It’s the first Saturday of the year!

The first Five Things Tech of the year starts with the depressing realisation that Trump has become Big Tech’s favourite useful idiot, tearing down AI safeguards, relaxing chip export rules and even waving through more powerful Nvidia silicon for China while the same CEOs who once postured about responsibility now just smile and cash in, and at the same time Meta has quietly turned Reels into a 50-billion-dollar business that keeps people glued to Instagram even longer than YouTube Shorts, all running on gigantic software systems that almost nobody fully understands but everyone prays will keep working somehow.​

On the slightly more hopeful side, clever people are teaching headphones “semantic hearing” so you can spotlight one voice and fade out the rest of the noise like an overenthusiastic sound engineer, and over here in Europe we once again promise to become the global DeepTech powerhouse with forecasts of a trillion dollars in enterprise value while still trying to fund insanely capital-intensive research startups with the same shallow VC pockets and software-style metrics that have held us back for years. We’ll see how that goes!

Welcome back to Five Things Tech!

From A.I. to Chips, Big Tech Is Getting What It Wants From Trump

Yet upon taking office, Mr. Trump vowed to continue a fight to break up Meta, imposed tariffs that would raise the costs of Apple’s supply chains and restricted the exports of artificial intelligence chips from Nvidia and other chip makers. It seemed that the tech industry’s efforts to woo the president would not pay off.

Now, however, the biggest tech companies have gotten almost everything they wanted from Mr. Trump.

Since the summer, he has eliminated many limits on A.I. chip exports, fast-tracked the building of data centers that power A.I. development and pushed legislation that gave government approval to a type of cryptocurrency. This month, Mr. Trump signed an executive order to kill A.I. restrictions set by states and greenlighted sales of a more powerful Nvidia chip to America’s top rival, China.

It is just appalling. But at least we know now that the Big Tech CEOs are just spineless guys who bend over backwards to make more money. It’s the EU that needs to enforce global regulation now. Again.

(…continue reading.)

How Meta’s Reels Became a $50 Billion Business

Instagram had to figure out how to do that much harder task, too.

It focused on promoting original content and paying creators to post on the platform. As people spent more time scrolling Reels, the algorithm got better at predicting what users wanted to see.

Five years on, something has started to click.

The average Instagram user is now spending 27 minutes a day watching Reels, versus YouTube Shorts users’ watching for 21 minutes on that platform, according to estimates from market intelligence firm Sensor Tower. (TikTok is still king with the average user spending 44 minutes a day scrolling its main feed.)

I too sometimes share reels with my family members, especially when I am on the subway or taking some time in the bathroom…

(…continue reading.)

Nobody knows how large software products work

Non-technical people - at least, ones without a lot of experience working with software products - often believe that software systems are well-understood by the engineers who build them. The idea here is that the system should be understandable because it’s built line-by-line from (largely) deterministic components.

However, while this may be true of small pieces of software, this is almost never true of large software systems. Large software systems are very poorly understood, even by the people most in a position to understand them. Even really basic questions about what the software does often require research to answer. And once you do have a solid answer, it may not be solid for long - each change to a codebase can introduce nuances and exceptions, so you’ve often got to go research the same question multiple times.

Ok, good that somebody finally admits it. But it’s true. When I was CTO I had to rely on so many components to just work, and like magic, they somehow did. Mostly…

(…continue reading.)

Welcome to the Future of Noise Canceling

One of the startup’s first big innovations was “semantic hearing,” which was the first project they approached, around three years ago. The team built a hardware prototype—a pair of on-ear headphones with six microphones across the headband, connected to an Orange Pi microcontroller—to test out a model that had been trained to recognize 20 different types of ambient sounds. This included things like sirens, car horns, birdsong, crying babies, alarm clocks, pets, and people talking, and then allowed the user to isolate say, one person’s voice as a “spotlight,” and block out all the other frequencies.

This kind of technology is truly fascinating as it blends audio signals with our surroundings and kind of remixes how and what we hear.

(…continue reading.)

Can Europe become the global centre of gravity for DeepTech?

Forecasts suggest that European DeepTech could generate around one trillion dollars in enterprise value by 2030. While DeepTech as a whole is attracting a record share of European funding, early-stage rounds are down by 30% since their peak in 2021.

This also happens to be the stage where founders face some of their toughest hurdles. Long development cycles, highly technical milestones, and limited commercial traction make traditional venture investors hesitant. Many VCs acknowledge the importance of DeepTech but still default to evaluating companies through software-style metrics.

DeepTech is certainly on the rise, but I am not so convinced that the notriously shallow pockets of European VC will be enough to carry research-heavy startups all the way to profitability and beyond.

(…continue reading.)

That’s all for now! Thanks for reading! If you missed last week’s Five Things Tech, you can find it here:

Five Things Tech: Tech Flops, RSS, AI Buildout, Electricity, SaaS December 20, 2025 Read full story

🤖

— Nico