@markus@tabletop.social (@MarkusRShepherd) header
@markus@tabletop.social (@MarkusRShepherd) profile picture

https://nitter.gq/t.co/Tb3HVzOer6 Human Approximation of Artificial Intelligence. He/him. Founder of @recommend_games

Helsinki 🇫🇮
Joined January 2011
Photo by @markus@tabletop.social (@MarkusRShepherd)Photo by @markus@tabletop.social (@MarkusRShepherd)Photo by @markus@tabletop.social (@MarkusRShepherd)Photo by @markus@tabletop.social (@MarkusRShepherd)Photo by @markus@tabletop.social (@MarkusRShepherd)Photo by @markus@tabletop.social (@MarkusRShepherd)Photo by @markus@tabletop.social (@MarkusRShepherd)Photo by @markus@tabletop.social (@MarkusRShepherd)Photo by @markus@tabletop.social (@MarkusRShepherd)Photo by @markus@tabletop.social (@MarkusRShepherd)
I revisited my 15-year-old master's thesis to learn how AI-for-maths works: proof plans, separate agents and AI reviewers on ordinary subscriptions. A nostalgic return to pure maths, too. riemannhypothesis.blog/2026/…
13
OK everyone, it's time to talk about the real part of every non-trivial zero of the Riemann zeta function (1/2)
80
502
60
5,234
208,781
I don’t really know what to say. A proof of the Riemann Hypothesis showed up tonight as 17 million lines of Lean, written with Claude. I’m stunned, giddy, a little lost. Three days ago I wrote about AI re-proving my master’s thesis: riemannhypothesis.blog/2026/…
1
32
Ich bin mir sicher, ein niedersächsischer Ministerpräsident names David McAllister hat ihn genauso gestört...
Mal abgesehen von den aberwitzigen politischen Vorstellungen der Linken in Berlin und ihrer Abhängigkeit von kriminellen Clans und islamistischen Terrororganisationen habe ich einfach ein Problem damit, dass eine Regierende Bürgermeisterin der deutschen Hauptstadt Berlin Elif Eralp heißt.
1
32
I've been rewriting the recommendation algorithm behind #RecommendGames and I need some human validation. Could I ask you to visit the two links in the thread, enter your #BoardGameGeek username and leave a like on the recommendation you like better? (Or like both…) Cheers! 🥰
1
1
1
39
Here are my predictions for #SpieldesJahres and #Kennerspiel tomorrow: blog.recommend.games/posts/s… As far as I can tell, both races are the closest I've ever observed, so the ceremony tomorrow will be extra exciting 🤩 @spieldes #SdJ2026
40
The simulated bracket at the moment according to my model (which has been shite so far). I hate that #FIFA inflates the tournament and then creates clashes like 🇫🇷 France vs 🇩🇪 Germany in the round of fucking 16. #WorldCup #2026WC #FIFAisamafia
77
#FIFA is a mafia and I love #football anyway, so I did the only reasonable thing: simulated the whole #WorldCup 10 million times. 🇪🇸 Spain are your filthy 35% favourites — and the betting market thinks I've lost the plot. ⚽️ blog.recommend.games/posts/w… #WorldCup2026 #MonteCarlo
26
#Spiel and #Kennerspiel des Jahres nominations and recommendations will be announced on Tuesday – high time for some speculations! 🤓 I've just released my annual predictions: blog.recommend.games/posts/s… What are your favourites? @SpieldesJahres #SpielDesJahres #SdJ #SdJ26 #SdJ2026
1
93
Those who have been around for the Python 2 -> 3 transition should understand why messing with established patterns is dangerous… 🙈
It's 2026. Why is Word's icon for "Save" still an abstracted floppy disk?
1
20
The computations for this plot took almost 14 days. I hope you enjoy it. 🤓
1
30
New post: Elo as a Skill-O-Meter 🧮 In part 3 of my Elo series I stop ranking players and start measuring how much skill a game actually shows. Toy universes, p-deterministic worlds, and snooker vs tennis. blog.recommend.games/posts/e…
2
1
122
New post: Elo as a Skill-O-Meter 🧮 In part 3 of my Elo series I stop ranking players and start measuring how much skill a game actually shows. Toy universes, p-deterministic worlds, and snooker vs tennis. blog.recommend.games/posts/e…
2
1
122
The answer is Alice. I will give the exact probabilities and much more after work, but for now, I will give a rigorous argument for why the answer is Alice.
Charlie puts 26 presents in 100 boxes, labeled 1 to 100. Each second, Alice and Bob look in one box. Alice opens them in order (1,2,3,…), while Bob opens the odds first, then the evens (1,3,5,…,2,4,6,…). Who is more likely to see all 26 presents first?
20%Alice
16%Bob
48%Equally likely
16%See results/cowardice
5,815 votes • Final results
12
19
4
404
61,328
The computations for this plot took almost 14 days. I hope you enjoy it. 🤓
1
30
From time to time I like to video myself solving maths problems in real time. Since the IMO has just happened, I plan to try a few of this year's questions. Here I get through Question 1. More to come at some point. youtube.com/watch?v=1G1nySyV…
10
55
13
431
119,142