Headworx

The energy crisis was the leading theme throughout the 2025 COSM conference. And people were really crying rivers about it. With the two most severe loses articulated: We will not be able to scale the AI infrastructure sufficiently enough China will overtake us Both are - I would say - purely "American" worries. First of all, nobody knows if AI progress is really limited by energy. Dwarkesh Patel has recently discussed this with Ilya Sutskever - and the headline is "We're moving from the age of scaling to the age of research". Returns from scaling out the AI infrastructure have been diminishing. Ilya argues we need to do other things than just building more capacity. And I could not agree more. The existing LLMs have already swallowed all that was ever written. And probably more. And they still remain language models, increasingly more sophisticated statistical engines, not having a clue what they are talking about. I'm not saying that LLMs are totally usel...

Last week I spent almost entirely in Arizona - attending the one of a kind COSM technology conference . It has been the only non-lighting related event for me in 2025. Very valuable, as it has a mixture of politics, philosophy, economy, and - of course - technology. Once Discovery Institute posts the conference videos, I will link and comment on some of them. I the meantime a number of assorted observations. I flew to Los Angeles on Lufthansa Airbus A380. There is a small sections of economy seats on the upper deck. The best seats ever. This plane is so unbelievably quiet. Like a Zeppelin blimp, only much faster. It is a pity they are being withdrawn. The in-flight internet sucked. Completely. I responded to the email with the purchase receipt and they acknowledged the problem, issuing a refund. Probably the simplest refund procedure ever. Then after some business meetings in LA, on the next day a short United flight from LAX to PHX. This one - on the other hand - was equipped with Sta...

I have very mixed feelings about the importance of good user experience (UX) in technology products. Very often ironizing on UX gurus' posts who think UX is super fundamental. Scott Jenson for example have cried rivers over control panels in microwave ovens. True, some of them are quite baroque, and myself - when it comes to microwave ovens - I do prefer just two mechanical dials: power and time. Same for air fryers: temperature and time. But really it does not take a super high IQ to use a microwave oven. A bit higher one is probably necessary to set up an alarm clock in a hotel: 99% of time they show wrong time and then it is unclear which time is the current time and which is the alarm time and if the alarm is on or off. Long gone are the times of travel clocks with the alarm hand and the pop-up "armed" button : nobody needed instructions to operate them. Nowadays even the iOS 26 alarm clock is misleading - if you have an alarm entry that is off and you click it and th...

Apple AirTags are really a great idea. They build on the popularity of iPhones. Periodically broadcasting tiny Bluetooth Low Energy beacons they last for a year or more on a coin cell. And any iPhone acts as a message relaying / geotagging repeater station. Physically being tiny buttons, they can be attached to almost anything - car keys, luggage, any item than can potentially be lost. What is even more exciting (but still not as ubiquitous as it could be), a Find My beacon is a very simple piece of software running on any Bluetooth chip. It can run together with other functions. One great example is the Insta360 GO 3S action camera. The one I actually lost on my way down from Mt Aorai . It was raining on the way down and the clay path quickly became quite slippery. I slid down on my butt a good couple of times (it has some quite steep sections). Upon reaching the shelter at 1400m I realized the zipper in my backpack's pocket was open and the contents (including the GO3S) was miss...

There is a very insightful podcast episode - the interview of Andrej Karpathy by Dwarkesh Patel . At 00:53:39 they mention for about 1 second the 2021 paper by Erik Hoel:  The overfitted brain: Dreams evolved to assist generalization . It is all just a hypothesis, but I must say I've fallen in love with this hypothesis. Not much proof exists yet, but Hoel is really on to something. Dreams are all undiscovered yet. Impossible to get into. Brain activity scans of course can do nothing, as it is like trying to figure out the logic of the software by scanning activity of the microprocessor transistors. Impossible. But surprisingly the recent advancements in AI may get us to better understand how brain works as we try to rebuild an intelligent system in silicon deep neural networks (DNNs). Erik puts forward the overfitted brain hypothesis (OBH): Notably, one of the most ubiquitous challenges DNNs face is a trade-off between generalization and memorization, wherein as they learn to ...

Very simple trick today, which you may actually find useful. I learned it a good while ago and admittedly have been using quite often. The default means of sharing things on a phone is taking a screenshot and reposting it. This is ugly as carries a lot of overhead and makes it (more) difficult to extract the actual information such as copy a portion of text. But is easy (press two buttons) and preserves the view / formatting. Selecting text on the other hand is still the most ugly part of touch screen UX. Wonder if we will have it ever solved. Anyway, back to the main thread. Sometimes someone asks you to send a well formatted document resulting from a web page (such as ecommerce order) or an email confirmation. Yes you can screenshot. But if the source document is longer than the screen, things become tricky. There are a number of screenshot applications (especially on Android) supporting long screenshots (a single screenshot consisting of multiple pages), but we are talking about pla...

Life has presented me with a unique opportunity to visit some remote islands on Southern Pacific. As I’m based in Europe, Southern Pacific is probably as far as I could go. New Zealand is probably the farthest, but not the most difficult to get to - usually within the range of two long haul flights. Having just finished US East Coast business activities - the very successful NYcontrolled show followed by the DALI America Summit  and a workshop with an important partner in NJ, I faced the dilemma of what to do over the next week ahead of yet another show - this time in Hong Kong . Of course I could simply get back to Poland and fly to China next weekend. Or hang around in the USA. Or… Or take a week long holidays on a multi-hop trip over the Pacific Ocean.  With the internet at hand planning such a trip these days is peanuts. The itaSofware Matrix is your friend. Ultimately I opted to stay a couple of days in Tahiti (French Polynesia) and the second half of the week in Fiji. ...