santiagobasulto

Do we really need a whole book for this?

I think this is just a tongue in cheek description, implying that "nobody needs this page and the `other one` makes everybody happy".

Cool project, I had to implement the same for my experimental AI Agent: https://github.com/quantized-ai/luca-py

LiteLLM and LangChain are AWFUL pieces of software and should be avoided at ALL costs.

Btw, you should update httpx to httpx2, and I think it's not much effort to remove openai's SDK compatibility.

I'd love to have a provider-agnostic LLM router (almost) dependency free (aside from httpx2).

of course I meant that a human made object traveling at those speeds was theoretical, not time dilation itself. Never thought I'd have to clarify that :)

Yes. Time dilation (with respect of the stationary observed) is calculated using the Lorentz factor.

At 99.999% of C, the Lorentz factor is ~223.6.

It grows pretty quickly as you add more 9s to the fraction. Every two additional 9s multiply the Lorentz factor by ~10.

So at 0.99999999999 c, it'd be ~223,607x.

2.5M years / 223,607 is: ~11 years

Of course this is all highly theoretical.

https://en.wikipedia.org/wiki/Lorentz_factor

A lot of hate for Apple as usual. I love my AirPods. They just works, that's what I need.

Some time ago I had the following shower thought: "the Speed of Light is pretty slow"

How i got there:

The closest major galaxy to the Milky Way is Andromeda, and is 2.5 million! light-years away. And this is the CLOSEST galaxy, the universe is extremely big.

Of course that as you get closer to C, the traveling object will experience time dilation (relative to observer), so the time passed will be less. At 99.999% C, the traveler would take ~11,000 years to arrive to Andromeda.

So again, even at 99.999% C, 11K years seems like a LONG time to reach even the closest galaxy.

My reasoning was: the speed of light is pretty damn slow.

But then I realized: no, it's not the speed of light that is slow, is my frame of reference.

For us humans, 11,000 years seems like A LONG time, but for the universe is not that long.

The universe's age is estimated to be 13.8 billion years. 11,000 years is 0.0000007971 of the age of the universe.

An average human lives 70 years, 0.0000007971 of that lifespan is approximately ~0.4 hours, or 29 minutes, so it's not that bad.

So yeah, frame of reference matters.

I was going to reply with "shit americans say" until I realize you're based in Paris. Isn't 90k a very good salary even for an expensive city like Paris? If it's not, MAN i'm out of touch with reality. In Italy 90K is great even in Milan.

Can you elaborate? Why is Python asyncio bad?

When python started incorporating asyncio my initial reaction was fairly negative. I was behaving like a grumpy graybeard programmer ("kids these days should learn how to use concurrency instead of working on a new runtime altogether!" [0]).

But after using it for a production project I have to say I'm deeply satisfied and surprised by the maturity, quality of APIs and performance in general.

I was proven wrong and I'm happy about that :)

[0] Old man yell at clouds type of meme

563 upvotes for a 2 paragraph post of someone complaining about AI images in someone else's blogs?

This is "Bicycle face" level of hatred (or fear) against new technology.

I love Google's and duck duck go AI suggestions. Am I crazy? I feel like 90% of the time they answer my question right there.

It will definitively be a problem in the long term for monetization of the authors. But for me as a user, it's the best.

LLMs are aligned to be cautious. And “good” financial advice is extremely simple. A conservative approach gets you there 80% of the time. Is when people want to get too smart (or they’re bordes) that money is lost (gambling mostly: literally or with bad investments). So yeah, I’d assume AI is good at this.

I'll take advantage of your domain knowledge :) Where's the best place to find local artists that produce them?

Didn't expect to see a MESHUGGAH reference in HN today. Not bad.

I loved it, i hated it. I'm so bad at this.

I think summarizing everybody's feedback the simplest solution is: "Game difficulty".

- Standard: what you have today

- Relaxed: 1 minute per word?

- Practice Mode: no timer whatsoever

And "Practice Mode" is a completely different mode that lets you skip questions, and instead of "you win / you lose" which is today's behavior, you end up with a score (14/18).

Forget AI agents, this DSL is better even for humans. Cool project!

Yours looks much nicer!

David's talks at PyCon are all amazing. My favorite is this one: https://www.youtube.com/watch?v=MCs5OvhV9S4

Funny enough, a few months ago we decided to start a new project and we chose async Python. I had no clue how'd work so I decided to read the PEP and learn a bit more. Big was my surprise when I realized I "knew" all the fundamentals just by understanding how coroutines and generators worked low level, thanks to David's talks.

EDIT: His talks about the GIL are also super informative!

I think we're missing the important point here.

The problem is not that the ID wallets require Google and Apple. The problem is that we're getting eaten alive by this Big Brother called EU (lead by the UK initiatives) that is starting an unprecedented control over the population.

These ID wallets should be all optional, there should NOT be any age verifications.

I remember ~10 years ago when Europe was laughing at China's face detection systems to track citizens.

We're becoming much worse than that now.

This is all super interesting, but I can't decide what's the MOST interesting thing:

1. The whole Pollen case (I didn't know about)

2. That Google can be tricked so easily?

3. The whole "industry" that seems to be in place to clean the image of some scumbags in the internet (this whole Ellie Piee from Bouvet Island)

I think the most worrying part is Google's fragility to hurt itself.

The fact that you're not naming any cost related variable reflects your own current personal situation (privileged, I'd assume). But this is NOT the situation for most people in the world.

Don't get me wrong, I'm also privileged. I can pay for pretty much any type of medical intervention that I'd need. So my variables are usually "comfort", "speed", "convenience", etc. But I know that this is NOT the most common scenario for everybody.

That is true, but I have to be honest and say that I didn’t make the connection until I saw Bellard’s project for the first time, and I said: “ah! That actually makes A LOT of sense”

Bellard has a very interesting project that is `ts_zip`, a compression algorithm powered by LLMs. It's just an "experiment" and should never be used in production, but very smart.

The description on his website is amusing: "The ts_zip utility can compress (and hopefully decompress) text files using a Large Language Model"

https://bellard.org/ts_zip/

> Senior technical Anthropic staff

doesn't seem to be a political meeting, but rather a technical one. Maybe they'll review the jailbreaks and demonstrate that can be replicated in any model out there?

Whatsapp is working fine for me (in Europe) btw

Not at all, I had the same feeling as yours the first time I read it. I think the key is that the "encoder" they're using is just a linear projection, which is probably pretty fast and memory efficient. A single matmul vs a ViT encoder is probably a huge win.

I don't think it's the same. It's a similar concept, but Gemma is using just a linear projection, which I assume is a lot faster. The developer guide has more details: https://developers.googleblog.com/gemma-4-12b-the-developer-...

    Vision embedder (35M parameters): Replaces the 27 vision transformer layers of the other medium-sized Gemma 4 models. Raw 48x48 pixel patches are projected to the LLM hidden dimension with a single matmul. A factorized coordinate lookup (X and Y matrices) attaches spatial location information directly to the input

the "single matmul" is the key here, I haven't tried it, but it's probably pretty fast and memory efficient.

I mean, "Cannes" the city right? lol

I don't own Apple Vision Pro, but I tested them two times already and I'm amazed by the technology. I honestly think it's the future. We're probably WAY too early yet, and the Vision Pro might fail badly, but it is indeed the future.

I did a special test session in Japan for "productivity" (the guys at the Apple Store were very friendly and agreed to let me install VSCode and Ghostty on the testing laptop. I cloned an open source repository and spent ~20 minutes just coding.

It was FANTASTIC. The Apple Store was full and I could still "black out" the noise and completely immersed myself in the experience.

I'm seriously considering buying a pair now, but I'm just concerned about the under-investment in the sector.

Regardless, I honestly think it's the future, maybe in 10/20 years, but it'll be the norm.