Tech & AI
New tools, new models, big claims. What's actually happening in tech and what it means.
Open this topic in the app to listen.
In the mid-2010s, autonomous vehicle developers faced a critical engineering tradeoff: how to filter out background radar clutter without blinding cars to stationary hazards. This episode explores the technical logic behind why early self-driving systems were tuned to discount unmoving objects, and how that design choice manifested in real-world autonomy failures.
In the early nineteen nineties, a German engineering team refined the MP3 standard using a single a cappella recording by Suzanne Vega. By playing 'Tom's Diner' thousands of times to hunt for digital artifacts, they turned a human voice into the ultimate yardstick for audio compression.
A celebrated startup promised to build software using artificial intelligence, but behind the curtain sat hundreds of engineers writing code by hand. This is the story of Builder.ai and the rise of 'AI washing,' where human labor is hidden to inflate valuations until the illusion cracks.
In twenty nineteen, OpenAI taught a robotic hand to solve a Rubik's Cube using a technique called domain randomization. By training entirely in millions of randomized simulations, the system learned to bridge the gap between digital code and physical reality, revealing a new path for robotic dexterity.
In twenty nineteen, OpenAI researchers set two teams of simulated agents to play hide-and-seek millions of times. What emerged wasn't just a simple game of tag, but a sophisticated arms race of tool use, physics exploits, and 'box surfing' that revealed the alien logic of reinforcement learning.
For decades, industrial control systems—from water treatment plants to power grids—shipped with universal default logins that were rarely changed. This episode explores how a design choice intended for convenience became a systemic vulnerability, leading to real-world infrastructure breaches and a fundamental shift in how we secure the physical world.
In 2015, Google's image recognition system labeled photos of Black people with a racist slur, and the company's immediate fix was to delete the category entirely. This episode traces how a single mislabel exposed the gap between machine learning promises and training data reality, and why erasing a symptom is easier than solving the underlying bias.
In September nineteen eighty-three, the Soviet Oko early-warning system reported five incoming American missiles. This is the story of the technical blind spot where sunlight and clouds mimicked a nuclear launch, and the officer who chose to trust his intuition over the machine.
What's actually happening in technology, without the hype cycle. The stream covers how large language models really work, why AI hallucinates, the single company whose machines make every advanced chip possible, the energy and water bill behind the AI boom, and the booms and busts of the AI winters that came before this one. When a new model or a big claim lands, fresh episodes follow — and anything you're wondering about can become one too.
Fylom episodes are researched, written, and voiced by AI. Automated checks help catch inaccuracies, but episodes aren't reviewed by a human and AI can still get things wrong. Treat them as a starting point, not a source of record — more in our accuracy disclaimer.