discodave

We desperately _want_ tests to prove our code is correct. Unfortunately, this is not possible. Instead, thinking of tests as a model is the way to make it an effective use of your time.

Pretty sure Nvidias dream is that OpenAI and Anthropic get into a hacking war using Nvidia cards.

The problem for them is that the leading provider of training and inference is actually AWS...

If they are mostly paid in NVDA stock and NVDA is also valued at 40x revenues or whatever insane multiple then it makes sense.

> Here have some of my monopoly money I can print and come join us at Nvidia!

AWS is too busy selling entire private AWS regions on "take or pay" agreements to Anthropic, OpenAI and Meta.

Lots of good points in there but when you say "swapped out data center" that sounds very un-cloudy and unreliable to me - workloads in AWS are spread across many buildings unless you try really hard or pay $$$$

Is there any scaffolding or harness runnning around the model on the cloud/server side? Seems like there's a lot of opportunity to make changes/improvements without making your customers call a different API or change the model parameter.

As an example, S3 team was able to migrate from eventually-consistent, to consistent without making any API changes, a complete re-architecture on the backend with 0 API changes.

CDOs and SPVs have two things in common:

They're acronyms and they are financial engineering.

Maybe this time it will be different?

Ed mostly avoids talking about LLMs and instead talks about money.

EC2 is more scalable than your wallet. What do you need Kubernetes for?

Amazon / AWS is very efficient at it. They are the largest cloud by a significant margin. Much larger than Hetzner.

Google and MS may be close behind.

Not sure about Meta but they are also renting from Amazon so...

Anthropic mostly rents from Amazon, Goog & MS.

EC2 may have higher margins than that.

But your point stands, ain't no way xAI competing in that game.

Nuclear weapons are a deterrent against somebody invading the US (or another NATO country) but that doesn't make conventional forces not a deterrent against other kinds of aggression. Many attacks have been made against the US and not resulted in nuclear retaliation, like 9-11.

India and Pakistan have nukes and have fought each other recently so your assertion that "has_nukes() == no_game_start()" is *false*. Nukes, however probably will deter India from doing the full-Putin into Pakistan.

Where you say AWS, you mean "a single AWS region"

But anyway, from your YCombinator blurb:

    "When you’re done editing data, it automatically flows back to S3 within a few minutes"

Does this mean Regatta trades consistency for cost (S3 and EBS and local storage are all CP systems these days)?