dilippkumar

Hobbyist harness builder here. Thank you for the article.

What would be helpful would be a cheap API call that would give me the energy cost estimates:

> Given some prefix-cache uuid, what is the cost (in mW of GPU energy) of adding X tokens that will branch off from what's been cached at offset Y tokens.

Suppose at some turn n, I read a file and pull it into the context. Now at turn n+m, I discover that the file has mutated on disk. The most energy efficient option might be any of these:

1. Replace the content of the file in the cache

2. Append the new file to the end of the cache

3. Add a thing to the context saying the file has been updated.

4. Compact memory then do option #3

As a harness builder, it's pretty hard for me to pick the most energy efficient strategy - I assume every model has a different cost for prefill vs decode and I also expect that a provider (an inference cloud) might have slightly different implementations that would impact how expensive a prefix-cache miss is.

Ofcourse, I also need to target aggregators like openRouter because as a solo hobbyist harness builder, it's really hard for me to track every cloud's unique API. Then there's the fact that I'm optimizing for lowest billed cost, not lowest energy consumption.

I'm waiting for the industry to standardize on a good way to express these in an API so that I can start doing these tradeoffs at runtime in my harness.

I’m sorry you went through this.

But I am interested in the monero aspect here.

Should I treat this as some datapoint on monero’s security having held up well so far?

Thanks for the pointer.

I went down the rabbithole, and as far as I can tell, you have to axiomatically assume infinities are real in order to prove Goodstein’s theorem.

I challenge the existence of ordinal numbers in the first place. I’m calling into question the axioms that conjure up these ordinal numbers out of (what I consider sketchy) logic.

But it was a really fun rabbithole to get into, and I do appreciate the elegance of the Goodstein’s theorem proof. It was a little mind bending.

I haven't studied math beyond what was needed for my engineering courses.

However, I also am starting to believe that infinity doesn't exist.

Or more specifically, I want to argue that infinity is not a number, it is a process. When you say {1, 2, 3, ... } the "..." represents a process of extending the set without a halting condition.

There is no infinity at the end of a number line. There is a process that says how to extend that number line ever further.

There is no infinity'th prime number. There is a process by which you can show that a bigger primer number must always exist.

> There are high speed police chases (100mph+) in Los Angeles — no exaggeration — on an almost daily basis.

How is anyone driving at that speeds in LA traffic?

Long time OpenBSD fan. Used it as my daily driver for years before standardizing all computers at home to macOS. I still think about going back to openBSD one day, but it's no longer very practical as a daily driver.

I want to use OpenBSD for the next project I'm building. However, I can't wrap my head around the old way of doing deployments (before containers). People who've built production grade systems with OpenBSD:

1. How do you deploy software? 2. How do you manage fleets of servers? 3. How do you spin up/turn down servers from cloud providers? (I only know of Vultr who provided an OpenBSD option out of the box).

I ended up having a work laptop that looks exactly like my personal laptop. After making an embarrassing mistake once, I picked up some stickers and stuck them on my office laptop.

Now, my work laptop looks clearly distinct, not just from my own personal laptop, but also from all the identical laptops other people at work bring to meeting rooms.

Yay! A tiny minuscule bit of my code is riding on these. While I no longer work there, I am absolutely thrilled at this milestone

1. Congratulations everyone! Yay!

2. I absolutely recommend Zoox as a great place to work. Believes me, I’ve sampled many jobs, Zoox is up there with Google in terms of what the experience feels like in my experience.

3. Yay again!

sigh That nightmare level

So pornhub needs to see how many terabytes of content they host and use AI to generate 2x more terabytes of cat pictures and add them to a compliance tab on their home page now?

Seems annoying but not impossible to do.

Edit: I am happy to build a cat pic to porn ratio audit company if anyone is interested. I want to participate in the funniest regulatory process this will create

This is really cool.

But I want to talk about how much free headspace Graham Hancock has managed to acquire. I thought about him the moment I saw this title. I'm sure some of you came to the comments looking for his name too.

> There are already open-weight models that match OpenAI's best models (at the time of their release)

Did I miss something? Do we now have o1-pro level performance in an open source model?

I use OpenBSD. I love it, but I recommend reading warnings on the label.

I would say FreeBSD is somewhat like Ubuntu is to Linux - easy to get setup, works for more people.

There isn’t anything like OpenBSD in the Linux world - where the primary focus is system correctness, even at the cost of user convenience at times.