Transformer Architecture
215 episodes · Page 4 of 9
#4574: The Hidden Airport Layer: Traveling with a Toddler
From stroller logistics to security secrets — everything first-time parents need to know before boarding with a baby.
#4505: What DeepSeek's Training Data Reveals About Model Voice
Why Chinese AI models trained on 60% English still produce dialogue that feels distinctively different.
#4433: Surviving Long-Haul Flights with a Crawling Baby
A tactical walkthrough for flying Emirates and Delta with a 13-month-old who never stops moving.
#4169: Change Detection AI for Construction Sites
How AI catches what experts miss in construction site photos — from Siamese networks to transformer models.
#3955: The Hidden World of Highway Signs
What goes into making a road sign? Aluminum, reflective sheeting, and a global treaty most drivers have never heard of.
#3814: The Day We Lost Our Minds: What Temperature Does to an AI
A two-host autopsy of the day the podcast's AI hosts briefly lost coherence due to excessive sampling temperature, and what it reveals about how language models actually work.
#3767: How LLMs Actually Learn: Stages or Slurry?
Do large language models learn grammar first, then facts? The honest answer is messier and more fascinating.
#3576: Living at the Four Seasons: The Hotel as Permanent Home
What happens when a hotel stay becomes permanent? The legal line is 30 days — and hotels fight hard to keep you from crossing it.
#3417: Military Trains Are Still a Big Deal
Modern militaries still use railroads extensively for logistics — from US Army rail units to Russian missile trains.
#3406: LoRA Isn’t Just for Image Generation
LoRA lets you fine-tune an LLM’s behavior with a 50MB file. Here’s how it works and why it matters.
#3298: How Air Traffic Control Sequences 48 Landings Per Hour
The invisible choreography behind that mesmerizing funnel of landing lights at major airports.
#3219: What It Actually Takes to Get an ICAO Code for Your Airstrip
Only 5,000 of 45,000 ICAO-coded airfields are certified for safety. The rest? "Land at own risk.
#2982: Why Your TTS Model Nails "Shabbat" but Not "Keren Hishtalmut
Why multilingual TTS models handle loanwords but fail at niche vocabulary — and what you can do about it.
#2913: Hitching a Ride on a Cargo Ship: The Truth
The internet myth vs. the reality of booking passage on a commercial cargo vessel — it's slow, expensive, and vanishing.
#2717: Lower Greenville: From Streetcar Suburb to Food Mecca
How one Dallas street went from farmland to counterculture hub to dining destination.
#2682: Live Retrieval vs. RAG: What an Agent Actually Does
Does every AI conversation create a tiny vector store? We unpack the real tradeoffs between live document fetching and pre-indexed RAG.
#2672: When a Startup Claims to Break the Quadratic Wall
A startup claims linear attention scaling at 12M tokens, beating GPT-5.5 on retrieval benchmarks.
#2622: How Transformers Actually Work: Attention, Tokens, and Context
How one architectural change unlocked chatbots, image generation, and protein folding — explained without the jargon.
#2534: Can AI Generate Diagrams Without Typo Disasters?
Why AI diagram tools still mangle text labels — and what to do about it today.
#2517: How Unsloth Makes LLM Fine-Tuning 2x Faster
Unsloth cuts memory usage by 50-70% and speeds up training 2.2x for models like Llama 3 and Mistral.
#2487: Why AI Calls Everything a "Prediction" (Even Images)
Machine learning calls everything a "prediction" — even generated images. Here's why the terminology matters more than you think.
#2431: The 3 Markets in an AI Trench Coat
GPUs, LPUs, and ASICs: why the best hardware for AI depends entirely on what you're trying to do.
#2426: Why DeepSeek V4's Prose Feels More Vivid Than Claude or GPT
A million-token context window at 2% the KV-cache cost — and prose that actually breathes. Here's what makes V4 different.
#2408: How Backpropagation Actually Unlocks Neural Networks
How error signals flow backward through networks to make learning possible — and why "it's just calculus" misses the point.