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
AI Wargame Memory: Beyond the Context Window 09.04.2026 34:47
In multi-agent wargaming, an AI general must remember decisions made forty-seven turns ago without dumping the entire conversation history into context every single turn. This episode explores the three-layer memory architecture required for serious simulations: shared world state, private context, and persistent long-term memory. We examine why naive approaches like full-history replay fail due t...
Housing as National Defense in Israel 09.04.2026 16:17
In Israel, political discourse has long been dominated by security concerns, but a growing movement is trying to shift the focus to the cost of living. This episode explores how grassroots organizers and politicians are reframing housing affordability and economic anxiety as issues of national resilience. By co-opting the language of security and building cross-partisan coalitions, they aim to bre...
Wargaming's Methodology, Not Magic 09.04.2026 33:01
Before plugging personas into an LLM, it helps to know what makes a wargame a serious decision-support tool. This episode traces the history and standards of professional wargaming—from the Naval War College and RAND to MORS and CSIS—and explains why most AI simulations skip the rigor of adjudication, repeatability, and structured output. We explore the difference between insight and prediction, w...
The Brutal Problem of AI Wargame Evaluation 09.04.2026 27:34
AI wargame simulations are moving from research labs into real policy planning, but how do we know they actually work? This episode explores the brutal evaluation problem: when simulating future crises, there's no ground truth to compare against. We walk through five candidate methodologies—backtesting, inter-run consistency, expert red-teaming, predictive calibration, and process validity—and rev...
Is Your AI Wargame Signal or Noise? 09.04.2026 29:47
As AI wargaming moves from hobbyist projects to policy workflows, the methodology behind running simulations becomes critical. This episode explores the tension between deterministic and stochastic runs, how temperature settings affect actor behavior, and why single-run simulations systematically underestimate risk. We break down the minimum viable run counts for different levels of rigor and tack...
The Fog-of-War Problem in AI Wargaming 09.04.2026 28:08
When both sides of a wargame run on the same AI model, how do you prevent information leakage? This episode explores the unique "fog-of-war" challenge in AI wargaming, where shared training data and inference servers create new vulnerabilities for accidental intelligence leaks. We examine real-world failure cases, including a 2025 RAND simulation where referee narration accidentally revealed class...
Engineering Geopolitical Personas: Beyond Caricatures 09.04.2026 29:22
What does it take to make an LLM convincingly play a geopolitical leader like Putin or Khamenei? This episode explores the full technical stack for building personas with strategic fidelity, moving beyond caricature to capture decision-making logic. We break down the layers: system prompting with doctrine, few-shot examples for voice, RAG for historical memory, and fine-tuning for character. The d...
The Referee's Dilemma: Epistemic Containment in LLM Simulations 09.04.2026 31:00
How do you keep a geopolitical crisis simulation from collapsing into a news commentary engine? This episode explores the critical role of the referee in curating a sealed world state, the subtle biases that can corrupt results, and why epistemic containment is the core engineering challenge. Episode #761228 — open it directly at myweirdprompts.com/761228
The CIA Is on GitHub 09.04.2026 37:16
The CIA's venture arm, In-Q-Tel, has an open-source wargaming project on GitHub. This episode explores why the intelligence community is embracing public AI tools, how Snowglobe generates narrative crisis simulations, and what it means for privacy and public-private partnerships. Episode #198205 — open it directly at myweirdprompts.com/198205
The Prediction Market Leak: Who Profited from the Iran Ceasefire? 09.04.2026 38:05
Hours before the Iran ceasefire was announced, anonymous accounts made millions on prediction markets. Was it brilliant analysis or an inside leak? Our panel forecasts the next 30 days through the lens of this unanswered question. Episode #263378 — open it directly at myweirdprompts.com/263378
Shifting Left on Hallucinations 09.04.2026 22:58
The era of vibe-based AI is ending. This episode explores how the industry is moving from post-hoc verification to architectural prevention—treating hallucinations as system errors, not creative flourishes. We examine tools like Guardrails AI, NeMo, and specialized judge models that catch fabrications before they leave the pipeline. Episode #795197 — open it directly at myweirdprompts.com/795197
The Victory Siren Sounds, But the Shelter Door Is Still Open 09.04.2026 34:20
When Netanyahu announced the end of Iran's threat, Daniel was still running to a bomb shelter. This episode explores the psychological and political gap between wartime victory narratives and civilian reality. We examine how information asymmetry, cognitive stress, and the evolution of media environments create a uniquely demoralizing experience for citizens caught between official statements and...
When the Siren Stops, the Brain Keeps Screaming 08.04.2026 18:26
When a conflict stretches from a twelve-day sprint to a six-week marathon, the human nervous system hits a breaking point. This episode explores the profound biological toll of living siren-to-siren, where the brain's ancient alarm system gets stuck in the "on" position. We examine how chronic hypervigilance degrades sleep, suppresses the immune system, and rewires the brain's predictive models. P...
Why Auto Wi-Fi Settings Fail You 08.04.2026 23:07
Most prosumer Wi-Fi problems come from trusting 'Auto' settings. This episode explains the physics of channel overlap, asymmetric links, and why high transmit power often makes your connection worse. Episode #742079 — open it directly at myweirdprompts.com/742079
Why Agentic Chunking Beats One-Shot Generation 08.04.2026 17:42
For years, generating long-form content with AI has been plagued by "token fatigue" and repetitive loops. This episode dives into the specific architecture—using a Planning Agent and Subagents with Claude Sonnet 4.6—that solves the context dilution problem. Learn why naive one-shot prompting fails for deep dives and how to structure a digital production team for books, briefs, and podcasts. Episod...
The Flashlight You Actually Need 08.04.2026 39:53
A power outage at 2 AM reveals the gap between a $15 hardware store torch and a purpose-built tool. This episode breaks down what actually matters in a flashlight for camping, emergencies, and home use—beyond the lumen wars. We cover the five brands worth trusting, the real baseline spend for reliability, and why battery tech and build quality matter more than marketing numbers. Episode #396200 —...
Human Reaction Time vs. AI Latency 08.04.2026 22:21
In the race for faster AI, engineers are burning compute to shave milliseconds off inference times. But there's a biological bottleneck that no amount of code can fix. This episode dives into the "Bio-Floor" of human reaction time—exploring the baseline of 250ms, how fatigue and alcohol degrade performance, and why sub-100ms optimizations are often invisible to users. Learn when it's time to stop...
Israel’s Pivot: From Europe to the Middle East 08.04.2026 34:53
The recent war with Iran has forced a dramatic geopolitical realignment in the Middle East. While headlines focused on missiles, the real story is the deepening integration between Israel and its neighbors, moving from ceremonial accords to an existential "Defense-Tech Corridor." This episode explores how shared security threats and economic gravity are creating a new regional bloc, examining the...
Israel's Russian Paradox: Arming the Enemy 08.04.2026 27:39
How can Israel maintain normal diplomatic relations with Russia while Moscow arms Iran with advanced air defense systems? This episode explores the strategic paradox at the heart of the Russia-Iran-Israel triangle. Episode #348639 — open it directly at myweirdprompts.com/348639
The 14-Day Ceasefire: A Tactical Halt, Not Peace 08.04.2026 29:22
Is it a ceasefire or just a pause in the fighting? As Israel and Iran agree to a 14-day halt in hostilities, we explore the fragile mechanics behind this "tactical timeout." From military logistics and intelligence gathering to the role of mediators and the risk of escalation, this episode unpacks why this truce may be more about repositioning than peace. Episode #393562 — open it directly at mywe...
Will Iran's Regime Collapse in a Year? 08.04.2026 11:39
After the April ceasefire and Khamenei's death, headlines say Iran's regime is collapsing within a year. But what does structural reality on the ground actually look like? We analyze the three pressures facing Tehran—economic strangulation, military degradation, and a legitimacy crisis—and explore why the IRGC may be consolidating power rather than fracturing. From the "democracy paradox" to the r...
Why a 14-Day Ceasefire Isn't Peace—It's a Reload 08.04.2026 39:29
When the sirens stop, the real work begins. This episode dissects the 14-day ceasefire not as a diplomatic breakthrough, but as a strategic pause dictated by manufacturing limits and logistics. We explore the "Interception Trap"—the staggering cost of defense versus offense—and the physics of rocket propellant curing that can't be rushed. From Israel's urgent need to replenish interceptor stockpil...
The Disciplined Engineering of Urban Search and Rescue 08.04.2026 24:29
When a building collapses, rescuers don't just dig. They perform structural triage, stabilize wreckage, and listen for micro-sounds. This episode explores the systematic protocols and engineering that turn chaos into a race against time. Episode #185267 — open it directly at myweirdprompts.com/185267
The Mathematical Nightmare of Air Defense 08.04.2026 21:00
When a single cheap missile can overwhelm a multi-million-dollar defense system, commanders face impossible choices. This episode explores the capacity gap that is reshaping air warfare in the 2030s. Episode #846378 — open it directly at myweirdprompts.com/846378
Why AI Answers Differ Even When You Ask Twice 07.04.2026 25:06
Why does an AI give you different answers to the exact same question? This episode dives into the trillion-dollar problem of AI non-determinism. We explore why "Temperature Zero" isn't enough, how GPU parallel processing causes numerical drift, and why your server's workload might be changing your code. Plus, learn the engineering workaround—moving determinism downstream—that developers use to bui...
Similar podcasts
Replaio is not a podcast publisher; show names, artwork and audio belong to their authors and are distributed through public RSS feeds.