How do you teach relational database design using only your voice? This episode explores why SQL concepts like foreign keys and constraints are fundamentally about relationships you can describe in plain English — no diagrams needed.

Episode #842994 — open it directly at myweirdprompts.com/842994

Where does "public data" end and "unauthorized access" begin? This episode traces the origins of cybercrime prosecution from the 1986 Computer Fraud and Abuse Act through landmark cases like hiQ Labs vs. LinkedIn and Van Buren vs. United States. We explore how laws born from the moral panic of WarGames still govern a world of APIs, scraping, and unauthenticated endpoints — and how courts are finally drawing clearer lines around technical authorization versus purpose. Plus, how the UK, EU, and Israel handle the same questions differently.

Episode #031396 — open it directly at myweirdprompts.com/031396

Can you teach someone to program using only spoken instructions? This episode puts that question to the test as two hosts coach a complete beginner through building a real TypeScript program, explaining every step and mistake along the way.

Episode #454132 — open it directly at myweirdprompts.com/454132

In this episode, we take on a challenge: teach someone who has never written a single line of code to build a working Python program from scratch using only our voices. We choose the guess-the-number game because it teaches importing modules, variables, user input, type conversion, conditionals, loops, and f-strings in just seven lines. We walk through every character, punctuation mark, and indent across five stages—from "Hello, World!" to a fully interactive game—and cover essential setup like installing Python, choosing a text editor, and running your program from the terminal. No prior experience required.

Episode #448942 — open it directly at myweirdprompts.com/448942

Most developers have only a fuzzy sense of what happens between their fingertips and the silicon. This episode traces a single Python `print("Hello")` statement through every layer of abstraction: CPython bytecode compilation, the virtual machine loop, glibc buffering, the system call boundary, the Linux kernel's VFS and terminal driver, and finally the hardware itself. We contrast this with C and Rust's approaches, examine why Python generates 562 system calls vs C's 34, and explore what "high-level" and "low-level" actually mean concretely — not as textbook definitions, but as countable layers with real costs. Understanding where your abstractions live and what they cost is no longer academic; it's a practical engineering decision for cold starts, edge computing, and systems programming.

Episode #667195 — open it directly at myweirdprompts.com/667195

When you open Chrome DevTools and watch the XHR requests fly by, you'll often find dozens of unauthenticated JSON endpoints sitting behind polished frontends. Are these data leaks, intentional public APIs, or just the natural plumbing of modern single-page apps? This episode explores what happens when LLM agents like Claude systematically discover and document undocumented APIs — and why the old assumption that "if it's not documented, it's private" no longer holds. We examine the spectrum from benign public data endpoints to genuine Broken Object Level Authorization vulnerabilities, the novel attack surface created by agent-driven DevTools access, and why every developer should adopt a "public by default" mindset for frontend-consumed APIs.

Episode #946770 — open it directly at myweirdprompts.com/946770

Want an LLM that doesn't default to cheerful, hedged over-explaining? This episode unpacks the state-of-the-art recipe for baking real personality into model weights — not just system prompts. We break down the Grumpy Italian Chef case study (a 1.2B model trained in 15 minutes on a consumer GPU), explain the SFT + DPO pipeline, and explore how much data you actually need for style transfer vs. robust persona alignment. Plus: tooling options (Unsloth, LlamaFactory, Axolotl), the beta personality dial, and the philosophical question of whether different alignment is misalignment.

Episode #078363 — open it directly at myweirdprompts.com/078363

This episode explores a listener's innovative approach to building a structured dataset from voice-dictated AI prompts. Daniel's method iterates between hand-annotating gold rows, running Claude Sonnet 4.6 with few-shot exemplars, and diffing outputs to find which rows changed most between iterations. We trace this to prior art in Active Prompt Engineering (APE), discuss why inter-iteration prediction change is a clever computational hack, and examine its blind spots — particularly rows that are consistently wrong across iterations. The conversation covers convergence criteria, the economics of active learning with cheap inference, and the critical distinction between converging a prompt versus converging an exemplar selection strategy. We also address few-shot leakage, held-out evaluation, and whether publishing a prompt as an artifact is meaningful without rigorous evaluation.

Episode #735620 — open it directly at myweirdprompts.com/735620

When you put structured data on GitHub, who's your real audience — future humans or AI agents? This episode explores Daniel's clever workflow of using public repositories as agent-accessible context, and the deeper question it raises about parallel documentation standards. We break down the emerging landscape of llms.txt, agenticweb.md, and AGENTS.md files, the surprising truth about whether any AI actually reads them, and why JSON (specifically NDJSON) is becoming the default format for agent consumption. Plus: the trust problem with agent-targeted content, the convergence thesis, and practical advice for anyone publishing information in an era where both humans and machines need to understand it.

Episode #661379 — open it directly at myweirdprompts.com/661379

How agentic code tools like Claude Code transform library selection from a high-stakes quarterly ritual into a fast, iterative loop—and why documenting decisions for both humans and future agents is the new challenge.

Episode #027529 — open it directly at myweirdprompts.com/027529