_hl_

Huh, very interesting- and counterintuitive that distance matters. If anyone here knows why I’d love to learn more!!

EDIT: maybe not so counterintuitive after all: if you scale everything up, you get a higher fidelity sensor and more signal to noise. Same reason telescopes want to be big: to collect more light especially from dim signals. Distance of the occulter then reduces perspective distortion that would eclipse the inner parts of the sensor more than the outer ones. Just my speculation though.

I understand that part of the objective is to demonstrate precise autonomous navigation and choreography, but for the "main" scientific objective re. observing the solar corona, why not just stick a big stick with a disk on the observing satellite?

You're not giving Apple $1M. You're merely giving them a letter from a bank saying "We're $TrustworthyBank and we trust creativeSlumber enough to be willing to give them a $1M loan whenever they want, according to such and such terms".

By giving that letter to Apple, you're just telling Apple "I could give you $1M but I'm not going to, instead take this letter that says that I could give you $1M if you wanted to."

No actual money ever changes hands.

It’s certainly not easy and apple is playing the malicious compliance game here. But it’s not impossible for a (even pre-product) startup to raise a $1M line of credit from a bank through some creative financing shenanigans without locking up too much working capital or otherwise.

I’ve always wanted this, amazing!! Thanks for making it.

Attention is, after all, not what you need.

I’ve always wanted to start building something much like this, but the sheer scale of the task put me off.

I tip my hat and then some to the fine folks rolling this stone up the mountain. And that’s coming from someone who is married to elisp.

@dang

> Building third party flakes takes forever since every flake uses its own version of nixpkgs.

The “official” (as in, community consensus) solution seems to be to bring your own nixpkgs, and manually override the nixpkgs of every dependency (and their transitive dendencies) and verify your build still works. Which just doesn’t seem right to me.

Flakes are great for building an ecosystem of composable, reproducable software. The more they succeed at that, the bigger the problem becomes. Surely it’s not too late to figure out a principled solution before stabilizing flakes?

Isn’t that what django’s html templating is designed for? It’s not interactive using js in the client, but it is rendered on the server.

FYI, there are lots of papers on parallel shortest path algorithms. If an approximate solution suffices, there’s also a lot of research available on that, often with some parameter that lets you trade more computation for a tighter approximation. It's not a problem that parallelises particularly well though, so not sure if you'll see good gains in practice.

If you can restrict the structure of your graphs (e.g. planar) then some very efficient methods exist.

> Now if only we can get a model in the range of 100bln parameters

Do we have enough data to train such a large model in a meaningful way?

While I mostly agree, I wouldn’t attribute buying middleboxes (or building shitty middleboxes) to incompetence. This is about doing the minimum viable effort for compliance, not about actual security, and everyone who bought or sold these boxes knew that.

You must have a larger phone than I do…

There's no need to round-trip through text, you "just" need to train an embedding space that captures both domains.

It's mostly businesses though, not people. If you have >1 employee you def. should have at least 250K in the bank.

With modern NLP models being extremely good, can't you bundle together all sites whose content embeddings are very close? Presumably there's an acceptable threshold somewhere that achieves a good tradeoff of false positives vs negatives?

It doesn't matter if the system can "automatically" implement an infinite loop. E.g. in excel it's fine if the user is involved by clicking a button repeatedly or something to drive the system forward. It's more about, is the description of this system rich enough to emulate the semantics of a turing machine?

But this is all for the enterprise grade hardware. There seems to be very little interest from AMD to extend their stack to consumer GPUs. From an outside perspective I really cannot understand their reasoning here.

That's because these are also just proxy metrics for "net value provided to society", and heavily distorted at that.

"Hire the best engineers?" doesn't read like bullshit to me. A lot of startuos start out with scrappy teams who only half know what they're doing in their domain. That's partly because the domain changes every couple of weeks.

When you scale, that changes. It's crucial that you hire competent people for whatever your startup ended up doing. Have a lot of frontend? Hire UX designers and senior frontend engineers with 10+ years experience. Doing advanced AI? Hire someone who's been in that space for 5-10+ years. Etc. They're worth every penny.

The logic of fake-it-till-you-make it doesn't scale, but many don't realize it early enough.

Wow, is there a tech blog where we can read more about the design and why it was made that way? Sounds absolutely fascinating!

If it's for clients, does it really matter whether you're spending $2 or $5 or $10 a month?

> The only way that happens is more ads displacing content, or appearing in formerly empty spots.

Or eeking out more dollars per ad through better targetting, or through diversifying the business and scaling new revenue streams, or ...

I think it's reasonable to expect Google to grow faster than general internet usage. Believing otherwise means you believe Google is in maintenance mode, incapable of improving their existing product or unlocking new revenue streams through innovation. Given the sheer size of Google, that's quite a bold belief to hold.

I usually write a script that idempotently spins these things up locally, and use the shellHook to call that script. That way whenever I cd into the dir the first time, direnv loads the nix-shell, which starts the db, message broker, etc. I also have a shutdown script that I run when I manually want to shut things down. Works quite well.

For more advanced use cases, such as k8s with tons of services, you're better off using something like tilt.dev.

This kind of comment is not useful. Enlighten us with your knowledge, or, if what you have to say is not better than silence, keep quiet.

A counterpoint is the typescript compiler, arguably one of the most complex pieces of dynamic language tooling out there. It's written in typescript itself, and reasonably fast given that it has to implement a hugely sophisticated type system.

This shows that it's entirely possible to write fast tooling in dynamic language if you know how to.

Ah yes, can't wait to run my offline-first app on a kubernetes-on-docker-on-linux-on-qemu-on-wasm stack in the browser to have closer alignment between my server and browser implementations.

JS has built-in syntax to do just that. It's in fact much less painful than making a deep copy.

> Getting rid leaded avgas

I was shocked to hear that leaded gasoline is still around, so I did a quick search. Turns out most (all?) piston-engine driven airplanes still run on leaded gas. Jet airliners don't, so I'm not sure if the overall impact is significant, but nonetheless shocking to hear that we're still spraying lead into the air we breathe.