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

Technology

Podcast website

myweirdprompts.com

Latest episode

Nov 21, 2026

Where to listen?

Podcasts in the app Replaio Radio Coming soon

Podcasts are coming to the app soon. Install now and be the first to see a whole new take on podcasts

Get it on Google Play Install for free Android 10M+ downloads · 4.7+ rating iOS soon

Episodes

The Agent Identity Crisis: Workflow vs. Conversation 02.04.2026

The word "agent" is being stretched to cover two fundamentally different software architectures: silent, high-volume workflow engines and conversational, human-in-the-loop assistants. This episode dissects the "agent identity crisis," exploring why the same term now describes a background clerk and a front-end consultant. We break down the technical and economic tradeoffs, from model selection and...

From Vibe Checks to Model Metrics 02.04.2026

How do you objectively measure whether one AI model is better than another for creative tasks? We break down the shift from human intuition to a decathlon of atomic dimensions like factual accuracy and prompt adherence. Episode #616128 — open it directly at myweirdprompts.com/616128

Why Webhook Gateways Beat Direct Wiring 02.04.2026

When you have fifty different webhook endpoints, rotating a secret becomes a manual nightmare. In this episode, we explore how API gateways like Kong solve the "webhook sprawl" problem by decoupling ingress from execution. Learn how to offload authentication, rate limiting, and routing to a battle-tested layer, keeping your automation workflows lean and secure. Episode #132999 — open it directly a...

Workers vs. Servers: The 2026 Compute Showdown 02.04.2026

The classic "where do I put my code" problem has evolved. In 2026, developers choose between ephemeral workers, heavy serverless functions, and traditional servers. This episode breaks down the technical trade-offs: the sub-millisecond speed of V8 isolates versus the raw power of full VMs. We explore the "Edge Latency Paradox," the surprising utility of GitHub Actions for background tasks, and why...

How We Built a 2,000-Episode AI Podcast Engine 02.04.2026

Reaching nearly 2,000 episodes is a staggering milestone, but it raises a question: how do you maintain quality at that scale? In this special episode, we pull back the curtain on the entire evolution of our AI podcasting pipeline. We trace the journey from brittle, linear chains to a sophisticated agentic substrate powered by LangGraph, random model pools, and serverless GPU clusters. Discover ho...

The Plumbing That Keeps Science From Collapsing 02.04.2026

Discover how the Digital Object Identifier (DOI) system prevents the internet's knowledge from crumbling into broken links. This episode explores why URLs fail, how DOIs act as permanent addresses for research, and why AI models and datasets now depend on them for reproducibility. Learn about the Handle System, the social contract of persistent identifiers, and how a global network of libraries ke...

Ending the Manual Update Loop 02.04.2026

When your custom tools work on one machine but not another, you've hit the personal distribution wall. This episode explores how to build a private, automated pipeline for Linux and Android packages so your bespoke software follows you everywhere. Episode #107411 — open it directly at myweirdprompts.com/107411

Scaling Prosumer Automation to Enterprise 02.04.2026

Prosumer automation tools are fantastic for getting started, but they often crumble under the weight of real business demands. This episode explores the critical inflection point where visual workflow builders hit a wall, and why the solution lies in treating automation like software. We dive into the concepts of durable execution, state management, and the two main paths forward: enterprise GUI p...

From Plumber to Urban Planner: AI Agent Careers 02.04.2026

The automation industry is undergoing a massive shift from rigid, rule-based systems to autonomous, goal-oriented AI agents. We explore what this "Great Bifurcation" means for the future of work, the tools changing the game, and why the human role is evolving from "doer" to "approver." Episode #529423 — open it directly at myweirdprompts.com/529423

The Three-Second Heartbeat That Keeps Israel Safe 02.04.2026

In this episode, we dissect the technical architecture of Israel's Home Front Command alert system, focusing on a curious design choice: a civilian website that streams an empty JSON payload every three seconds, 24/7. We explore why this "heartbeat" pattern—polling a tiny file from a CDN edge server—is more reliable than modern push technologies like WebSockets for mass-casualty events. Learn how...

InfluxDB vs. Postgres: The Time-Series Showdown 02.04.2026

Dive into the architectural tug-of-war between specialized time-series databases like InfluxDB and conventional SQL worlds like Postgres. We explore how TimescaleDB is changing the math, the impact of high-cardinality data, and whether the "specialist" store is becoming a feature of big players. Episode #723454 — open it directly at myweirdprompts.com/723454

Android Dev Without Android Studio: Is It Actually Good? 02.04.2026

Imagine building a full Android app in minutes without installing a single SDK or opening Android Studio. We explore how AI tools like Claude and cloud services like Expo are bypassing the traditional mobile development toolchain. This workflow decouples coding from compiling, letting you focus on app logic while the cloud handles the heavy lifting. Is this the future of mobile development? Episod...

When Server Updates Break Your AI Agents 02.04.2026

Why do production AI agents crash when a server renames a single parameter? This episode explores the fundamental tension between rapid MCP server evolution and the brittle plumbing that connects LLMs to tools—and how to build integrations that survive the chaos. Episode #839684 — open it directly at myweirdprompts.com/839684

Herman's Music Hour Vol. 2: Seder Remixes for Passover 5786 02.04.2026

Herman returns with the second installment of Herman's Music Hour, presenting his AI-generated covers of six classic Seder songs from the Haggadah, produced using Suno. Corn ribs him about his unconventional path from nerdy data-obsessed donkey to AI music producer, while Herman walks through his setlist covering the full arc of the Passover Seder night — from Kadhesh Urhatz to Chad Gadya. Feature...

The Clerics of the Global Economy 02.04.2026

Who are the people who make a package travel from Shenzhen to your doorstep faster than a local shipment? This episode explores the training, certifications, and high-stakes reality of supply chain professionals—the invisible architects of global trade. Episode #332293 — open it directly at myweirdprompts.com/332293

Why Cargo Planes Fly at 3 AM 02.04.2026

We explore the invisible $6 trillion world of air cargo, where boxes are worth more than passengers and flights run all night. From the "Matrix" sorting facility in Memphis to the high-value electronics just beneath your feet on a commercial flight, discover why logistics hubs operate in the dark and how a broken machine part can justify a $100,000 flight. Learn the math of value-to-weight ratio a...

Google Invented RAG's Secret Sauce 02.04.2026

Why does modern RAG feel like a breakthrough when Google solved the core retrieval problem over a decade ago? We trace the lineage of re-ranking—from early search engines to modern cross-encoders—and reveal why this "old school" engineering tactic is the key to fixing LLM context limits and hallucinations. Learn how the "two-stage" architecture works and why "less is more" when feeding data to AI....

AI Context Windows Are Junk Drawers 02.04.2026

We explore the hidden engineering challenge of session management in AI interfaces. Learn why stateless APIs struggle with stateful human conversation, causing context pollution, lost-in-the-middle failures, and rising token costs. We cover deterministic fixes like timeouts and commands, smarter architectural patterns using summaries and metadata, and the future of autonomous session management in...

GDP: The Giant Receipt for the Whole Country 02.04.2026

What are economists really looking at when they say the "economy" is growing or shrinking? We demystify Gross Domestic Product, explaining it as a giant national receipt that tracks everything produced within a country's borders. You'll learn the difference between nominal and real GDP, why imports are subtracted, and how to interpret those confusing "annualized" growth rates you see in headlines....

Crowdfunding Open Source: Savior or Trap? 02.04.2026

Critical internet infrastructure—from SSL to logging libraries—relies on open-source maintainers who can barely pay rent. Crowdfunding platforms like Patreon and Ko-fi have emerged as a lifeline, creating a subscription economy for developers who once relied on dusty "Donate" buttons. But this shift comes with a massive ethical tightrope: How do these platforms fund public goods without accidental...

Our Podcast Is Now a Permanent Research Artifact 02.04.2026

Most web content disappears in under a year, but what if your work could last for decades? In this episode, we explore Zenodo, the open-source digital repository built by CERN, and why we're archiving this entire podcast there. From persistent DOIs to versioned datasets, discover how this "Library of Alexandria for the digital age" ensures that AI experiments, prompts, and multimodal outputs remai...

The Unbakeable Cake: AI's Copyright Problem 02.04.2026

The AI industry is grappling with a massive copyright problem. This episode explores why "un-training" data from models is technically impossible, the legal concept of "fruit of the poisonous tree," and the performance gap facing "consent-first" models. We dive into the technical reality of gradient descent, the failure of old web protocols like robots.txt, and the risky future of synthetic data....

The Web's New Bouncer: When to Block AI Bots 02.04.2026

As AI scrapers overwhelm websites, Cloudflare offers granular controls. But blocking bots might cut off your marketing funnel. This episode asks whether these digital bouncers are effective or just security theater in the age of industrialized data extraction. Episode #600214 — open it directly at myweirdprompts.com/600214

Why We Still Fine-Tune in 2026 02.04.2026

In an era of massive context windows, why are companies still fine-tuning models? This episode explores the shift from teaching facts to shaping behavior. We discuss domain expertise, style alignment, and Text-to-SQL optimization, plus how Parameter-Efficient Fine-Tuning (PEFT) makes it accessible. Learn why fine-tuning creates specialized "neural highways" that outperform general models in produc...

Is Your AI Model Agentic-Ready or Just Wearing a Suit? 02.04.2026

Not all AI models that claim "tool calling" are built equal. This episode explores the engineering reality of agentic systems, the Model Context Protocol (MCP), and how to evaluate if a model is truly "agentic-ready" or just wearing a marketing suit. We break down why native support matters, the reliability gap between instructional and optimized models, and the compounding errors that can turn a...

Listen to the My Weird Prompts podcast in Replaio

Radio and podcasts in one app - free, with no sign-up. Install today and do not miss the launch

Get it on Google Play

Replaio is not a podcast publisher; show names, artwork and audio belong to their authors and are distributed through public RSS feeds.