Skip to content
Artwork for My Weird Prompts
TechnologyScience

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.

Play
  • 1,808 episodes
  • Avg 25 min
  • English

Support the show

Goes straight to the publisher. podnod takes nothing.

Episodes 1,808

Counted on this page — what you have heard stays on this device, so it is not something the list can be paged by.
  • #6523
    Yesterday · 25 min

    Unpacking Israel's National AI Strategy

    Israel set a national target of 100,000 AI accelerators, but primary documents commit no state budget to building them.

    • Transcript
    • Chapters
  • #6522
    Yesterday · 28 min

    Nebius and the GPU Cloud Pivot

    Born from Yandex's post-invasion breakup, Nebius transformed leftover assets into a $65 billion GPU cloud backed by Nvidia.

    • Transcript
    • Chapters
  • #6521
    Yesterday · 28 min

    Multitenancy in Qdrant Vector Databases

    Migrating a hundred small tenants into a pooled Qdrant collection reduced resident memory by roughly eighty percent.

    • Transcript
    • Chapters
  • #6520
    Yesterday · 27 min

    Mapping Multi-Tenancy Across Vector Databases

    Isolating tenants physically into namespaces can cut vector query costs a hundredfold compared to using metadata filters on shared indexes.

    • Transcript
    • Chapters
  • #6518
    Monday · 22 min

    Why Export Bans Cannot Stop Remote Compute

    Data centre colocation vacancy in Johor, Malaysia, sits at just 0.7 percent as foreign AI developers rent access to controlled Nvidia hardware.

    • Transcript
    • Chapters
  • #6515
    Monday · 19 min

    Creative Schemes Behind the Global GPU Hunt

    Engineers carried eighty terabytes of training data across international borders in four checked suitcases to reach rented servers in Malaysia.

    • Transcript
    • Chapters
  • #6514
    Monday · 22 min

    Why Even Solo Developers Face a GPU Shortage

    H100 rental contracts jumped nearly 40 percent in five months as on-demand GPU capacity sold out across legacy and cutting-edge hardware.

    • Transcript
    • Chapters
  • #6512
    Monday · 22 min

    Build Servers Never Actually Build Anything

    A Woodpecker CI server runs entirely as an idle control plane using roughly 100 MB of RAM while external agents handle compilation.

    • Transcript
    • Chapters
  • #6511
    Monday · 22 min

    Why Immigrants Speak Hebrish

    A fine-tuned speech model proved nearly four times better at recognizing untranslated Hebrew loanwords in English-language speech.

    • Transcript
    • Chapters
  • #6510
    Monday · 20 min

    Voice Isolation Before Speech Recognition

    Running voice isolation on clean speech without a competing talker increases transcription word error rates from 0.04 percent to nearly three percent.

    • Transcript
    • Chapters
  • #6508
    Monday · 25 min

    Why Speech Transducers Omit Common Symbols

    A transducer-style speech model cannot learn to emit an at-sign through fine-tuning if the character is missing from its fixed output vocabulary.

    • Transcript
    • Chapters
Showing 1–20 of 1,808 episodes