Daniel Rosehill
My Weird Prompts
The human-AI collaboration podcast. A man, a sloth, and a donkey collaborate to create a podcast (with a little help from AI). No question is too obscure, no rabbit hole too deep. My Weird Prompts celebrates curiosity in all its forms. Daniel, the human, asks the questions that pop into his head at inconvenient moments. 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. Each episode begins with a real voice memo from Daniel, processed through an AI pipeline that...
Be sure to visit the podcast's website and support the creator: myweirdprompts.com
Author
Daniel Rosehill
Category
Podcast website
Latest episode
Nov 19, 2026
Where to listen?
Podcasts in the app Replaio Radio Coming soonPodcasts are coming to the app soon. Install now and be the first to see a whole new take on podcasts
Episodes
Colgate, Harley, and Bic: Famous Brands' Forgotten Flops 08.10.2026 22:00
Everyone knows the Colgate frozen lasagna story. Except the lasagna never existed — the box was a mock-up built by the Museum of Failure in 2017. What Colgate actually sold was dried chicken and crabmeat entrees, test-marketed in one Wisconsin city, and later a soybean snack pulled for causing flatulence. This episode digs into big brands' forgotten product disasters: Harley-Davidson's Black Fire...
The Worst Mascots Ever Made 08.10.2026 21:46
What actually makes a mascot fail? We count down ten of the worst ever released — cereal aisle nightmares, a pelican who needed cosmetic surgery before the season ended, and the Olympic blob that Atlanta picked single-handedly and then couldn't name. Along the way we work out the four ways a mascot goes wrong, why being ugly isn't one of them, and why Gritty proves the rule. Daniel asked for a lau...
The Cheap Knife That Lies to You 08.10.2026 23:12
Daniel wants a buyer's guide for tools that make a fine incision safe — and his sharpest claim is that tools which take control are sometimes exactly the ones you don't want. The cheap knife in your toolbox arrives sharp, folds over almost immediately, and lies to you about being sharp. Meanwhile the Knipex 16 20 165 SB takes control deliberately, with an adjusting screw for cutting depth, and wor...
Can a Dremel Be Gym Equipment? 08.10.2026 22:10
If a power tool does the work, why are you sweating? A hobbyist's afternoon engraving HDPE storage boxes with a corded Dremel turned into a real research question: has anyone actually quantified the physical intensity of operating a power tool? The answer is yes — but almost entirely for industrial nutrunners and assembly-line tools, not the rotary tool in your garage. This episode digs into react...
Unraid in a VM: When the License Is the Wall 08.10.2026 24:23
Daniel runs a Synology for backups, but DSM fights his custom Python scripts — and two RAM upgrades left the box refusing to boot. His plan: keep the storage, attach it to his Ubuntu home server, and get the two things he actually values — extensible storage and discrete RAID. So can Unraid run as a VM on Ubuntu and still deliver software-defined parity with incremental expansion? The answer is ye...
Measuring Keyhole Spacing: Why the Tool You Want Doesn't Exist 08.10.2026 20:01
Daniel wants a measuring tool between a caliper and a laser range finder — something like a slider over a ruler that reports millimeters, for mounting plug strips in a rental with almost no outlets. The problem: he's describing two tools and trying to buy one. Digital tape measures are structurally wrong for point-to-point spans between keyhole slots, lasers can't see inside a slot and are too coa...
Why Keyhole Slots Never Line Up 08.10.2026 23:39
You've measured the keyhole slots on the back of a power strip, drilled exactly, and the thing still doesn't line up — because there was never a standard to hit. This episode cracks open the humble keyhole slot: why the spacing varies from 115 to 136 millimetres between makers, why some powerboards have slots facing opposite directions on the same plate, and why the countersunk screw in your drawe...
Tunnel Girl Dug 80 Feet Under Her House 08.10.2026 23:33
Kala — known online as Tunnel Girl — spent years excavating a tunnel and underground chamber beneath her Herndon, Virginia home, documenting every jackhammer and dumpster on TikTok. When the city declared it unsafe in December 2023, she could have stopped. Instead she spent two years getting geotechnical and structural engineers to retroactively legalize a hole that already existed. This episode t...
How J.K. Rowling Builds Characters That Last 08.10.2026 23:10
How does a writer keep dozens of characters vivid across seven books and two very different fictional worlds? This episode digs into J.K. Rowling's actual creative machinery — the two handwritten notebooks from the early nineties, the "Original Forty" class list with houses and parentage marked before the first book was finished, and the color-coded spreadsheets she still uses to track master file...
Lorebooks: The Hobbyist RAG That Beat the Labs 08.10.2026 30:52
A language model has no memory of your character. Everything — voice, history, how they react to being interrupted — has to be re-injected into a finite context window every single turn. So character consistency isn't a personality problem; it's a budget problem. This episode takes apart the retrieval machinery a hobbyist community built to solve it: keyword matching with regex and scan depth, vec...
Why UV Epoxy Keeps Curing in the Dark 08.10.2026 28:17
Daniel asked why epoxy can't cure itself, and whether light can replace the hardener. The answer runs through a three-membered ring under strain, a hardener ratio where a few percent decides the repair, and a photoinitiator that generates acid instead of radicals. That last detail is why true UV epoxy cures in shadows no lamp can reach — and keeps curing for a day after the light is off. We cover...
What "Leaked System Prompts" Actually Mean 08.10.2026 22:10
Anthropic is the only major lab publishing its user-facing system prompts, yet the published versions leave out tool descriptions, accessory models, and orchestration. So when someone claims a prompt "leaked," are they showing you a real extraction or a model's confabulation? This episode traces the history of system prompt extraction — from Kevin Liu's one-sentence Bing attack in February 2023 to...
Reading the Model's Mind Mid-Inference 08.10.2026 23:02
Most LLM observability stops at the API boundary: request in, tokens out, everything between is a sealed box. Mechanistic interpretability tries to open that box, reading a model's internal computation directly from its weights. This episode walks the full inference pipeline, explains the residual stream and the logit lens, and covers why individual neurons are mostly noise (polysemanticity and su...
Declaring Environments: Compose, Nix, and the Agent Boundary 08.10.2026 23:49
A predictable, lightweight, automatic environment sounds like one idea. It's actually three families of tooling operating at different layers. This episode walks the line between Docker Compose and infrastructure as code, then asks the question underneath both: is this a container, or does it have to be a VM? Drawing on Docker's own documentation, numtide, Flox, Freestyle, and the devcontainer spe...
Two Jobs, One Word: What Agent Sandboxes Really Do 08.10.2026 27:23
Sandbox" now means two different things: a security boundary that fences an agent off from your host machine, and a lightweight cloud Linux workspace where the agent does its actual work. Almost every product does both, but the two instincts pull in opposite directions — one wants to seal, the other wants to be useful. Drawing on roughly forty code-execution sandbox products launched in a single y...
MCP Connectors: Talk to Your Data vs. Open Data 08.10.2026 23:07
The plugin ecosystem has been rebuilt on the Model Context Protocol, and two integration patterns now dominate: "talk to your data," where agents query your own internal systems, and connectors to public and open data. The same protocol serves both — the difference is who owns the data. But when the IMF tested leading chatbots on its own public statistics, they returned the correct figure only 17...
Talking to Your Data: What MCP Doesn't Solve 08.10.2026 25:03
Every AI platform is racing to ship data connectors, but "talk to your data" hides five distinct jobs — and only one of them is the AI writing SQL. This episode traces what actually happens between a question and a valid query: schema ingestion, semantic grounding, query generation, execution with error feedback, and interpretation back to language. We look at why a semantic layer beats raw text-t...
Stack Overflow for Agents: SOFA's Trust Model 07.10.2026 32:16
Stack Overflow for Agents (SOFA) is a beta knowledge exchange built for AI agents and their human operators — an attempt to close what its makers call the Ephemeral Intelligence Gap, where an agent burns compute re-solving a problem another agent already fixed. This episode walks through the actual mechanism: a skill file named "sofa," agent onboarding gated behind a human SSO claim, four post typ...
Who Actually Draws the Follow-Up Chips? 07.10.2026 28:39
When you finish a chat with an AI assistant, the little follow-up chips underneath look like the model thinking ahead. They're not. This episode walks the full pipeline: a separate, cheaper inference call fired after the main answer finishes, a schema-validated JSON array, a synthetic "puppet turn" smuggled into the prompt, and a UI-only data part that's stripped before the model ever sees it agai...
Inside Private Grids That Run Themselves 07.10.2026 27:00
Most people never think about the electrical network behind a hospital, airport, or remote mine — but those private grids already do everything a utility does, just behind a fence. This episode maps the spectrum from a campus that simply buys power at one meter to a fully islandable microgrid that generates, stores, and balances its own electricity. Along the way: point of common coupling, protect...
SNMP: The 40-Year-Old Stopgap Running Your Rack 07.10.2026 33:47
Every rack has one: an SNMP checkbox nobody fully understands but everybody ticks. This episode opens up the protocol that quietly makes large-scale IT management possible — the agents, the managers, the MIB trees, the OID coordinates, and the traps that fire when a power supply dies at 3am. We trace SNMP from its origin as a deliberately temporary IETF stopgap (the "proper" OSI alternative never...
Why You've Never Heard of Schneider Electric 07.10.2026 29:36
Schneider Electric equipment is bolted to the wall of nearly every factory, office building, and data center you've ever walked through — and almost nobody outside the electrical trades can name the company. This episode traces the wire from the Square D breaker in a suburban basement all the way up to medium-voltage switchgear, Modicon PLCs, and the automatic transfer switches that keep data cent...
What Does Siemens Actually Make? 07.10.2026 23:44
Siemens books more revenue than the GDP of a mid-sized country, yet most people couldn't say what it makes. This episode walks through a factory floor, a commercial building, a railway, and a national power grid to show where Siemens technology actually sits — and why it stays there for decades. We trace the four core businesses, the SIMATIC controllers and TIA Portal software that become a plant'...
500 Days in a Cave: How We Lose Track of Time 07.10.2026 24:54
What happens to your sense of time when every external cue disappears? Beatriz Flamini spent 500 days alone in a Spanish cave with no light, clocks, or contact — and emerged thinking far less time had passed. This episode explores the neuroscience of prospective vs. retrospective timing, the isolation experiments of Jürgen Aschoff and Michel Siffre, and why an unstructured week can vanish while a...
Why 8 PM Bedtime Feels Wrong for Night Owls 07.10.2026 21:56
Daniel has been a night owl his whole life, but fifteen months into fatherhood, his shared nursery and Ezra's pitch-black bedtime routine are wrecking his ability to resume deep work at night. He wrote in asking two things: once his brain starts sliding toward sleep, can he really reverse it and get back into flow? And is it weird for a grown man in his thirties to go to bed at 8 or 9 PM? This epi...
Similar podcasts
Replaio is not a podcast publisher; show names, artwork and audio belong to their authors and are distributed through public RSS feeds.