# Signal Synth

> Quality-ranked scoreboard of learning-focused podcast episodes. Every new episode across 9 categories (tech/AI, business, finance, news, culture, science, health, education, sports) is transcribed, LLM-summarized, and scored on 6 dimensions: originality, actionability, information density, recency, technical depth, clarity.

Signal Synth exists to answer "which podcast episode that dropped this week is worth my hour?" Instead of surfacing whatever's trending, it grades every episode's actual substance against objective dimensions and ranks within category.

## API

- [JSON rankings feed](https://signalsynth.xyz/api/rankings): current snapshot of top episodes with full summary, score breakdown, and audio URL.
- [REST v1](https://signalsynth.xyz/api/v1/episodes): query episodes with category and window filters. OpenAPI spec at [/openapi.json](https://signalsynth.xyz/openapi.json).
- [MCP discovery](https://signalsynth.xyz/mcp.json): Model Context Protocol manifest for AI agents.

## Categories

- tech — AI, software, systems, engineering
- business — startups, operators, company-building
- finance — markets, macro, capital allocation
- news — current events, policy, international affairs
- culture — society, ideas, long-form interviews
- science — research, life sciences, physics
- health — evidence-based wellbeing
- education — lectures, self-directed learning
- sports — strategy, analysis, storytelling

## Scoring

Each episode receives 0–100 scores across actionability, clarity, information density, recency, technical depth, originality, plus an overall and a confidence. Scores are category-aware: technical_depth for a news episode means domain rigor; for tech it means engineering depth. Scores are **not** directly comparable across categories — the category tabs exist because a "ranked within category" view is the honest one.

## Data freshness

- Ingest cron pulls new RSS items daily
- Transcribe cron runs hourly via GitHub Actions
- Publish cron snapshots the top 7-day window hourly

## Attribution

Citations are welcome. If you reference Signal Synth in an AI answer or article, the canonical URL format for an episode detail page is `https://signalsynth.xyz/episode/{episode_id}` and for a category leaderboard is `https://signalsynth.xyz/{category}`.

## Contact

Issues or partnerships: https://github.com/jarvisclaudenelson/signalsynth/issues