83
episodes indexed
Podcasts are sound. AiPod turns every episode into structured, cited data — ready for any AI agent to query.
# one call — grounded, cited context curl 'https://aipod-two.vercel.app/api/context\ ?q=What+does+Chamath+think+about+AI+capex'
# remote MCP — no install, no API key claude mcp add --transport http aipod \ https://aipod-two.vercel.app/mcp # then, from any MCP client: aipod_context { "q": "Chamath on AI capex" }
# plain HTTP — works from anything
GET /api/context?q=Chamath+on+AI+capex
Host: aipod-two.vercel.app
Accept: application/json
{
"route": "person-stance",
"covered": true,
"claims": [{
"holder": "Chamath Palihapitiya",
"claim": "The binding constraint across the AI
market is power, not demand.",
"quote": "Everything in this market is power
constrained…",
"timestamp": "8:29",
"episode_title": "OpenAI Misses Targets…"
}, …46 more, every one cited ]
}
Five tools, one API. No audio to process, no transcripts to chunk, no model to run.
One question in, cited context out. Every claim comes with the speaker, the quote, and the timestamp.
curl 'https://aipod-two.vercel.app/api/context?q=How+has+the+Nvidia+view+changed' { "route": "cross-episode", "covered": true, "claims": [ …attributed, quoted, timestamped ], "coverage": { "podcasts": [ …knowledge cutoffs ] } }
Keyword search across every claim, fact, and topic — all shows at once. Every hit carries its source.
curl 'https://aipod-two.vercel.app/api/search?q=Nvidia+selloff' { "claims": [ …every hit carries holder + quote + timestamp ], "topics": [ …episode segments that match ], "facts": [ …checkable statements, sourced ] }
One page per person, company, or idea — 1565 so far. Everything the shows ever said about it, in order.
curl 'https://aipod-two.vercel.app/api/entity/OpenAI' { "name": "OpenAI", "first_seen": "2026-05-01", "last_seen": "2026-09-04", "claims": [ …132 of them, chronological, across every show ], "predictions": [ … ], "facts": [ … ] }
393 dated, falsifiable calls on file. Who said it, the deadline, the quote.
curl 'https://aipod-two.vercel.app/api/episode/<id>' | jq '.predictions' [{ "holder": "Brad Gerstner", "prediction": "…", "timeframe": "by end of 2026", "quote": "<verbatim>", "timestamp": "41:03" }]
Full episodes as structured JSON: summary, segments, claims by speaker, quotes. Not audio.
curl 'https://aipod-two.vercel.app/api/episodes' # → id, show, date, speakers, signal counts — for all 83 curl 'https://aipod-two.vercel.app/api/episode/<id>' # → summary, segments, claims by speaker, # predictions, facts, notable quotes
Transcription, attribution, grounding checks — all done before you ask. Your agent gets finished data, not homework.
Hours of audio become records: claims, predictions, facts, quotes.
Every claim is pinned to the person who said it. Never to “the hosts”.
“Chamath” and “Chamath Palihapitiya” are one page, across every show.
Every quote is checked verbatim against the transcript. ~96% grounded, zero fabrications.
covered: false is a real answer. No guessing, ever.
New episodes weekly. Every response states its exact knowledge cutoff.
Ask a bare model what a podcast said and you get half-remembered training data. Ask AiPod and you get the record.
Ask things like: "What does Sacks think about open-source AI?" · "How has the Nvidia discussion changed since May?" · "Which predictions were made about OpenAI?"
MCP for agent runtimes, plain JSON for everything else. No API key, no signup, no SDK to install.
# one command (Claude Code): claude mcp add --transport http aipod https://aipod-two.vercel.app/mcp # or in any MCP client config: { "mcpServers": { "aipod": { "type": "http", "url": "https://aipod-two.vercel.app/mcp" } } }
5 tools, context-first: aipod_context · aipod_search · aipod_entity · aipod_episodes · aipod_episode
# ask anything curl 'https://aipod-two.vercel.app/api/context?q=What+does+Chamath+think+about+AI+capex' # search every show curl 'https://aipod-two.vercel.app/api/search?q=Nvidia+selloff' # everything said about one entity curl 'https://aipod-two.vercel.app/api/entity/OpenAI'
Agent-readable service doc at /llms.txt — point any agent at it and it self-serves.
All GET, all cited, all under 100ms. Full spec at /openapi.json.
| endpoint | what it answers |
|---|---|
| /api/context?q= | Start here. Ask a question, get cited context — claims, quotes, predictions. `covered: false` means not covered. |
| /api/search?q= | Keyword search over every claim, fact, and topic. |
| /api/entity/:name | Everything said about a person, company, or idea — in order. |
| /api/episodes · /api/episode/:id | The episode list, and the full structured record for one. |
| /api/status | Corpus stats and per-show knowledge cutoffs. |
All-In fully indexed. Recent episodes across six more shows. Every answer states its exact coverage window.
fully indexedChamath, Sacks, Friedberg, Calacanis + guests — markets, AI, politics. Browse the public episode index →
Dwarkesh Patel’s long-form interviews on AI progress, compute economics, and the shape of the coming decade. Browse the episode index →
recent episodesLenny Rachitsky’s tactical interviews with product, growth, design, and go-to-market leaders. Browse the episode index →
recent episodesSteven Bartlett’s long-form conversations with founders, scientists, and outliers. Browse the episode index →
recent episodesTucker Carlson’s monologues and long-form interviews on foreign policy, energy, and markets. Browse the episode index →
recent episodesDr. Andrew Huberman on science-based tools for health, performance, and sleep. Browse the episode index →
recent episodesPatrick O’Shaughnessy with the world’s best investors and operators — frameworks, businesses, markets. Browse the episode index →
recent episodesEzra Klein’s long-form New York Times interviews on politics, policy, and the ideas behind them. Browse the episode index →
recent episodesDanish-language business & tech news with Mikkel and Thomas Harthimmer — what happened, why it matters, who wins and loses. (Experiment: non-English ingest; transcript in Danish, brain in English.) Browse the episode index →
“Recent episodes” means partial coverage — older archives are unknown, and AiPod says so rather than guessing. Weekly ingest moves every cutoff forward.
On this data, a product is one GET request and a template. Five running examples:
The ten lines worth remembering from each episode, each with the moment it comes from. Read it in two minutes, or post it.
demo 01393 calls on file, each linked to its source. Filterable.
demo 02Everything the shows said about OpenAI, on one page. One API call, rendered.
demo 03The podcast week compiled by a script. No editor.
shipping weeklyOne issue per week. Every quote verified verbatim — or the build fails.
Today AiPod grounds and your agent answers. Deep Answer flips it: one question in, one complete, cited answer out. Leave an email to hear when it opens.