Nov 19, 2026 · 6 min listen · Last updated November 19, 2026
From storyflo. This is your daily audio brief. Hey, it's Theo. November 19th. Five things in tech that mattered this morning — let's start with the one that surprised me most. Let's get into it. First, from jmarriott.
Listen · storyflo · tech
Daily Tech Brief · November 19th
0:00-6:18
Pick your daily storyteller
Subscribe to match with Theo, Jessica, Chloe, Mason, Brock — your voice, every brief.
Audio pre-rendered by Storyflo · cached + delivered from the edge
Martin Luther rewired your brain, the AI memorisation crisis, The mysteries of global shipping, A Walk After Dark and the genius of George Orwell
Welcome to Cultural Capital! This week in The Times I wrote about the twenty-first century decline in philanthropy and the age hyper-individualism. I’m increasingly persuaded the idea that the individualist chaos of modern societies may have at least something to do with the fading of Christian-derived moral ideas about charity, humility and turning the other cheek. Ideas now replaced by the omnipresent individualist clichés about the importance of “being yourself” and “following your dreams”.
Is parliament now post-literate?, beware declining superpowers, thankfulness, awe and wonder, and how AI destroys institutions
Welcome to Cultural Capital! In The Times this week I wrote about the under-discussed problem of boomer phone addiction. Everyone panics — rightly — about kids and phones. But there is evidence that total screentime among over 65s (including TV and iPads) may be even higher than it is for users in their twenties. I missed this piece in The Economist from last year. In the UK parliament speeches are getting shorter and less complex.
The return of nepotism, the death of the secretary, the genius of Shakespeare, how AI could save public discourse
Welcome to Cultural Capital! My column in The Times this week is about the virtues of doubt. I don’t know how anyone can feel dogmatically certain about anything in a world that is undergoing a series of profound economic, political and technological transformations: Dogmatism is a luxury of easy times. We can all strike impressive attitudes when there’s nothing at stake. It’s why so many university students are Leninists.
“Did you never invest in US tech stocks?” a young friend, all of 23 and fresh into the world of investment banking, asked recently. The question’s not invalid. Over the past decade or so, technology has made the world wealthier than ever. A very close friend who recently got promoted to a mid-senior managerial role at a top Big Tech firm whimsically mused that not a single member of his family had ever breached the ₹1-crore salary threshold before him. He wasn’t wrong—for most in India, this kind of salary was unimaginable for the most part, beyond the very limited few.
So, Redis has this new thing called Iris, and it's pretty neat for AI agents. You know how sometimes you ask a support bot the same thing in different ways? Like "where's my refund?" versus "did my refund go through?" A regular system might just store the answer, but what if the answer changes, like with a refund status? That's where Iris comes in. Instead of just storing the answer, it stores the *steps* to get the answer. So, when you ask, it doesn't just give you an old answer; it actually goes and checks the live data, like the current refund status, before replying. It's like reusing the recipe instead of just reusing a pre-made dish, which means it's always fresh and accurate.
And about GPUs, it turns out they're not just about raw speed. A big bottleneck is how fast they can actually *get* the data they need to do calculations. Think of it like having a super-fast chef but a really slow delivery service for ingredients. The chef can chop and mix incredibly quickly, but they're always waiting for the next item to arrive. This is especially true when generating text, where the model has to constantly fetch its "weights" – basically, its knowledge – from memory.
To make things faster, GPUs are designed to do more work with the data they *do* manage to fetch. When you process multiple requests at once, the GPU can use the same data for several tasks, making better use of what it has. Plus, they have this layered memory system. There are super-fast, tiny storage spots right next to the processing units, then slightly bigger caches, and then the main, slower memory where all the big stuff like model weights are stored. The trick is keeping the data you're using often in those faster, closer spots.
This is why something like FlashAttention is cool. Instead of loading a huge table of data into the main memory and then pulling it back out, it works on smaller chunks and keeps the intermediate results close by. It's all about minimizing how much data has to travel back and forth from that slower main memory. So, it's not just about how many calculations a GPU can do, but how efficiently it can move and reuse data.
As you all know, I’m a senior staff software engineer at Meta, and that job is incredibly time-intensive. But my passion, the thing I do for fun, is playing with AI tools and making videos about them. Honestly, I wouldn’t be able to do any of this without AI now. My entire workflow is fully agentic, and I’m going to show you how it works, in the video / …
As the Ukraine War enters it’s 5th year, and the Oct 7th Wars start their 4th… we’re sorta getting a feel for the ultimate shape of modern drone warfare, AI , and its limits. I’ve written previously on the use of AI surveillance against insurgents and assassins and how to counter them in my hit pieces “Getting Away With It: The Art of Urban Infiltration and Exfiltration” and “Tactical Espionage DollarStore Hobo”… But now I want to address conventional warfare in the Cyberpunk 2020s we find ourselves in… Lets discuss “AI Drone Swarms”… and why they don’t and almost certainly won’t exist. . .
Google AI Ultra plan brings ad-free YouTube to the whole family
Google sweetens its pricey AI Ultra plans with perks that go well beyond Gemini, from generous cloud storage to bundled subscriptions. Earlier this year, it introduced a new $100-per-month Ultra tier and upgraded both Ultra plans to bundle YouTube Premium Individual subscription for free. Now, Google is making the plan even more attractive by upgrading the bundled YouTube Premium subscription from Individual to Family.
New Study Shows How Young People Feel About AI Companions
An anonymous reader quotes a report from 404 Media: A new study from Gallup looked at how young people across the world are making friends, both digitally and in person. The good news is that around 93% of the young people surveyed reported having at least one strong in-person friendship. The weird news is that 33% said they had an AI companion. Gallup's study is lengthy and nuanced. It partnered with the Boston Children's Digital Wellness Lab to conduct the study and used online surveys of more than 16,000 young people from the ages of 15 to 22 in 16 different countries.
Discord Rewrites Desktop and Mobile Clients in Rust Using Zed's GPU-Accelerated Rendering Framework
Discord’s CTO Stanislav Vishnevskiy recently announced rewriting its desktop and mobile clients in Rust, moving away from Electron and React Native. The project eliminates layers of abstraction to directly leverage the device’s GPU, native mobile views, and a Rust core shared across platforms. Early results hint at a significant reduction in RAM usage and boot time for chat, voice, and streaming.