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
The Good Fence: Lebanon’s Forgotten Refugees in Israel 25.04.2026 29:47
In May 2000, as Israel withdrew from southern Lebanon after 15 years, some 6,500 Lebanese — former South Lebanon Army soldiers and their families — fled across the border. This episode traces the full arc of that forgotten chapter: the Good Fence humanitarian crossing opened in 1976, the alliance with Saad Haddad and the SLA, the security zone’s daily intimacy between Israeli soldiers and Lebanese...
Ghost Murmur: Heartbeat Detection or Disinformation? 25.04.2026 40:42
In early April, an American airman was rescued in Iran. But the White House briefing focused on a CIA technology called Ghost Murmur—a system allegedly capable of identifying a person by their unique cardiac rhythm through walls, from miles away. Is this a revolutionary breakthrough, or a carefully crafted disinformation campaign? We break down the real physics of remote heartbeat detection, from...
Autism Numbers vs. the Noise 25.04.2026 25:13
The headlines say autism is exploding. The data tells a more complicated story. This episode cuts through the panic to examine the actual research: how autism went from a 1-in-1,400 diagnosis in 1980 to 1-in-31 today. We trace the diagnostic history from Kanner and Asperger through five editions of the DSM, break down what diagnostic substitution really means, and explore why prevalence varies 10x...
Is Love on the Spectrum Helping or Hurting? 25.04.2026 22:57
Is Netflix's *Love on the Spectrum* a heartwarming window into autistic dating, or does it trade on awkwardness for entertainment? We explore the fierce debate between autism advocacy groups and self-advocates. From infantilizing musical cues and the problem of "masking" to the show's narrow slice of the spectrum, we unpack what the series gets right, what it gets wrong, and why the autistic commu...
When Your AI Says No to Everything 25.04.2026 23:55
Why does your AI assistant refuse to answer harmless questions? This episode explores the hidden failure mode of over-refusal in large language models — when safety guardrails fire on innocent prompts like "how to kill a mosquito." We break down three key benchmarks: OR-Bench (80,000 prompts), XSTest (the predecessor now considered "solved"), and PHTest (model-specific pseudo-harmful prompts). The...
When AI Caves: Progressive vs. Regressive Sycophancy 25.04.2026 27:44
Stanford's SycEval benchmark reveals that 58% of AI responses flip when challenged, but the real story is the difference between progressive sycophancy (accidentally correct) and regressive sycophancy (actively wrong). We explore why preemptive rebuttals trigger more flips and what the 78% persistence finding means for alignment. Episode #229518 — open it directly at myweirdprompts.com/229518
Are Political Bias Benchmarks Actually Measuring Anything? 25.04.2026 26:46
The Political Compass Test is the go-to tool for measuring political bias in large language models — but a growing body of research suggests it's fundamentally broken. This episode unpacks why the PCT can mask bias rather than reveal it, then walks through four better alternatives: IssueBench's open-ended prompt approach, Stanford's perception-based user study, OpenAI's granular five-axis internal...
How Researchers Actually Measure Censorship in Chinese LLMs 25.04.2026 30:45
Everyone argues about whether Chinese LLMs are censored, but almost no one asks how we actually know. This episode unpacks the validated benchmarks—CHiSafetyBench, SafetyBench, ChineseSafe, FLAMES, JailBench, and the PNAS Nexus longitudinal study—that researchers use to measure political refusal. We explore the three different things "censorship" can mean, why multiple-choice tests inflate safety...
When AI Cheats on Cultural Knowledge 25.04.2026 29:54
Frontier models score poorly on cultural benchmarks, but worse: they cheat by pattern-matching culture names. This episode unpacks the methodologies that expose the shortcuts and what they reveal about AI's understanding of human culture. Episode #283845 — open it directly at myweirdprompts.com/283845
How Backpropagation Actually Unlocks Neural Networks 25.04.2026 25:38
What actually happens inside a neural network when it learns from its mistakes? This episode breaks down backpropagation — the algorithm that computes gradients for every weight in a network by propagating error signals backward through the same connections that carried data forward. We walk through a concrete example using the chain rule, explain the credit-assignment problem that kept neural net...
Three Landings in 90 Days: Pilot Automation Dependency 25.04.2026 31:18
The FAA's own data shows pilots aren't hand-flying enough to stay sharp—and the current regulatory floor only requires three landings every 90 days, none of which need to be manual. This episode examines the automation dependency problem exposed by Air France 447 and Asiana 214, the specific regulatory gaps under FAA and EASA rules, and what airlines like Lufthansa, Delta, and Cathay Pacific are d...
Why Million-Token Context Windows Can't Handle 3 Reasoning Steps 25.04.2026 28:00
The needle-in-a-haystack benchmark is saturated — every frontier model hits 99% on it. But that doesn't mean they can actually reason across long documents. This episode explores four benchmarks that replaced it: RULER, BABILong, NoCha, and Michelangelo. We break down why models that ace million-token retrieval tests collapse at 11,000 tokens on multi-hop reasoning, and what this gap between claim...
LLM Benchmarks Are Full of Noise: Statistical Rigor in AI Evals 25.04.2026 28:03
Almost every model release blog post you've read has a statistical problem. When OpenAI, Anthropic, or Google claims their new model beats the previous one by two points on MMLU, that difference is often well within the noise floor. This episode gets into the weeds on power analysis, McNemar's test for paired evaluations, bootstrapped confidence intervals, and why the decimal-place precision in be...
What Tool-Calling Benchmarks Miss About Production Failures 25.04.2026 27:59
Most tool-calling evaluations hide more than they reveal. This episode breaks down three fundamentally different benchmarks — the Berkeley Function Calling Leaderboard, tau-bench from Sierra Research, and Nexus from Nexusflow — and explains what each one actually measures versus what it misses. BFCL's AST evaluation catches structural errors but is blind to semantic wrongness. Tau-bench grades on...
Choosing Your LLM Eval Framework 25.04.2026 27:26
An opinionated architectural shootout of four major LLM evaluation harnesses — Inspect, Promptfoo, DeepEval, and Braintrust — stress-tested on multi-turn conversations, tool-using agents, async execution, and CI integration. No hedging: we pick winners for research labs, startups, and enterprise teams. Episode #108034 — open it directly at myweirdprompts.com/108034
Geospatial Gold Rush: Who's Hiring Satellite Sleuths? 24.04.2026 23:13
Satellite imagery isn't just for spies anymore. Industries from agriculture to insurance are scrambling to hire analysts who can turn pixels into profits—using tools like QGIS, Planet Labs data, and Python’s geospatial stack. This episode maps out the hottest job markets (spoiler: John Deere paid $300M for this tech), the surprising crossover between war zones and soybean futures, and the exact sk...
Designing Data Models That Mirror Your Work 24.04.2026 22:05
Why off-the-shelf software fails niche small businesses and how building custom tools around your actual workflow—using Airtable, Firebase, or AI agents—can solve the mismatch without coding from scratch. Episode #505584 — open it directly at myweirdprompts.com/505584
Claude Code’s Hidden Context Tax 24.04.2026 23:03
Bigger context windows don’t mean unlimited working memory—Claude’s hidden overhead can quietly degrade performance. This episode breaks down the hierarchy of context costs, from subagents (400-800 tokens each) to lazy-loaded MCP tools (near-free). Learn why eager vs. lazy loading matters more than raw size, how plugin sprawl inflates your startup tax, and practical optimizations to reclaim 25-30%...
When Permanent Means Surviving 400°C 24.04.2026 20:46
Consumer markers fail under heat, chemicals, and abrasion. Industrial markers like the Edding 780 are engineered for survival. This episode explores the chemistry and physics that separate a toy from a tool. Episode #484035 — open it directly at myweirdprompts.com/484035
Your Taste, Your Data: Owning Your AI Preferences 24.04.2026 24:11
We’re terrible at articulating our own preferences but brilliant at recognizing them. Netflix and Spotify exploit this paradox, using our behavioral data to train their recommendation engines—while locking that data away. What if you owned your taste profile instead? This episode explores a radical alternative: a local, portable database of your preferences that any service can query (but never ke...
When Data Becomes the Decision Framework 24.04.2026 24:30
How do emergency responders turn a flood of live feeds into a prioritized action list? This episode explores the technology and philosophy behind dashboards that don't just display data—they become the decision itself. Episode #268265 — open it directly at myweirdprompts.com/268265
Predicting War: The Science of Geopolitical Forecasting 24.04.2026 21:47
What if you could foresee the next major conflict months before the first shot is fired? Geopolitical forecasting blends game theory, expert analysis, and AI to predict wars, economic fallout, and humanitarian crises. From Cold War-era scenario planning at RAND to modern machine learning models crunching satellite data, this episode dives into the methods—and limitations—of anticipating global con...
Engineering a News Pipeline for the Edges 24.04.2026 24:38
How do you build an automated news feed that reliably surfaces underreported developments—like drug price spikes or infrastructure damage—alongside major headlines? This episode explores the technical challenge of balancing consensus news with novel, credible data points in fast-moving conflicts. Episode #107740 — open it directly at myweirdprompts.com/107740
How SITREPs Cut Through Geopolitical Noise 24.04.2026 25:24
Geopolitical news cycles drown us in speculation, but structured formats like the SITREP (Situation Report) force clarity. Developed for military decision-making, this method separates verified facts from analysis, turning overwhelming events into a clear, actionable picture. We break down how to apply SITREP discipline to fast-moving crises—like Taiwan Strait tensions—and explore its tradeoffs co...
Is Israel Really a High-Tax Country? 24.04.2026 20:50
Many assume Israel is a high-tax jurisdiction, but the data tells a different story. This episode compares Israel's tax burden to OECD averages, explores the lowest-tax countries, and examines how taxing income versus capital gains reveals a nation's economic priorities and social contract. Episode #428646 — open it directly at myweirdprompts.com/428646
Similar podcasts
Replaio is not a podcast publisher; show names, artwork and audio belong to their authors and are distributed through public RSS feeds.