Editor’s note

Friends, if you enjoy this newsletter, please share it to your own friends! So we can all be friends together. I am enjoying the replies and the conversations that I have with some of you. Do write if you disagree, agree, or merely for praise or criticism. Or just to say hi! Please enjoy!

Unrelated-but-interesting

Robots are farming tomato fields in Japan. There are labor shortages there because of demographic pressures, so another notch on the belt for the idea that necessity is the mother of invention. But tomatoes are hard to harvest because you have to pick ripe ones and leave the rest on the vine. Instead of detecting a merely red tomato, the robot now uses statistical analysis to evaluate the “ease” of picking the fruit. Which improves its return to 81% success. Which tells us that sooner rather than later, if technology can replace human workers in agricultural contexts, the choice will be easy. For better or worse? — phys.org

How does science progress? We often perceive scientific progress as accumulative. People conduct research that trickles into knowledge. Humanity advances incrementally. Thomas Kuhn, in his 1962 book “The Structure of Scientific Revolutions” suggests a different model. One where dramatic shifts that reject the previous equilibrium force us to adopt a new paradigm. Here's how this goes, more or less: there is a reigning model, such as Ptolemy's geocentricism which said that every astral object rotated around the Earth. Copernicus comes along and proposes that, au contraire, it is the Earth that rotates around the Sun (heliocentrism). Copernicus still used the same tools as Ptolemy, the new paradigm rests on the old. Over time, anomalies pile up. Galileo says that there is an idea such as inertia, and objects stop because of friction, rejecting Aristotle. Then came Newton who unified the discoveries. Ptolemy's model was rejected and a new paradigm adopted. Note that this is still not the consensual view of scientific progress, hence this summary — wikipedia.org

What were the consequences of DEI for millennial white men? That is Diversity, equity, and inclusion. I may start with the following disclaimer that I do not feel represented by the article, but the message is about enforcing fairness in a fair way (repetition intended). Throughout the last century, the lives of women and people of color i.e. political minorities, were systematically made unfair by some kind of gatekeeping by white men. For Gen X and boomers, the DEI movement starting in 2014 may have felt like rebalancing. But for some millennial men, working in corporate or media jobs, being 30 in 2014 felt discriminatory. Some were refused jobs because they were white men. An interesting account, that sheds light on a story that may be boring and irrelevant for many, notwithstanding its truthfulness — compactmag.com

Can color in artworks reflect the economic reality of a country? Yes, apparently. A team of economists studied the color distribution of historical artworks in France, the UK, the Netherlands and such from the 17th to the 19th century, and they discovered that colors mapped to economic results e.g darker for recessions, and brighter for periods of growth. Correlation is not causation but isn't it interesting that the chromatic properties of visual art encode information about material conditions? — ssrn.com (PDF)

Why is ice slippery? One hypothesis was pressure: the weight of a step exerts pressure that melts the surface making it more water-like and thus more slippery. But it was disproved. Friction, mayhaps? Friction creates a water layer made by whatever is sliding on it. Also disproved, although still prevalent: ice can be slippery immediately, before any friction. But the hypotheses are right in the abstract sense as ice is slippery because of structural changes on the surface. We still do not know why for certain, even though amorphization, or shapelessness, is gaining traction. Lots of fun in the scientific world! — quantamagazine.org

AI & software

In late December, Anthropic and OpenAI released Opus 4.5 and GPT 5.2 which pushed the intelligence of LLMs into a new territory. Models are now powerful enough to automate coding fully, as they can write tests and make sure that the code that is generated passes them. So the prediction of the Big Labs was true: in 2026, AI will be able to engineer production-grade programs. My engineer friends are not writing code anymore, they are merely steering the agents. Many questions come downstream for society as a whole. Any task that is computer-based would now be possibly automated sooner than we think. This has not been registered by the general population because most people use less-intelligent models. The new ones are kept behind the $100/month subscriptions. But they will most certainly be made cheaper soon. If you truly want to see where AI is at, I encourage you to buy the “expensive” subscriptions for a month and make sure to use the models written as the smartest. You will discover capabilities that may surprise you. I am very curious to know what you think and I remain available if you need any help evaluating different models

Elon Musk says there is no need to save money because we are fast-approaching a post-scarcity world. Where we will play the piano, sing, and paint, while our robot slaves will produce economically viable work for us. The universal basic income will make us all whole. My intuition is that the opposite is true. So my advice is to save as much money as you can! If wages are not guaranteed because of machine intelligence, we cannot continue with the model of the consumerist society which rests on the idea that we have enough to buy goods and services we do not need. This is especially true for America where it is likely this transition will be most brutal.

Two diverging theories about the “AI bubble” are running through Twitter circles. One is Eric Jang's, in “As Rocks May Think”, which states that the velocity of progress is not priced in by markets. That this time, it is different. That in 24 months, AI capabilities will be unrecognizable. The most salient idea is that just like air conditioning lifted the Global South out of poverty by enabling bureaucratic systems (Singapore) and A/C now represents 10% of electric demand, data centers (currently 1%) will represent 10% of future demand, which justifies the investment of the Big Tech companies. A 10x return such as this would change the world.

The other theory is Steven Sinofsky's. He says that we have seen this movie before. PCs, smartphones, fiber optics. All these revolutions were always wildly optimistic on the timeline. This time, he argues, is not different. The transition will take an entire career. Wall Street concluding that “software is dead” because anyone can build their own, is nonsense.

I think, unfortunately, that the former is most likely to be correct. Simply because the demand for “machine intelligence” is infinite. The more supply we will have, the more demand we will have. This is Jevons's Paradox in action: increased efficiency leads to more demand, rather than less, because the cost is lowered, and we find more uses for it. Who would not want to have an army of AI agents doing all kinds of research, data-crunching, idea generation tasks 24/7? If doing that gives you an edge, you may do it. If the providers make this transition seamless, which it currently appears to be for software engineering, then the sky is the limit (or more prosaically, our ability to generate electricity) — evjang.com - twitter.com

Loose ends

A curation of free, high-quality philosophy lectures on YouTube. A truly extraordinary resource — substack.com

Magnus Carlsen against a novice opponent. Playing with increasingly unfair rules (opponent gets two moves per turn, opponent starts with 23 queens, Magnus starts with 23 pawns). This is a good example of the difference between the top 1% and the top 0.0000001% — youtube.com

CleanMyMac, DaisyDisk, and more packaged in a nice TUI named Mole — github.com

Download 1300 high-quality stills from the official Studio Ghibli website — ghibli.jp