Yesterday, we bought two Knoll Sapper monitor arms. Most arms on the market use gas springs to counterbalance the display weight. The Sapper separates vertical motion, which works against gravity, from horizontal motion. Vertically, it uses a screw on the main column and the arm is held by friction. Horizontal motion requires no special mechanism. Gravity exerts no force in that plane, and friction is enough to resist small disturbances.

It’s so simple and so well thought out that I’m surprised it’s not the most popular arm. Its look is the expression of its function and not a gram of material is wasted.

We went to National Print Museum. Seeing Univers in metal and Linotype/Monotypes machines was surreal.

Finished reading Stoner by John Williams and it will leave the long lasting mark on me.

There’s nothing particularly interesting about the plot or the main character. He’s just a man. He’s born, you see his life from a long distance, and then he dies.

I cannot say I admired or hated William Stoner. I don’t think you’re supposed to. Instead, I felt sad, melancholic and sometimes happy for this man. Having been able to feel that delicate mix of emotions for someone who never lived was a sobering experience. You start seeing yourself as another person who lives his life through bad and good moments. And all that makes a life. A good life.

When I look at old textbooks there’s something poetic about them. The fact that pages turned yellow and corners were damaged, but the ideas remained. And then you look at someone’s name written at the corner and you understand that there was another person from a different country bending over this same book. It has been useful in the deepest sense. It made people know something they didn’t know before. This is a remarkable thing for a small stack of paper to have done.

Perfect.

Run FSRS optimization for my Mochi cards. Using train_csv.py from fsrs-rs-python. The new weights are below.

[0.26720747351646423, 1.629839539527893, 2.9071414470672607, 10.455878257751465, 6.474127292633057, 0.7291164994239807, 3.0826733112335205, 0.06065294146537781, 1.7888823747634888, 0.3533310890197754, 0.7168195843696594, 1.4715564250946045, 0.03916575014591217, 0.24278202652931213, 1.603986144065857, 0.6014000177383423, 1.8729000091552734, 0.5113826990127563, 0.07216055691242218, 0.09571979939937592, 0.1193043515086174]

I’m using 95% target retention rate for my science deck and I haven’t found it as “hellish” as many people in the SRS community see it. I use 90% for all other decks, so maybe it counterbalances my workload. The algorithm is more cautious with this target — it shows me cards right at the top of the forgetting curve resulting in intervals that grow quite slowly.

I hope these new parameters will optimize the spacing. If not, I may drop the target rate by a few percentage points. The anecdotal evidence is 5% increase in retention can nearly double your review volume.

Expected Return and Probability for Irish Prize Bonds

Irish Prize Bonds are a government-backed savings product where bondholders have the chance to win tax-free prizes in regular draws. The bonds do not pay interest; instead, bondholders are entered into a lottery for cash prizes.

The entire fund is approximately €4.5 billion. With each bond costing €6.25, the total number of bonds is $N = 7.2 \cdot 10^8$. All these bonds participate in the lottery.

There are 8,761 winning bonds per draw, and the total prize pool is €734,000. This means the expected value of a single prize is:

$$E_p = \frac{734,000}{8,761} \approx 83.78 \text{ euro}$$

The probability of a single bond winning in one draw is:

$$P_d = \frac{8,761}{7.2 \cdot 10^8}$$

The probability for a single bond to win in one year (52 draws) is:

$$P_a = P_d \times 52 \approx 0.0006327$$

For an investment of €25,000, a person would own 4,000 bonds. The expected number of wins for 4,000 bonds in one year is:

$$E_w = 4,000 \times P_a \approx 2.53$$

Thus, the expected annual return is:

$$E_r = E_w \times E_p \approx 2.53 \times 83.78 \approx 211.96 \text{ euro}$$

And the return on investment (ROI) is:

$$R = \frac{E_r}{25,000} \approx 0.00848 \text{ or } 0.85\%$$

Historical data analysis

A discussion on Boards.ie provides an opportunity to compare theoretical conclusions to practical results. One user reported an average annual return of €235 over 15 years on a €25,000 investment.

Another user shared a 20+ year history with total winnings of €3,550 since 1991. This gives an average annual return of approximately €162.7. Given their total investment of €84,950, this represents an average return of approximately 0.19% per year.

Conclusion

While some articles and reports criticize prize bonds, none make the mathematical disadvantage immediately obvious. The Irish Times states:

Of course, you can improve the odds of winning by buying more bonds.

While true, this statement overlooks an important fact: ==the return on investment does not depend on the number of bonds owned==. Whether you own 1 bond or 4,000, the expected return remains approximately 0.85% per year. You will win more frequently with more bonds, but you will lock away more capital for the same rate of return.

Update

The Irish Prize Bond website actually provides "Prize Fund rate", which is currently 1%. This is the same thing as the expected return calculated above. If you put €100 into prize bonds, you can expect to get back €1 on average per year in winnings.

Received Purcell's Electricity and Magnetism. Much like The Feynman Lectures, this book is a product of the Sputnik era when textbooks were still written for serious students. This first edition published by McGraw-Hill (ISBN: 9780070048591) is lean and tight and simply well made; the latest version from Cambridge University Press feels a bit bloated by comparison.

There's something magical about the fact that you can buy access to such knowledge for the price of a meal.

SwissMicros makes beautiful calculators. Will definitely get one when DM42n is back in stock.

Found Croscore fonts, designed by Steven Matteson (Segoe, Open Sans). They are metrically compatible with Arial, Times New Roman, Courier New, which means that text doesn't jump when the web font loads.

I found a photo from the previous edition of The Feynman Lectures on Physics published by Pearson/Addison-Wesley. It looks better than the New Millennium Edition from Basic Books that I have. I love Verdana, but the move from Futura to Verdana is sad. Futura feels very "Space Age" and academic. Verdana feels "odd" and digitally generic. The figures also look warmer when they aren't as sharp as they are in the newer edition. The pages were thicker, too; the new edition’s pages are so thin that you can see the text from the next page.

The only, but very big reason to choose the newest edition is the huge number of errors they fixed.

Added infinite scroll to this blog using htmx. Works pretty well. My website doesn’t have a footer, so it’s a pretty nice use case. Writing feels easier.

I bought two new math textbooks: George F. Simmons’s Calculus with Analytical Geometry and his Differential Equations with Applications and Historical Notes. I’ve only started the calculus book, but I already love it. It complements Spivak’s austerity and rigor very well with a more practical approach.

The Holdovers, 2023. Watched it a second time and already it’s my favorite Christmas movie. The atmosphere reminds me of The Catcher in the Rye in the best possible way.

Today I deleted my Instagram and Facebook, and deactivated my Twitter account. I hadn’t actively used Instagram or Facebook for years, posting only occasionally. Instagram stopped prioritizing real friend connections ages ago, so it was easy. I loved reading a couple of people on Twitter, so not being able to read them again makes me sad.

From E.T. Jaynes Probability Theory:The Logic of Science

The writer has learned from much experience that this primary emphasis on the logic of the problem, rather than the mathematics, is necessary in the early stages. For modern students, the mathematics is the easy part; once a problem has been reduced to a definite mathematical exercise, most students can solve it effortlessly and extend it endlessly, without further help from any book or teacher. It is in the conceptual matters (how to make the initial connection between the real-world problem and the abstract mathematics) that they are perplexed and unsure how to proceed.

How true this is.

The Station Agent. What a great film.

There are no options left to buy a non-integrated bike that looks good. I would expect Standert to produce some, but they also switched to running everything through a headset. So I went ahead and bought a previous-generation Aethos. It’s the cheapest (9r) frame they produce. I don’t think there’s a big difference between them, aside from it weighing 100 g more. It comes with some aluminum wheels and Shimano Di2.

I’ve changed the stock wheels for Roval Alpinist CL II wheels. When you look at the modern wheels, they have even more gimmicks than modern bike frames. Most of them use proprietary components, so good luck finding freehubs, bearings, spokes, whatever in a few years.

Roval Alpinists are as standard as they can be. They use DT Swiss spokes. They use DT Swiss 350 hubs. These hubs are bulletproof and require no tools for service.

The wheels are feather-light. I’m not sure yet how that translates to ride quality — if it does at all — but they look great on the bike. The 33 mm rims give it a tidy, balanced look that fits the frame’s lines well.

P.S. You know what this bike reminds me of?

Specialized launched the Aethos in 2020 and it turned out to be a pretty unique bike.

While all other brands embraced gimmicks — integrated cables, aero tubes, custom seatposts, press-fit bottom brackets — that are either useless for mere mortals or create more problems than benefits — Aethos had none of them. It was a non-aero, light bike that had almost no custom parts with the goal of providing superior ride quality. It, obviously, also catered to people who loved fixing and tuning their bikes themselves.

Recently, Specialized launched Aethos 2, and I think it killed the original spirit of the bike. Now it looks and feels like any bike from any other brand. Not in a good way.