Castuner

My Weird Prompts

Daniel Rosehill

No question too obscure, no rabbit hole too deep. A man, a sloth, and a donkey talk through one question at a time (with a lot of help from AI). My Weird Prompts celebrates curiosity in all its forms. Daniel, the human, asks the questions that pop into his head at inconvenient moments, and now joins the hosts in the studio to ask them himself. Corn the Sloth offers laid-back, thoughtful takes. Herman the Donkey brings boundless enthusiasm and energy. Together, they explore topics ranging from the mundane to the mind-bending. Every episode is made by an AI pipeline that researches the question, writes the script, voices it with AI text-to-speech and assembles the final podcast. Every voice you hear is synthetic, and some are voice clones, Daniel's included. Listeners can send prompts too. Stay curious.

Podcast website

Listen in your app

Episodes

6267 episodes

  1. Smart Home Emergency Buttons and Mental Labor 17:53
  2. Why Named Entity Linking Fails After the Span Is Found 26:25
  3. Why Model Benchmarks Fail at Wit 23:06
  4. How Political Scandals Exposed Boring Email Bugs 31:21
  5. Why 3M Primer 94 Sticks to Everything 18:16
  6. Why API Email Still Relies on SMTP 20:36
  7. Why Model Verbal Tics End Up Driving Script Benchmarks 19:16
  8. Verifying Generated Arabic Text as a Non-Reader 18:53
  9. How Forensic Tools Search Android for Spyware 25:07
  10. Why Editing Vector Databases Silently Re-Embeds Text 18:33
  11. Decaying Old Memories in Vector Search 25:13
  12. Why Swipe Decks Make Better Training Data 18:40
  13. What Really Runs When You Compact an Agent's Context 21:30
  14. Where Harmonizable Processes Actually Work 19:33
  15. Why Weighting Software Scores Fails to Pick Winners 21:00
  16. Why LLM Output Tokens Cost More Than Input 23:25
  17. Choosing Class Breaks for Skewed Price Data 18:39
  18. Choosing AI Models on Cost and Performance 18:54
  19. Turning Anonymous Diarization Clusters into Names 23:39
  20. Choosing the Right Voice Activity Detector 32:08
  21. Porting Voice Models to Core ML and MLX 30:46
  22. Running Speaker Identification Locally on Phones 22:34
  23. How Audio Models Tell Voices Apart 28:23
  24. Why Financial Centres Outlive the Forces That Built Them 25:12
  25. Why Modern AI Rests on Statistics 23:16
  26. What Theoretical Physicists Actually Do All Day 29:42
  27. MWP 4 Kids: Why the Moon Changes Shape 9:43
  28. Where Email Attachments Actually Live and Go 23:12
  29. Why Private IP Addresses Look the Way They Do 22:45
  30. Why Modern Audio Datasets Ship as Parquet Files 31:00
  31. The Most Popular GPUs - Today's Cloud Demand 23:51
  32. What Does It Take To Train An AI Model From Scratch? 28:18
  33. Why Windows Struggle with Bass 21:24
  34. Why Training Recipes Matter More Than Model Size 28:07
  35. Teaching Speech Models to Transcribe Bilingual Slang 22:56
  36. Collecting Speech Data on Outdoor Walks 22:13
  37. When Automated Proofs Leave Mathematicians Grieving 28:47
  38. Where Mathematicians Actually Work and How AI Fits In 24:10
  39. Why the Kibbutz Had to Unbundle Security 29:50
  40. What Counts as a Mathematical Proof 29:01
  41. When Eating Triggers Breathlessness Without Reflux 29:13
  42. Why Cross-Age Friendships Depend on Shared Places 27:52
  43. Why Viral Campaign Videos Fail to Build Coalitions 27:43
  44. Mapping Hasidic Dynasties and Their Modern Centers 28:36
  45. Why Model Downgrades Are Rarely About Quantization 23:43
  46. What Remote Teams Can Learn from the Linux Kernel 28:29
  47. How Ubuntu Actually Updates Your Firmware 30:13
  48. Hilbert Flumingtop on Sixty-Three Boxes 36:07
  49. Why GPG Builds an Entire Department Instead of a Key 28:48
  50. Why Modern SSH Keys Cannot Fix Harvested Traffic 23:44