pedrosbmartins

Number 4 is particularly interesting. Before Hattusili, the Hittite empire was ruled by his nephew, Mursili III. He was deposed by his uncle and exiled, first to Syria then to Egypt. He posed a great threat to Hattusili's rule, so of course the treaty had to include an extradition clause.

> Like the classifiers we were building 15 years ago with random forrests and logistic regressions

I think no one mentioned here, but the obvious difference is that Jev can spit out decisions directly from natural language input. None of these ML models could do that, and other than a full-fledged LLM (which is optimized for conversation and agentic tasks) or some classical NLP models (which underperform compared to LLMs, AFAIK), there is nothing right now that rivals Jev-like models.

Of course it's not perfect, but seems like the right step forward for quick classification/decision tasks based on natural language.

"That is a lot to ask of one number. Here is how it fits into one. Built one decision at a time." it sure sounds like Claude.

Pretty interesting how DeepSeek V4 Flash 0731 has such disparate (and cheap!) results. I would never guess they come from the same model and prompt.

> What is the probability that you are sharing the same birthday with people around you?

> What if I told you that in a room with only 23 people there’s already a 50% chance for two of them to have matching birthdays?

I guess it's the subject shift from _you_ to _any two people from a group_ that creates the surprise in the birthday paradox. You definitely need way more than 23 randomly sampled people to get to a high probability that _you_ specifically share a birthday with one of them, and the result does not contradict that notion.

KSPS was my favourite surfing game growing up. What I dislike about it though is the infinite/perfect waves (always a truck-sized barrel opened, you can do tricks anywhere, almost no differentiation between waves) and the focus on impossible airs and tricks.

In my game the waves start, break and end, with different sessions and hollowness, so there's more wave reading involved. Also the focus is on being able to stay on the wave and generate speed, doing cutbacks, snaps, off the lips, etc.

I believe realistic fluid dynamics in the scale of a full breaking wave is unheard of for real time graphics, there's just no way to do it in 30-60FPS (I could be wrong tho!)

What I do instead is to transform a procedural plane mesh into wave-like geometry. For added realism, I base this transformation on bathymetry data (ocean floor height), so you can get left/right breaking waves, different breaking sessions, etc, just by defining different heightmaps.

I'm working on a (somewhat) realistic surfing game. Tired of arcade-style games, I decided to try my hand at something closer to the real sport, focusing on realistic breaking waves, speed generation and carving, rather than impossible air combos.

After one year of development, it's going better than I expected, so I'm considering building a demo to gather feedback and see if there's enough traction for working towards a Steam release.

Even if that's not the case though, it's been a blast learning about game dev in Unity/C#, as well as 3D modeling and animation in Blender!

I'm not sure what you mean by that. Neutral networks are pretty good statistical learning tools, and in this kind of application you'll need some stochastic learning, regardless of using a laptop or a supercomputer. It's not like they used an LLM to predict the simulation steps. If you read the paper, they seem to use a simple fully-connected 5-layer neural network architecture, which is a completely different beast from, say, trillion parameters transformers used for LLMs.

First time I came across integer programming (and mathematical programming generally) was when studying hydroelectric power generation planning, for a masters I ended up not pursuing. Then, when selecting a masters in CS, I ended up working with an advisor who used mixed-integer programming applied to classic machine learning models (mainly optimal decision trees). A fascinating and widely applicable method, indeed!

what kind of sick, twisted, inhuman comment is that?

Sebastian Lague is one of those content creators that show what a delight the internet can be when the content is high-quality. I particularly enjoy how he presents his implementation thought-process, eventual dead-ends, and possible future improvements. If anyone here has suggestions for similar content, I'd be more than interested!

cool project! E-mail seems like a good channel for small chunks of language-learning content + reminders.

If I may ask you, how do you plan on building vocabulary from these e-mails? Do you use anki or some other method?