My Weird Prompts

Jerusalem's statutory land-use plan, the Taba, is a legal object that no human mind has ever fully contained. It carries a British mandatory skeleton, Ottoman land law in the basement, and 171 amendments stacked on top. On this episode, we trace how a single Rehavia lot can be governed by four overlapping plans, why 60% of local plans between 2016 and 2022 were authorized as spot-planning, and what Paris's single-document PLU reveals about whether complexity is necessary — or whether the planning machine now exists partly to sustain itself.

Key sources:

• OECD, 2025-08-04 — https://www.oecd.org/en/publications/redefining-spatial-planning-and-development-in-israel_89036c98-en/full-report/the-institutional-framework-for-spatial-planning-and-development-in-israel_2fa805d6.html

• Paris bioclimatic PLU revision, 2024 — https://cdn.paris.fr/paris/2024/11/07/2024-du-142-delib-mFHZ.pdf

• Rachelle Alterman, Technion Platform of Experts in Planning Law [primary/expert] — https://internationalplanninglaw.net.technion.ac.il/countries/israel/

Episode 6018 — myweirdprompts.com/n/6018 (permanent short link: myweirdprompts.com/471451)

Israel's average time to a building permit hit 319 days in 2019 — roughly three times the Western average of 114 days. But where does that year actually go? This episode walks the full pipeline for private houses, apartment buildings, and high-rises: the five strictly sequential approval stages, the relief track that costs architects a year, and the automatic-rejection rule that quietly resets the clock. Then it compares Israel against Dallas, Germany, and the U.K., and digs into the deeper question — prescriptive versus performance-based codes, and why Israeli architects are expected to be legal experts in a law nobody taught them in school.

Key sources:

• Israel Planning Administration citizen guide, updated 2023-04-13 — https://www.gov.il/he/pages/guide_citizen?chapterIndex=2

• Dallas Building Permit Review Time Summary, updated 2020-10-21 — https://dallascityhall.com/departments/sustainabledevelopment/buildinginspection/DCH%20documents/pdf/PermitReviewTime_102620.pdf

• Dallas RSVP program bulletin, 2022-12-27 — https://dallascityhall.com/departments/sustainabledevelopment/DCH%20documents/Service_First_Bulletins/Service_First_Bulletin%20_201_Final_12.27.22.pdf

Episode 6017 — myweirdprompts.com/n/6017 (permanent short link: myweirdprompts.com/709547)

Seventy billion parameters gets repeated on every earnings call, but almost nobody who says it could tell you what one parameter is. This episode builds the artificial neuron from the ground up — weights, bias, activation — then works through the math objects underneath it: scalars, vectors, matrices, and tensors. Along the way, the conversation lands on why stacked linear layers collapse into a single matrix without a nonlinearity, why the word "rank" means three unrelated things depending on who's talking, and which parts of the biological neuron the artificial one actually kept. It closes by stress-testing the metaphor against a real transformer, where the perceptron equation survives intact but the perceptron as a unit of meaning does not. By the end, that number with a B on the end feels like something you could almost picture.

Key sources:

• Anthropic, Towards Monosemanticity, published 2023-10-04 — https://transformer-circuits.pub/2023/monosemantic-features

• Goodfellow, Bengio, Courville, *Deep Learning* Ch. 2 — https://www.deeplearningbook.org/contents/linear_algebra.html

• François Chollet, *Deep Learning with Python* Ch. 2 — https://deeplearningwithpython.io/chapters/chapter02_mathematical-building-blocks/

Episode 6016 — myweirdprompts.com/n/6016 (permanent short link: myweirdprompts.com/808714)

What actually happens when someone wires you a million dollars? The answer splits in two: the number is trivial, but the plumbing is a whole industry. This episode walks through three transfers — a hundred dollars, a million, and a billion — and shows where the paths diverge. Along the way: why money never physically moves, how CHIPS netting meant one large payment never had to move as a large payment, why the Fed's instant rails now reach ten million dollars, and what really stops a ten billion dollar transfer (hint: it isn't the message format).

Key sources:

• Fedwire Funds Service Product Sheet — https://www.frbservices.org/binaries/content/assets/crsocms/financial-services/wires/funds.pdf

• Fedwire Funds Service page — https://www.frbservices.org/financial-services/fedwire-funds-service/

• FedNow transaction limit increase to $10M — https://www.frbservices.org/news/press-releases/090525-fednow-transaction-limit-increase/

Episode 6015 — myweirdprompts.com/n/6015 (permanent short link: myweirdprompts.com/104972)

Encoder-only, encoder-decoder, decoder-only — three shapes, one skeleton. This episode pulls apart the architectural fork at the center of modern language models: how attention masking and training objectives separate BERT from GPT, why the decoder-only design swallowed the field, and why encoder-only models still quietly run classification and retrieval at a fraction of the cost. Along the way: the four unrelated things this field calls "encoding," the parts bin that top labs actually assemble models from, and where the real architectural innovation is happening now.

Key sources:

• Google Developers Blog, T5Gemma: A new collection of encoder-decoder Gemma models, 2025-07-09 — https://developers.googleblog.com/en/t5gemma/

• Yi Tay, What happened to BERT & T5? On Transformer Encoders, PrefixLM and Denoising Objectives — https://www.yitay.net/blog/model-architecture-blogpost-encoders-prefixlm-denoising

• arXiv:2510.26622, Encoder-Decoder or Decoder-Only? Revisiting Encoder-Decoder Large Language Model, 2025-10-30 — https://arxiv.org/abs/2510.26622

Episode 6014 — myweirdprompts.com/n/6014 (permanent short link: myweirdprompts.com/609796)

A listener asked a deceptively simple question: what actually happens during an LLM forward pass? This episode walks through the whole thing — input tensors, linear layers, why nonlinearities are the only reason depth means anything, and how logits become a prediction. Then it gets to the part that trips people up: a classifier produces its answer in a single forward pass, but an autoregressive model needs one pass per token, because each token depends on the one before it. The architecture isn't different. The dependency chain is. Along the way: why training can score many predictions at once while inference can't, why generation latency has a hard floor no GPU can fix, and how speculative decoding sneaks around it.

Episode 6013 — myweirdprompts.com/n/6013 (permanent short link: myweirdprompts.com/624779)

The model scores; the engine chooses. That split is the whole story of autoregressive decoding, and this episode walks through it token by token. Starting from a warm KV cache, we trace a single decode iteration — the forward pass, the query-key-value projections, the logits row that actually gets used, and the sampling step that feeds the next token back in. Along the way we work a concrete example ("The cat sat.") to show why keys and values get cached but queries never do, and why skipping the cache would mean reprocessing over eight million positions to generate a four-thousand-token answer. Then we put the loop on a GPU, where the real bottleneck isn't arithmetic at all: it's memory bandwidth, and the only parallelism that exists is across independent requests.

Key sources:

• All About Transformer Inference, JAX/Google DeepMind scaling book, published Feb. 4, 2025 [primary/technical] — https://jax-ml.github.io/scaling-book/inference/

• Sebastian Raschka, What is a KV cache… [technical] — https://sebastianraschka.com/faq/docs/kv-cache.html

• Transformer 101, Inference [technical] — https://www.transformer101.com/transformer-inference

Episode 6012 — myweirdprompts.com/n/6012 (permanent short link: myweirdprompts.com/325414)

A single request to an LLM isn't one computation—it's a stack of stages wearing a trench coat. This episode walks the full pipeline: flattening a chat transcript into one flat document with special role tokens, tokenizing text into IDs, the massively parallel prefill pass that produces your first token, and the strictly sequential decode loop that produces every token after. Along the way: why forgetting one flag called add_generation_prompt makes the model write your half of the conversation, why "encoding" and "decoding" each mean two different things, and how a KV cache can quietly outgrow the model weights it's serving. If you've ever wondered where your time-to-first-token actually goes, this is the tour.

Key sources:

• Nick Gustafson, What an Inference Engine Actually Does — https://thegustafson.com/blog/what-an-inference-engine-does

• BentoML, How does LLM inference work? | LLM Inference Handbook — https://bentoml.com/llm/llm-inference-basics/how-does-llm-inference-work

• JAX-ML / Google DeepMind, All About Transformer Inference | How To Scale Your Model, published Feb 4 2025 — https://jax-ml.github.io/scaling-book/inference/

Episode 6011 — myweirdprompts.com/n/6011 (permanent short link: myweirdprompts.com/350177)

Wearable tech promised to change how we live. It gave us step counters, a haptic buzz when someone likes your photo, and a robot that feeds you tomatoes mid-marathon. This episode tours ten of the strangest wearables ever manufactured — not failures, exactly, but objects built with complete sincerity that look absurd on a human body. From Necomimi's brainwave-controlled cat ears (a real product, torn down by Adafruit, sold at Comic-Con for $99.95) to the 1880 umbrella hat that's still solving a real problem for wheelchair users, the through-line is what we call the seriousness gap: nobody at any of these companies was winking. Along the way we hand out three entirely unofficial prizes — Most Unnecessary, Weirdest Overall, and Most Surprisingly Useful — and argue about why "ridiculous" and "useless" are not the same word.

Key sources:

• NeuroSky Support — https://support.neurosky.com/kb/necomimi/how-does-necomimi-work-what-do-the-ear-movements-mean

• NeuroSky Support — https://support.neurosky.com/kb/necomimi/what-inspired-necomimi-to-be-developed

• Google Patents — https://patents.google.com/patent/US8944300B1/en

Episode 6010 — myweirdprompts.com/n/6010 (permanent short link: myweirdprompts.com/346314)

Vertex AI was Google's 2021 bet on unifying ML tooling — pipelines, feature stores, model registry — into one platform. In April 2026, Google renamed it the Gemini Enterprise Agent Platform and said all Vertex services would be delivered exclusively through it. So what actually changed? This episode traces the lineage from AutoML and AI Platform through the 2021 launch to today's agent-first rebrand, untangles the five Google products that all sound identical, and lands on a decision rule for when the enterprise endpoint is worth it over the plain Gemini Developer API — plus how it stacks up against Bedrock and Azure.

Key sources:

• Gemini Enterprise Agent Platform docs , accessed 2026-10-15 — https://cloud.google.com/vertex-ai/docs

• Google Cloud Blog, 2026-04-22 — https://cloud.google.com/blog/products/ai-machine-learning/introducing-gemini-enterprise-agent-platform

• name-changes table , 2026 — https://docs.cloud.google.com/gemini-enterprise-agent-platform/vertex-ai-name-changes

Episode 6009 — myweirdprompts.com/n/6009 (permanent short link: myweirdprompts.com/511455)