Hacker News: leourbina

New comment by leourbina in "New statue in London, attributed to Banksy, of a suited man, blinded by a flag"

leourbina — Mon, 04 May 2026 01:32:50 +0000

And yet here here we all are taking about it. Art is about inciting a response, and he’s done it. Whether we think he’s a hack or not is irrelevant - he has the world’s attention.

Show HN: Idea File for LLM Cycling Coach

leourbina — Wed, 15 Apr 2026 03:00:07 +0000

This is heavily inspired by Andrej Karpathy's LLM Wiki, and could be used to create many other types of "Agentic Apps" or however you want to call them.

My specific implementation uses Claude Code, TrainingPeaks, Todoist and Apple health. So far it is better than anything else that I've tried, and I love that I can use it to plan my structured bike training, gym workouts, and meal plan.

Curious to see if others have similar idea files.

Comments URL: https://news.ycombinator.com/item?id=47774188

Points: 1

# Comments: 0

New comment by leourbina in "Ask HN: Is Claude Down for You?"

leourbina — Thu, 22 Jan 2026 22:21:05 +0000

Yes

New comment by leourbina in "Trump says Venezuela’s Maduro captured after strikes"

leourbina — Sun, 04 Jan 2026 12:17:29 +0000

And that’s where my hopes vanish. A dictator-wannabe that has a track record of not respecting the peaceful transition of power in _his own_ country and has run his entire campaign on hatred for immigrants and more recently Venezuelans in particular, is going to, all of a sudden, grow a conscience and do what’s right for the Venezuelan people? Let’s call it what it is: this is about oil. I don’t have to say it, because he already did in front of the cameras. He doesn’t care what happens to Venezuela or Venezuelans.

New comment by leourbina in "Trump says Venezuela’s Maduro captured after strikes"

leourbina — Sun, 04 Jan 2026 11:55:17 +0000

What we’ve been living through in the last 28 years is Chavismo and yet Chavez is not around. When Chavez died in 2013 we celebrated that he was gone, and what we got after was much, _much_ worse. Now Maduro is gone, and we can celebrate it too. That said, Trump has signaled that he’s not interested in removing Chavismo: he’s keeping Delcy, Diosdado, et al, as they continue to be the brokers of power as long as he gets access to oil. This is just Chavismo aligned to American interests. Time will tell whether this is better or worse.

PS: As an aside, since I was a child growing up in 90’s Venezuela, the overall political mentality of people was that things were so bad that they couldn’t get any worse - and yet they continued to worsen. A lesson that I’ve learned is that in politics things need to be intentionally built - there is no “rock bottom”, the fact that things have been horrible doesn’t mean that they can’t get even worse. Thus my hesitation with what’s going on. There are no guarantees that this isn’t going to be a deal with the devil that leaves us in an even worse state…

New comment by leourbina in "Trump says Venezuela’s Maduro captured after strikes"

leourbina — Sun, 04 Jan 2026 11:24:36 +0000

Venezuelan here. It’s not that simple: Maduro was an _absolutely_ horrible dictator and yes many Venezuelans (myself included, and likely many of the 8+ million that left) are overjoyed with him being ousted, we haven’t seen any change in over two decades. And yet, it is transparently clear that the Trump admin is here not to save Venezuela, or Venezuelans… it’s here to line its pockets and that of its shareholders.

There was a very evident omission during Trump’s press conference: Any mention of Edmundo Gonzalez Urrutia, the duly elected president-elect of Venezuela (who won with a super majority last July - backed by Maria Corina Machado). Instead, Trump bad mouthed Maria Corina saying that “she does not have the support or respect of the country to run it”. They ousted Maduro, but they kept his VP (Delcy Rodriguez - which along other things is in charge of running the torture centers for political prisoners) as “she will do anything we ask her”. Trump doesn’t care about democracy or regime change - these things take time and are a long, thorny road (this wouldn’t be the US’ first rodeo). Instead they’ve chosen to keep the regime obedient with the threat of force, and instead just come in and extract as many riches as humanly possible…

Dark times ahead for Venezuela and the Venezuelan people

New comment by leourbina in "Size of Life"

leourbina — Thu, 11 Dec 2025 14:07:22 +0000

Came to say exactly the same thing :)

New comment by leourbina in "Most RESTful APIs aren't really RESTful"

leourbina — Wed, 09 Jul 2025 08:25:34 +0000

This post follows the general, highly academic/dogmatic, tone that I’ve seen when certain folks talk about REST. Most of the article talks about what _not_ to do, and has very little details on how to actually do it.

The idea of having client/server decoupled via a REST api that is itself discoverable, and that allows independent deployment, seems like a great advantage.

However, the article lacks even the simplest example of an api done the “wrong” vs the “right” way. Say I have a TODO api, how do I make it so that it uses HATEOAS (also who’s coming up with these acronyms…smh)?

Overall the article comes across more as academic pontification on “what not to do” instead of actionable advice.

New comment by leourbina in "Ask HN: I'm an MIT senior and still unemployed – and so are most of my friends"

leourbina — Mon, 07 Apr 2025 17:08:21 +0000

I’m a MIT grad from ‘12. PM me (email is on my profile)

New comment by leourbina in "A few words about FiveThirtyEight"

leourbina — Thu, 06 Mar 2025 06:50:36 +0000

Anyone can say anything on YouTube. I think you’re confusing confirmation bias with “citing your sources”.

New comment by leourbina in "How bloom filters made SQLite 10x faster"

leourbina — Mon, 23 Dec 2024 14:37:13 +0000

UUIDs are very wasteful [1]. For most use cases you can replace them with much shorter strings and still have very low chances of collisions [2]

[1] https://henvic.dev/posts/uuid/

[2] https://alex7kom.github.io/nano-nanoid-cc/

New comment by leourbina in "Starbase: SQLite on the Edge"

leourbina — Mon, 07 Oct 2024 02:30:20 +0000

Just fyi, the person that replied wasn't just "someone" at Cloudflare. It was Kenton Varda (kentonv here). He's the creator of Cloudflare workers, he's an incredible engineer.

New comment by leourbina in "Codestral Mamba"

leourbina — Tue, 16 Jul 2024 16:31:55 +0000

If you can run this using ollama, then you should be able to use https://www.continue.dev/ with both IntelliJ and VSCode. Haven’t tried this model yet - but overall this plugin works well.

New comment by leourbina in "Ask HN: How are engineers leveraging AI in your org?"

leourbina — Tue, 25 Jun 2024 13:16:36 +0000

I've been using co-pilot as well as Claude (Sonnet 3-3.5) and ChatGPT (4-4o) since the beginning of the year with somewhat mixed success. My general take is that, at least so far, the improvement is very much marginal. These tools save me a few seconds by quickly surfacing information that otherwise would require me having to google and search through docs/stackoverflow/etc, or by writing code that's generally tedious (like tests).

Co-pilot is mostly useful to stay in the zone, allowing me to focus on a larger task and letting it get some of the details right (which it does ~90% of the time).

On the other hand, I use ChatGPT/Claude for more open ended tasks (e.g. "I got this error", "how do I configure this framework so that xxx") which previously I would have googled hoping to find a stackoverflow answer or a doc page somewhere. For this use case I'd say it's ~50% successful, but I often have to deal with hallucinations - some times just following up with "Are you sure?" does help, but it's hit/miss.

As I said a the beginning, mostly marginal improvement. It's definitely saved me time, but thus far nothing that I couldn't do myself by spending a little bit more time. Largely it is a nice to have, not a need to have.

New comment by leourbina in "Does ChatGPT make us dumber?"

leourbina — Fri, 21 Jun 2024 14:07:11 +0000

I’d take a slightly different perspective. I’d like to think that these tools are in some ways “humanizing” - we can offload things that we’re not particularly good at (like memorization tasks) and instead we can use our capacities to do things that (at least for now) we humans are uniquely capable of doing. As an example: Back in the 90’s people knew many phone numbers by heart, nowadays I don’t think people know more than a handful, if that. Does that mean that “phone contacts” are making us dumber? Or perhaps we can use the time/effort/capacity to better use.

New comment by leourbina in "Fly.io has GPUs now"

leourbina — Wed, 14 Feb 2024 02:25:00 +0000

Paperspace is a great way to go for this. You can start by just using their notebook product (similar to Collab), and you get to pick which type of machine/GPU it runs on. Once you have the code you want to run, you can rent machines on demand:

https://www.paperspace.com/notebooks

New comment by leourbina in "Kullback–Leibler divergence"

leourbina — Wed, 23 Aug 2023 06:11:32 +0000

It just needs to be clarified that KL divergence isn’t a proper mathematical norm, so it doesn’t behave the way we intuitively think a distance should. As mentioned, it doesn’t satisfy the triangle inequality, which is a basic property for any distance-like function.

In comparison, both of your examples are much closer to norms as they both satisfy the triangle inequality.

For reference, this is what I’m referring to when I say a “norm”:

https://en.m.wikipedia.org/wiki/Norm_(mathematics)

New comment by leourbina in "Kullback–Leibler divergence"

leourbina — Tue, 22 Aug 2023 13:07:58 +0000

One intuition is that KL-divergence represents a sort of “distance” between probability distributions. However, this isn’t quite right as it doesn’t satisfy some basic properties a real distance (a norm) would satisfy, including the fact that it isn’t symmetric: KL(Q, P) != KL(P,Q), and it does not satisfy the triangle inequality. Nonetheless, KL(P,Q) gives you a good idea of how “far” is P is from Q: in the context of encoding, if you wanted to come up with an ideal encoding of symbols coming from P, but you guessed Q as the distribution of these symbols, then KL(P, Q) is the extra number of bits you’d have to use. One nice property is that in the case that KL(P,Q) = 0, P and Q are equal (almost everywhere, which for most applications is irrelevant). This makes it useful in the ML context as you can minimize KL divergence and know that the resulting “guessed” distribution is getting closer to the data distribution you’re trying to guess using some parametrized function (an NN).

New comment by leourbina in "Making home Internet faster has little to do with “speed”"

leourbina — Tue, 18 Apr 2023 17:39:52 +0000

https://webcache.googleusercontent.com/search?q=cache:s7pKN1...

New comment by leourbina in "Beej's Guide to C Programming"

leourbina — Sun, 26 Feb 2023 21:58:35 +0000

Just wanted to say thank you. your content was crucial in getting me to understand network programming and was crucial for my masters thesis. Your content has always been crisp, to the point and accesible. Keep it up!