 targeted practice prompts that reference the actual content by name

Delivery layer:

• Evening email (8pm) with 2-3 practice prompts
• Each prompt specifies: the skill being practiced, the content it came from, and the exact output format (60-second video, LinkedIn post, one written sentence, pre-meeting frame)
• v2: push notifications, calendar-timed prompts before relevant meetings

The skill taxonomy (v1 uses 6-layer communication stack as proof of concept):

• Layer 1: Physical voice
• Layer 2: Prosody
• Layer 3: Structured thinking / BLUF (current focus — communication is designed, not reactive)
• Layer 4: Pause placement
• Layer 5: Listening
• Layer 6: Strategic communication

Reflex is layer-aware: prompts are biased toward whichever layer the user is currently working on.

GTM angle

Start personal (dogfood it yourself). Chrome extension as the onboarding artifact — install and forget. First cohort: people who are already doing deliberate practice and frustrated that their content consumption doesn't compound. Content angle: "I built a tool that turns everything I watch into practice prompts. Here's what happened after 30 days."

Riskiest assumption

That AI-generated practice prompts tied to consumed content are actually better than generic prompts — and that users will act on them. The output quality depends on whether the specialist agents generate prompts that feel specific and earned, not generic. If they feel generic, nobody does them.

Open questions

• What skill domains beyond communication work well? (Technical skills? Business judgment? Language learning?)
• Does the local server model work in practice, or does it need to be cloud-first from day one?
• Is evening email the right delivery mechanism, or is a mobile push notification better compliance?
• What's the monetization model? Subscription? Per-skill-track?
• Can this become a B2B product for companies running L&D programs?

Current build status

• v1 built and tested: Python + GitHub Actions, Recall API as capture layer, evening email delivery, skill-specialist agent hierarchy
• v1.5 in progress: Chrome extension replacing Recall (passive capture, no dependency on Recall subscription)
• v2 planned: cloud API, push notifications, calendar integration, completion tracking