nchagnet

That is the same impression I got reading this. I would enthusiastically use any OS-first method if it means one less brittle dependency, but the current usage of environment variable is rather different than what is described in the post.

Where .env files shine is that: - they act as a declaration of expected environment variables, - they are project-scoped, which is one big issue with using "/etc" for example

I personally like to set a .env.example for my collaborators to know what's expected, and I use a .envrc with direnv. And to make it more secure, I always have .envrc in my global gitignore so I can't just forget it.

The drawback is that for any non-interactive run (debugger) I have to manually add each variable each time.

Considering I'm not Dutch, you may feel reassured there is no superiority feeling at play here.

I agree with another commenter that while flat, the Netherlands have their own hurdles (biking with a strong headwind on the banks of the IJ is not easy, even if flat), and I definitely agree that their city design is what makes this unique.

I lived in various parts of France growing up, and I can assure you there are flat cities there, yet biking in them felt very risky at best.

Oh I agree. When I lived in Lyon, who is also quite bike-friendly, it was a lot more challenging than Amsterdam.

But with electric bikes becoming more affordable, hopefully the gap can eventually close.

That's a really good point, I hope at the very least it enables a "car -> public transport -> bikes" flow. So even if these people were taking the metro, all that extra metro space can accomodate car-owners who wish to switch.

I live in the Netherlands where the weather is arguably tougher than in Paris (rain, cold and wind for large portion of the year) yet everyone bikes year in year out.

And not just young active people, it's a habit found across all age groups, parents bike their children to school (or with them if old enough, etc.)

All that to say I wouldn't worry too much about the feasibility issue, it's really more of a mindset to adopt, and it's happening more and more in France.

They're likely saying that at equal usage, the user with mixed usage will cost less because the cost of B is lower than A.

The same reason people generally don't regret having kids even though the commitment and overall change of your life are much greater than what you described for dogs.

It's definitely quite impressive. You can still really tell that each shot is jarringly different from the previous one. Already in the building it felt like the "background" always changes. It was even more jarring in the car section as the terrain just randomly changes in weird ways. But I guess it makes sense.

Also, I don't know if it was supposed to be a covert ad for the cybertruck, but it reminded how ugly this car is.

This looks absolutely wonderful. Any plans for an Android version?

You called your product in the LLM space mistral.rs but there is no mention on your page of the name relation to Mistral AI? That's kinda confusing.

This looks wonderful, I've dreamt of something like this back when I was in academia!

You mention some advanced stuff like abstract index notation which is usually the hardest part to get right with things like simpy and even mathematica, how does your package handle simplifications of complex tensor expressions (GR, hydrodynamics, etc.)?

Just fyi that Skapa is already the IKEA design system used to harmonize their design across platforms and websites:

https://m.youtube.com/watch?v=UPTLA2aNASA&t=2s&pp=2AECkAIB

> While LLMs have massive valuations right now. Pretty sure the public has spoken when it comes to the differences you fail to illustrate actually mattering.

Are you seriously using market valuation as an indicator of worth?

I can see why, when I work/focus, I like to use my computer instead of my phone because that's where my headphones are connected (easy switch for meetings, etc.) and I generally like to be nice to my phone battery.

Why would you lose your job for posting about and publishing your personal projects?

HN has become my main feed for non-news stories and I quite like it, particularly because it has a higher density of small web tech articles and those are my favourite. It doesn't have to be highest quality or most innovative, but reading someone's articulated opinion on their blog is what I go on internet for.

I also generally enjoy the comment section because people here tend to at least argue intelligently, even if I don't always agree with the viewpoints. I will say I find the dissonance between the "hacker" and the "tech startup" ethe to be quite amusing. I may work for a large company in their 'tech' area, but I can never feel fully at home among the more entrepreneurial-minded people. I really struggle to cheer for tech companies' stocks like they are football teams, but to each their own.

Thank you HN for being a unique place on the Internet, it's truly refreshing.

We're using ducklake with data storage on Google cloud storage and the catalog inside a postgres database and it's a breeze! It may not be the most mature product, but it's definitely a good setup for small to medium applications which still require a data lake.

Sounds like a fun initiative!

Mine is https://nchagnet.pages.dev

I really like the pattern presented in the article. I find myself guilty of designing errors which are useful to me, but maybe not to my user (which tbh in my area is always a bit of a nebulous entity). I really like the idea of separating those two intents, and to make explicit the possible action.

This is really interesting, I'm always finding weird bugs in my joins once they grow a certain size and this has helped me understand why. Btw you still have a todo comment in section "Why you should only use ID equality"

I generally agree with you on the point that a "good Pokémon team" can be better encapsulated by other attributes, including those you mentioned. I would disagree on assumptions being very wrong, because I am not assuming that the objective and constraints chosen are ideal or even good enough, I am choosing them simple for illustration purposes.

I actually found it interesting that in spite of what is a clearly overly simple model, the non-legendary non-multi-starter you eventually get is quite a good one, in my opinion better than what the naive constraints would lead me to think.

Also, keep in mind that I'm not talking about competitive matches here, just mainline gaming. For that end, types are usually all you need, and in that area the main thing I would do is generalize type constraints to not be just defensive but also ensure each resistant Pokémon has a good enough attack against that type.

In my opinion, abilities, nature, objects are: 1. Too complex for such models (MIPs are still exponential-time) 2. Overkill strategy when all you wanna do is beat the league

But that last part is just my opinion.

That's a great idea!

Thank you for your suggestion, I agree with you (and another commenter) that base stat is not that useful, and availability is actually what I would prioritise on in a next iteration. I tried to keep it simple here, mostly because it was interesting enough as an analysis. But if I were to redo this to get _the best_ team in a generation, I'd definitely go with what you suggested!

I do comment on that in the article, I think it's a nice example of how your model can only know what you tell it (the one I used in the article doesn't know about abilities).

Absolutely! In general I would expect a better model to incorporate a lot of weighed terms in the objective to choose less "extreme" solutions, but here I was mostly interested in illustrating the method.

Oh that's really cool, I didn't know about this! I just linked to the wikimedia-hosted illustration, but that's a good perk too.

I was planning in a future sequel/update to do this but with "better" constraints like only including Pokémon available in a game, etc... Maybe even separate it into early/mid/late-game availability since most optimal Pokémon are late-game anyway.

Thank you, you're right! For some reason I always forget mew comes after mewtwo in the pokedex...

Good catch! Especially since I ended up drawing y - x = C but didn't update the legend. I updated it!

"QM works in the Hamiltonian picture and I recall from my undergrad days that you get there from a Legendre transformation on the Lagrangian (or something to that effect) so I'm trying to understand the justification of that approach before moving up the conceptual ladder."

That is only one approach to QM. As you rightly point out, Hamiltonian and Lagrangian approaches are always two sides of the same coin: one is only the Legendre transformation of the other and so they describe the same physics.

So to that end there is a neat QM Lagrangian representation: the path integral formulation. You can apply it to basic QM as well as to QFT or even QFT in curved spacetimes and string theory.

So if your goal is to "trace your way from simple postulates", that is a good way: assume your system can be described by an action, and go from there. It works in pretty much every scenario. In most research I've been involved, you always end up constructing generic actions whose coefficients eventually determine the behaviour of the theory.

And to get back to what I assume is your real conundrum (why do we extremize something to begin with), I just don't think there's any true answer as to why nature behaves this way.

What we can answer is: what is the action, what does it represent, what does it mean to extremize it? The short answer to this (provided by the path integral formulation I mentioned earlier) is that the action is essentially controlling a probability distribution of paths in a given geometry. In quantum mechanics, we interpret this distribution with that of actual particles. When you extremize the distribution, you essentially find the most likely trajectory, and if your distribution is peaked enough around that trajectory, then you can take this path as representative of your system when fluctuations are ignored.

So in the classical limit of QM, that trajectory is all that's left (and that would be the classical mechanics trajectory).

Interestingly a similar interpretation exists in statistical physics. If you "complexify" your time dimension, your action is again on a Euclidean (instead of lorentzian) spacetime and the time direction behaves like a circle whose radius sets a scale akin to a temperature. This might sound a bit complex but where I'm going with this is that once again, you can think of this euclidean path integral as a distribution of paths over fluctuations (this time thermal), and the extremum is this time the system's behaviour when at equilibrium.