winddude

I've read / hear the cloud providers like AWS started extending amortization periods in 2020ish

agreed, cloudflare says 20% of all websties, which makes sense, digging into it the "20% of all web traffic" seems to have been like a game of telephone in media, social media, and/or AI.

I can't believe they only have 295,000 paying customer, that puts me in a small minority. lol

https://w3techs.com/technologies/overview/proxy, they are tiny compared to CF, their revenue is high because they focus on large enterprise clients.

They don't, akamai has had several outages as well jsut no one notices. Akamai is way way smaller than cloudflare, 20% of internet traffic passes through CF networks, not sure it's even measurable on Akamai.

8T is the high-end of the McKinsey estimate that is 4-8T, by 20230. That includes non-AI data-centre IT, AI data-centre, and power infrastructure build out, also including real estate for data centres.

Not all of it would be debt. Google, Meta, Microsoft and AWS have massive profit to fund their build outs. Power infrastructure will be funded by govts and tax dollars.

"But we bought too many GPUs! We spent billions on infrastructure! They have to be put to work!"

right... none of them are saying that. They could probably use more GPUs considering the price of GPUs and memory are skyrocketing and the supply chain can't keep up. It's about experimentation, they need real users and real feedback to know how to improve the current generation of models, and figure out how to monetise them.

why use a name, although spelled differently onyx vs onnx, that's already used and known in the ML/AI community?

linode was better and had cheaper pricing before being bought by akamai

but like a normal dev no unit test.

As a self taught engineer who hasn't read the article or done any research I can confirm.

It's a bit old, but I bench marked a number of the web extraction tools years ago, https://github.com/Nootka-io/wee-benchmarking-tool, resiliparse-plain was my clear winner at the time.

I'm being pydantic, "So unlike the thermostat in your house, which doesn’t have to contend with any other control systems, all of the governors of the mind have to fight with each other constantly.", but what about automated blinds, self tinting windows, automatic skylights, humidistats, and humans open and closing windows.

It's also important to note that other control systems in the body that affect control systems in the mind, eg. endocrine.

as a car guy and software engineer I just want to say car's need way less software, way more separation of concerns, more standardisation and more open platforms, but most of the money is made on service, so the manufactures are incentivized to make closed systems.

difference to something like duckdb?

this is for type 2 diabetes, a VERY IMPORT distinction. So it probably just effects eating habits. <https://www.theguardian.com/society/2024/oct/31/less-sugar-i...

unlikely, over the summer they announced that they were going to be more opensource, <https://www.elastic.co/blog/elasticsearch-is-open-source-aga...>

I had that one happen as well, after launching a project. I could even post in a messages to friends.

probably why OpenAI wants to build there own 5gw nuclear plant.

'''"None of the Service’s technologies has been trained on a comprehensive and current corpus of federal and state laws, regulations, and judicial decisions or on the application of those laws to fact patterns," the FTC found'''

Wow!! That seems so simple, and literally a few weeks to do in today's ecosystem, now thoroughly testing make take a little more time, but wow, I wonder if it was evening attempting to do RAG.

the vision models can also do tool calling according to the docs, but with text-only inputs, maybe that's what you meant ~ <https://www.llama.com/docs/model-cards-and-prompt-formats/ll...>

Do you want to finish it or do you want to start something else? lol

but yea, I agree, which to do. My random take would be do the opposite of what you normally do... my scientific expertise on this, I've started ~100 projects in the last year, and mostly "finished" 3...

or shopping carts in the store, need to cut down on the number of sorries.

Interesting idea, I'm also in Canada, I briefly looked at extracting grocery store price info, don't forget that prices from the same retailer can vary greatly by geo and even locations. Also several stores, in store prices don't match online prices (costco is the obvious example, but nearly every store has in store exclusives that don't hit online advertising).

I also can't explain why it's easier than java, but it is. My theory is the syntax is just insanely clean, making it quick to read. There's also a lot of approachable scripts for it, that make it easy to get momentum. Javascript has too many "gotchas" to make it quick to learn. Sure BASIC is quick to learn, I learned the basics when I was kindergartenish, but it's not something you would use for the past 20+ years, and wouldn't be suited to modern applications.

There's also something about python, that it was/is extremely fast to learn and pick up.

> However, it does mean that Google will have to give up on its mission, “to organize the world's information.” Though that slogan looks benign, it is in fact anything but. Being the organizer of the world’s information is far far too much power for anyone to have. It’s time to give up on it.

Giving up on that is stupid, and would set society back. We need more companies and organisations doing it. AI has brought a few new startups into the search space, but yes, splitting the companies up may make more opportunities for them.

That said, if you can raise the capital you can get into search, because of opportunities presented by AI, and google has been sacrificing quality for optimising ad revenue. Email has also seen some startups making big leaps by adopting AI. I wouldn't touch video sharing platforms, or ad networks.

Seems to really fall apart on subsequent prompts, and a few times I've had code end up in the "thinking" tokens.

I'm guessing most of the training data was single-turn, instead of multi-turn, but that should be relatively easy to iterate on.

PhDs aren't relevant. It's more just a certificate that you can learn to learn and stay committed to hard and challenging things. It does give bonus points to VCs, because it's seems to be easier to market to other VCs, same applies for hedge funds.

And with fine tuning, there's zero math needed, it's a bit of common sense, and a lot's of data optimization.

holy fuck is your application long.

Ehhh, I'd rather hear that they stopped supporting SB 1047.