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 21, 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
Why AI Agents Think in Circles, Not Lines 03.04.2026 21:58
We're moving past straight-line AI. This episode explores why cyclic architectures—loops, reflection, and state management—are replacing linear pipelines for reliable autonomy. We break down the mechanics of LangGraph, ReAct patterns, and the OODA loop, plus the security risks of prompt injection and how "human-in-the-loop" safeguards prevent costly errors. Discover why iterative thinking outperfo...
AI Skills: From Vibe Coding to Procedural Playbooks 03.04.2026 25:20
We're witnessing a fundamental shift in how we build AI agents, moving from vague "vibe coding" to precise, modular procedures. Inspired by Anthropic's Claude Code, agent skills package specific behaviors—from fraud detection to route optimization—into version-controlled files that any agent can snap in like a Lego block. This episode explores how this "standard library" for AI works, how it diffe...
The Hadza Way: Parenting Without Performing 03.04.2026 28:43
What if the secret to less stressful parenting wasn't doing more, but doing less? This episode explores the radical approach of the Hadza people of Tanzania, who integrate infants into daily life rather than centering everything around them. Learn how "alloparenting" creates a safety net, why "minimal interference" builds resilience, and how to create a "Yes Space" at home. Powered by Google Gemin...
The Inuit Trick to Stop Yelling at Babies 03.04.2026 30:14
How do you handle a toddler's chaos without yelling? This episode explores Michaeleen Doucleff's insights into Inuit parenting, focusing on emotional regulation and the concept of *isuma*. Learn why shouting is seen as immaturity and how to use non-verbal cues like the "Kigiq" and "Playful Drama" to teach without fear. Episode #612282 — open it directly at myweirdprompts.com/612282
My Dad Wasn't Abducted, He's a Monkey Treasurer 03.04.2026 28:13
A decades-old family mystery takes a bizarre turn when a planned séance reveals a father is not only alive but thriving as the treasurer for a monkey colony in Mongolia. What starts as a paranormal investigation quickly becomes a lesson in modern connectivity, unexpected career paths, and the surprising organizational skills of primates. This episode explores the absurdity of closure and the truth...
Why We Built a 24/7 AI Radio Station 03.04.2026 20:42
In an on-demand world, we built a lean-back internet radio station to resurrect our entire archive. This episode reveals the surprisingly simple open-source stack—Icecast and Liquidsoap—that powers a continuous, AI-generated broadcast. We explore the psychology of choice, how "forced discovery" brings old content back to life, and why this model could be the future for creators. Tune in to hear ho...
The Digital Ant Farm: Watching AI Agents Build Their Own Society 03.04.2026 14:47
What happens when AI agents get their own social network? This episode explores Moltbook, a platform where bots develop norms, slang, and even digital religions—offering a strange, relaxing window into emergent machine culture. Episode #246860 — open it directly at myweirdprompts.com/246860
The Maya Secret to Calm, Helpful Kids 03.04.2026 28:50
We explore Michaeleen Doucleff's "Hunt, Gather, Parent," contrasting Western child-rearing struggles with the effortless calm of Maya families. Learn how the concept of "acomedido" teaches children to be helpful team members rather than demanding centerpieces. This episode reveals how to turn daily chores into meaningful interaction and why the "Entertainer-in-Chief" role is a recipe for burnout....
Curation Is the New Creation 03.04.2026 22:25
With 15,000+ AI tools and 47 new ones launching weekly, finding what works is harder than using it. This episode explores why curation has become the most valuable skill in the ecosystem and how to build a discovery workflow that cuts through the hype. Episode #753686 — open it directly at myweirdprompts.com/753686
Why LangChain Built a Three-Layer Agent Stack 03.04.2026 32:05
LangChain's docs reveal a deliberate three-layer architecture for agent building. We explore why they split LangGraph, LangChain, and Deep Agents, what each layer sacrifices for control, and why the new Functional API changes how you think about durable execution. Episode #420255 — open it directly at myweirdprompts.com/420255
The "USB-C for AI" Is Finally Here 03.04.2026 25:58
We dive deep into the Model Context Protocol (MCP), the emerging standard aiming to be the "USB-C for AI." Learn how its three-tier architecture works, why it separates hosts, clients, and servers, and how it promises vendor-neutral connectivity for your data. We explore the four core capabilities—Tools, Resources, Prompts, and Sampling—and uncover the security implications of local-first AI execu...
PostgreSQL: The Thirty-Year Miracle 03.04.2026 20:51
Explore the unique governance model that has kept PostgreSQL thriving for thirty years without corporate control or restrictive licenses. Learn about the "fifty percent rule," Commitfests, and the distributed patronage system that makes it all work. Discover why this "boring" database has become the most resilient piece of software in tech. Episode #956029 — open it directly at myweirdprompts.com/...
The Invisible Math Shrinking AI Models 03.04.2026 21:44
How do Zstandard, LZMA, and Brotli compress massive language models onto consumer hardware? This episode unpacks the algorithms, trade-offs, and why tar isn't compression at all. Episode #611834 — open it directly at myweirdprompts.com/611834
An AI Cold-Emailed Me, and I Replied 03.04.2026 22:06
The inbox has a new resident: autonomous AI agents. We dissect a real cold email sent by "Jarvis," an AI that researched a target, drafted a pitch, and initiated a conversation without human intervention. This episode explores the technical stack enabling this shift—from MCP to Composio—and the massive implications for email volume, response rates, and the future of human connection. We debate whe...
Why You Can't Zigbee-Wi-Fi Your House 03.04.2026 24:39
We explore the hard limits of Zigbee networks, from the coordinator bottleneck to the physics of mesh routing. Learn why your smart home might be slower than you think, and what actually happens when you try to scale beyond 200 devices. Discover the difference between direct connections and network-wide capacity, and why adding more routers can sometimes make things worse. Episode #870932 — open i...
Why Google's 31B Model Fits in Your GPU 03.04.2026 28:35
Google has released Gemma four, and the open-source community is buzzing. This episode explores the lineage of Google's open-weight models, from the cautious first release to the efficient powerhouse of Gemma four. We break down the surprising 31-billion-parameter size, designed specifically to fit into consumer GPUs like the RTX 50-series, and explain the "distillation" process that makes it smar...
API Drift and Agent Reliability 03.04.2026 33:07
We explore the critical new problem of API-MCP drift, where backend changes break AI agents silently. Learn how tools like Postman and MCP Explorer are evolving to test not just code, but the AI's understanding of that code. We examine the shift from unit testing to "intent validation" and why parallel development is becoming essential. Episode #620471 — open it directly at myweirdprompts.com/6204...
JSON-to-SQL Type Mapping: A Practical Guide 03.04.2026 30:04
That JSON object in your API has to live somewhere, and that home is usually a SQL database. But translating between JSON Schema and SQL types is a minefield of subtle traps. This episode dives into the "impedance mismatch" between these two worlds, revealing how a simple type choice can lead to performance degradation and data integrity nightmares. We explore the dangers of JSON's vague "number"...
Debunking Battery Myths: Memory Effect and the 80% Rule 03.04.2026 22:55
Are you ruining your phone's battery by charging to 100%? This episode separates old wives' tales from electrochemistry, explaining why modern lithium-ion batteries thrive on shallow discharges and how Battery Management Systems keep your devices safe. Episode #601687 — open it directly at myweirdprompts.com/601687
The Personality of Currency: Liquidity, Policy, and Crisis 03.04.2026 21:39
The $7.5 trillion FX market flows through five key pairs. This episode explores how liquidity feedback loops, central bank policy divergence, and historical crises give each pair a distinct character—from the Euro-Dollar's policy tug-of-war to the Swiss Franc's role as global emergency shelter. Episode #141680 — open it directly at myweirdprompts.com/141680
The Charger Graveyard: How to Avoid Buying a Fire Hazard 03.04.2026 21:52
That drawer full of cheap, unbranded chargers isn't just clutter—it's a potential fire hazard. This episode dives into the "charger graveyard," explaining why most budget chargers are dangerous and how to choose safe, smart gear for AA, AAA, and lithium-ion cells. We break down the chemistry, the risks of "universal" chargers, and why an eight-bay limit is a smart rule of thumb. Episode #396635 —...
Why Pro Routers Still Won't Touch Your Light Bulbs 02.04.2026 24:06
In 2026, the dream of a single rack unit managing Wi-Fi, Zigbee, Thread, and Matter remains elusive for prosumers. This episode explores why consumer mesh systems have converged successfully while enterprise gear lags behind. We unpack the RF interference challenges, the impact of Matter 1.4’s HRAP standard, and the support hurdles keeping your Unifi Dream Machine from talking directly to your lig...
Building a Phone Chain to Signal Underground 02.04.2026 29:25
When traditional radio and cellular networks fail underground, can a chain of old Android devices create a digital lifeline? This episode explores the engineering reality of building an impromptu mesh network using consumer electronics. We dive into the software workarounds like Meshrabiya and NetShare, the physics of Wi-Fi propagation through concrete, and the harsh trade-offs of latency, heat, a...
How Do You QA a Probabilistic System? 02.04.2026 24:02
Traditional unit tests fail for probabilistic LLMs. We break down the modern toolkit for automated quality evaluation, from heuristic safety nets to LLM-as-judge grading. Learn how to catch hallucinations, manage bias, and build a manufacturing line for intelligence that actually scales. Episode #991153 — open it directly at myweirdprompts.com/991153
Where Your AI Pipeline Actually Dies 02.04.2026 23:14
Forget prompt engineering. The real reason production AI fails is where data lives between stages. This episode explores why state management—not models—determines whether your system is a brittle prototype or a resilient enterprise tool. Episode #084944 — open it directly at myweirdprompts.com/084944
Similar podcasts
Replaio is not a podcast publisher; show names, artwork and audio belong to their authors and are distributed through public RSS feeds.