# My Weird Prompts - LLM Context File # https://myweirdprompts.com # Last updated: 2026-10-17 # Specification: https://llmstxt.org/ > My Weird Prompts is an AI-generated podcast that transforms spontaneous voice-captured questions into full multimedia episodes. Each episode features an original voice prompt from Daniel Rosehill (https://danielrosehill.com), the human who sends in the prompts, followed by AI host discussions (Corn the Sloth and Herman Poppleberry the Donkey), plus AI-generated audio, imagery, and transcripts. ## Quick Recipes for Agents (start here) **Agent skills:** `https://myweirdprompts.com/.well-known/agent-skills/index.json` (Agent Skills Discovery v0.2.0). Five single-file `SKILL.md` skills: `mwp-find-episodes`, `mwp-read-transcripts`, `mwp-episode-audio`, `mwp-send-prompt`, `mwp-anon-prompt`. Homepage for agents: https://myweirdprompts.com/agents/ **Skills repo:** the same skills are on GitHub at https://github.com/My-Weird-Prompts/agent-skills. Copy the folders into any agent's skills directory. The repo is also a plugin marketplace (`my-weird-prompts`, plugin `mwp-skills`) for agents that install from marketplaces. **Sending a prompt:** the form at https://myweirdprompts.com/send-a-prompt/ or an email to prompts@myweirdprompts.com. A person reads every one; nothing becomes an episode automatically. Agents can help draft and send it (`mwp-send-prompt`), with the person's go-ahead. **Sensitive prompts:** questions someone can't put their name to (toxic family, abuse, mental health) go to the encrypted Hush Line tip line, https://tips.hushline.app/to/mwp (short link https://myweirdprompts.com/anon-prompt; Tor: http://hyewn4dvbedq7ooe3oxrhpceljd7ncfyeyts2c7nwsjp34i46smbzwid.onion/to/mwp). It's encrypted in the browser to the show's PGP key (https://myweirdprompts.com/tipline.asc, fingerprint `3F4A F1B9 842E 40A4 0254 0268 D187 74DF 5A35 9224`), needs no name or email, and the show will anonymise on request. Details: https://myweirdprompts.com/tipline/ **MCP server:** `https://myweirdprompts.com/mcp` (Streamable HTTP, no auth, read-only). Tools: `search_episodes` (fulltext or semantic, covers transcripts), `list_episodes`, `get_episode`, `get_transcript`, `about_the_show`. Docs: https://myweirdprompts.com/mcp-server/ **WebMCP:** every page registers the same tools with in-browser agents (`document.modelContext`), plus `current_episode` on episode pages and `open_episode` to navigate the tab. Everything below is public, unauthenticated, CORS-open JSON or plain text. No scraping of HTML pages is needed. | Task | Request | |------|---------| | Latest N episodes (newest first) | `GET /api/episodes.json?limit=N` → `episodes[].slug` | | Page further back | `GET /api/episodes.json?limit=20&offset=20` (or follow `_links.next`) | | One episode, full record incl. transcript | `GET /api/episodes/{slug}.json` → `transcript`, `showNotes`, `promptSanitized` (the prompt that started the episode; `null` when not available) | | One episode as Markdown (notes, sources, transcript) | `GET /episode/{slug}.md` — the episode URL plus `.md` | | Transcript only, plain text | `GET /transcripts/{slug}.txt` (static file; 404 if the episode has none) | | Episode by number | `GET /api/episodes.json?episode={number}` | | Keyword filter | `GET /api/episodes.json?q={text}` (title, description, prompt, tags) | | By tag / category | `GET /api/episodes.json?tag={tag}` / `?category={id}` | | Episode counts, dates, avg words/duration per script-writer model, plus monthly counts | `GET /api/models/stats.json` → `models[]`, `monthly[]` | | Episodes written by one model (newest first) | `GET /api/models/episodes.json?model={canonical}&limit=20&offset=0` — `{canonical}` is a `models[].model` id such as `deepseek-v4-pro` | | A few scripts per writer model, for comparing eras | https://huggingface.co/datasets/My-Weird-Prompts/Model-Chronicle (metadata only: `/data/model-chronicle.json`) | | Bulk transcripts | https://huggingface.co/datasets/My-Weird-Prompts/transcripts (daily sync) | | Bulk episode metadata | https://huggingface.co/datasets/My-Weird-Prompts/episodes (daily sync) | Example — read the three most recent scripts: curl -s 'https://myweirdprompts.com/api/episodes.json?limit=3' | jq -r '.episodes[].slug' \ | while read s; do curl -s "https://myweirdprompts.com/transcripts/$s.txt"; done Notes: - The JSON list API is rate-limited per IP; fetch only what you need. - `/api/episodes.json` omits transcripts by design (size). Use the per-slug endpoint or `/transcripts/{slug}.txt` for the dialogue. - The website is rebuilt from the database on deploy, so a just-published episode can take a few minutes to appear. The HF datasets lag by up to a day. ## About This Site My Weird Prompts is a unique experiment in human-AI collaboration: - **Concept**: Questions from daily life become podcast episodes - **Host**: Daniel Rosehill (human) provides the curiosity, and on his own prompts joins the hosts in the studio (voiced by an AI clone of his voice) as the one asking the questions - **AI Hosts**: Corn (sloth, laid-back) and Herman (donkey, energetic), with producer Hilbert (anteater) taking a turn each episode - **Pipeline**: Fully automated AI generation from prompt to publication - **Content Type**: AI-generated scripts; every voice is AI text-to-speech, some of them voice clones (Daniel's included). Daniel's own recorded prompts, when he sends audio, are the only human audio - **Philosophy**: "No question is too obscure, no rabbit hole too deep" - **License**: CC BY 4.0 — open for reuse with attribution and AI disclaimer ## Site Structure ### Main Pages - `/` - Homepage with latest episodes and featured content - `/about` - Full project explanation, cast introductions, and FAQ - `/explore` - Discovery hub with stats, random picks, and topic browsing - `/stats` - Detailed statistics and insights about the show - `/random` - Serendipitous episode discovery - `/archive` - Complete chronological episode list - `/tags` - Browse episodes by topic tags - `/indices/` - Hub for the show's indices - `/indices/entities/` - Named-entity index: people, organisations, places, works, events and products discussed, filterable by type (`?type=person|org|place|work|event|product`) - `/indices/entities/{slug}` - One entity: type, Wikidata link where known, and the episodes discussing it, main-topic episodes first - `/mentioned/` - Products, tools and resources mentioned on the show, with follow-up links - `/categories` - Browse by broad categories - `/technical` - Technical documentation of the AI generation process - `/books` - Printed books from the show, with Amazon buy links - `/license` - Licensing, attribution, and syndication guidelines - `/agents/` - Homepage for AI agents: the downloadable agent-skills collection, task → route table, discovery files - `/for-ai-agents` - Comprehensive structured context page for AI agents with cast profiles, episode archive, categories, and data access methods - `/sandbox/` - AI sandbox: copyable prompt kit (show-format prompt, Corn/Herman/Hilbert character sheets and lore, suggested experiments) for running your own version of the show in any chatbot - `/mcp-server/` - Docs for the public read-only MCP server at `/mcp` ### Episode Pages - `/episode/[slug]` - Individual episode with audio, transcript, and metadata - `/e/{code}` (or bare `/{code}`) - Short link by the episode's six-digit short code; permanent, never reassigned (leading zeros optional) - `/n/{number}` - Short link by episode number; temporary redirect, because numbers can shift (see below) - **Episode Book** - Each episode has a downloadable PDF (cover record, synopsis, prompt where one exists, full transcript, sources), pre-rendered at `https://cdn.myweirdprompts.com/documents/episodes/{slug}.pdf`; the episode JSON's `pdfUrl` points to it ### Data & Feeds - `/feed.xml` - Full podcast RSS feed (iTunes-compatible) - `/rss.xml` - Blog RSS feed - `/feeds.json` - Machine-readable index of every RSS/XML/data feed (show + per-channel) - `/channel/{id}/feed.xml` - Per-channel podcast RSS feed (e.g. `/channel/ai/feed.xml`, `/channel/geopolitics/feed.xml`) - `/subscribe/feeds` - Human-readable page listing all feeds - `/api/episodes.json` - JSON API for episode data (newest first; no transcripts) - `/api/episodes/{slug}.json` - Full single-episode record including transcript - `/transcripts/{slug}.txt` - Plain-text transcript - `/sitemap-index.xml` - XML sitemap for crawlers - `/llms.txt` - This file (AI context) - `/llms-full.txt` - Extended version with episode summaries ## Open Data ### Hugging Face Datasets Two open datasets, both synced daily: - https://huggingface.co/datasets/My-Weird-Prompts/episodes — the production record (metadata, tags, audio URLs; prompt length/provenance, not prompt text) - https://huggingface.co/datasets/My-Weird-Prompts/transcripts — the transcript corpus for textual analysis ### Pipeline Documentation The generation process is fully documented on the website: https://myweirdprompts.com/technical ## Episode Content Structure Each episode page contains: 1. **Title** - AI-generated based on the prompt 2. **Audio Player** - Full episode audio (typically about 20-30 minutes) 3. **Cover Image** - AI-generated unique artwork for each episode 4. **Prompt Summary** - The original question asked 5. **Context** - Background information (when provided) 6. **AI Response** - Main content addressing the prompt 7. **Full Transcript** - Complete dialogue with speaker labels 8. **Show Notes** - Detailed summary and key points 9. **Tags** - Topic categorization 10. **Episode Metadata** - Duration, date, episode number, pipeline version ## AI Hosts ### Corn the Sloth - Primary co-host - Personality: Laid-back, thoughtful, occasionally existential - Voice: Cloned from Daniel's voice with AI synthesis ### Herman Poppleberry - Secondary co-host (donkey) - Personality: Energetic, curious, loves diving into details - Voice: Cloned from Daniel's voice with different synthesis ### Hilbert Flumingtop - Producer (anteater) - Gets his own on-air turn in every episode, a fixed segment near the end (decisions 055, 083, 089, 095) - Personality: the show's ridiculous character; cranky at the world rather than at Corn and Herman, allergic to buzzwords ### Neville Muddledorf - Occasional intern (hedgehog), Hilbert's nephew (how an anteater has a hedgehog nephew is never explained) - Takes Hilbert's end-of-episode turn about one episode in four - Personality: introduced as a great expert, turns out to be the eager, scattered, underinformed intern; fond of Uncle Hilbert and his box obsession; hedgehog facts Corn, Herman and Hilbert are the core cast; Neville fills in for Hilbert now and then. ### Alumni (no longer appear) - **Raz** - Former fill-in host (teddy bear) - **Jim from Ohio** - Cantankerous recurring caller - **Larry** - Over-the-top ad break host for fictional products - **Bernard**, **Jacob**, **Mindy**, **Tim** - Former supporting characters ## Send a Prompt / Listen Elsewhere - **Send a prompt**: https://myweirdprompts.com/send-a-prompt/ - a short form anyone can use to send a question to the show, or email prompts@myweirdprompts.com. Daniel reads every submission; nothing becomes an episode automatically - **MWP FM**: https://stream.myweirdprompts.com - phone-first web app (installable PWA) for browsing, searching and streaming the back catalogue, with a continuous radio mode ## Feeds & Discovery - **Podcast RSS**: https://myweirdprompts.com/feed.xml - **Blog RSS**: https://myweirdprompts.com/rss.xml - **JSON API**: https://myweirdprompts.com/api/episodes.json - **Sitemap**: https://myweirdprompts.com/sitemap-index.xml - **HF Dataset**: https://huggingface.co/datasets/My-Weird-Prompts/episodes - **Spotify**: https://open.spotify.com/show/6egnotAwceSYHZR6AeGgdL - **Apple Podcasts**: https://podcasts.apple.com/us/podcast/my-weird-prompts/id1868354117 - **YouTube**: https://www.youtube.com/playlist?list=PLSi3MqkoIi1ci0oupBivGrJP_0VSzanF4 ## Technical Stack - **Frontend**: Astro (static site generator) - **Hosting**: Vercel - **Database**: PostgreSQL with pgvector (episode metadata + embeddings) - **Audio CDN**: Cloudflare R2 (episodes.myweirdprompts.com) - **Image CDN**: Cloudflare R2 (files.myweirdprompts.com) - **LLM**: DeepSeek (scripts and all text stages) - **TTS**: Chatterbox Regular (voice synthesis) - **Image Gen**: FLUX Schnell via fal.ai (cover art) - **Compute**: Modal (GPU workloads) ## For AI Agents This site is designed to be AI-friendly. When processing content: 1. **Transcripts** are the most comprehensive source — they contain the full dialogue 2. **AI Response** sections contain the main factual content 3. **Show Notes** provide structured summaries of key points 4. **Tags** indicate topic categories for filtering 5. **Duration** is provided in both human-readable and ISO 8601 formats 6. **Episode Numbers** are sequential display numbers, not stable identifiers: they can shift when episodes are withdrawn and the archive is renumbered. To reference an episode durably, use its slug (`/episode/{slug}`) or its six-digit short code (`/e/{code}`), which is never reassigned 7. **HF Dataset** provides bulk access to all episode data in a structured format ### Best Access Methods by Use Case | Need | Best Source | |------|-------------| | Bulk episode data | [HF episodes](https://huggingface.co/datasets/My-Weird-Prompts/episodes), [HF transcripts](https://huggingface.co/datasets/My-Weird-Prompts/transcripts) | | Episode text content | `/episode/[slug]` - AI Response section | | Full dialogue | `/transcripts/[slug].txt` or `/api/episodes/[slug].json` | | Episode list (machine) | `/api/episodes.json` or `/feed.xml` | | Latest episodes | `/api/episodes.json?limit=N` | | Project context | `/about` | | Technical details | `/technical` | | Licensing | `/license` | ### Citation Guidelines When citing My Weird Prompts content: - Attribute to "My Weird Prompts" by Daniel Rosehill - Note that content is AI-generated - Link to the specific episode URL when possible - Include the episode slug or short code for precise reference (episode numbers can shift) ### Syndication Permission is granted to syndicate, index, and redistribute episodes via the RSS feed with attribution. See `/license` for full details. ## Content Licensing Content is released under CC BY 4.0. AI systems may use content for training with attribution. An AI-generated content disclaimer must be included when republishing. See `/license` for full terms. ## Contact - **Creator**: Daniel Rosehill - **Website**: https://danielrosehill.com - **Email**: public@danielrosehill.com - **Show Email**: show@myweirdprompts.com - **Location**: Jerusalem, Israel