jaredsohn

But that still costs you $5-$10 instead of having $5 in free credits :)

Yeah, this is practical and I did this myself to use MacGPT.

I think a benefit of a smaller number of larger packages is reduces the number of parties you need to trust.

The people in charge of each larger package can review contributions made to it and have the authority to reject code that includes attacks. If a larger package doesn't do a good job of preventing attacks, its reputation will decline and because there are fewer packages involved it is easy to remember these reputations.

Could see if GPT can come up with more names for your existing scheme; I think naming things is one of its strengths and you don't have to worry about hallucinations.

programmers he subcontracted to.

Also explains why he was so fast

I think a part of the purpose of pseudocode is it helps you focus on what the code is trying to do and it removes the distractions of syntax, style, etc.

You can probably just use LLMs to convert the pseudocode anyway.

Just having some extra money can make life a little smoother. Many of these things aren't that expensive but might not have been considered when acting frugally.

1. Buy extra copies of things where it provides convenience i.e. extra power cords for computers/phones that you leave where you need them. 2. If you travel somewhere often (i.e. family for the holidays or a significant other's house), consider buying an extra copy of things and leaving it there instead of packing/unpacking. 3. Buy personal items in bulk so you don't have to go to the store as often when things run out and you can save money this way over time, too. 4. Throw out clothing that you only wear when you have no more clean clothes and replace it with clothing you do like. Consolidate socks to one or two types to simplify matching. 5. Outsource laundry, cleaning, lawncare if you spend time doing that. 6. Spend extra on more convenient flights.

Agree. Anything you ask could be typed in by a human using chatgpt or google so requiring an answer quickly should be able to detect it was a bot. Could also ask a large number of questions and expect rapid responses for extra assurance.

Maybe check stolen accounts?

So this post is not about VCs and startups getting people addicted to engineering :) (That was how I first parsed the headline.)

Sure - but the important part in such conversations is the contrast of what different networks give access to.

The differences here compared to unit tests are that the breaking is outside of your control and the updating process is tedious. Also, testing requires making real API calls rather than using stubs so it requires additional infrastructure.

One thing I haven't heard much discussion about is the fact that ChatGPT is constantly being updated.

This means that if you build a prompt for classification and become confident that you've whacked all of the moles so that it is pretty solid with all of the edge cases, it can later start breaking again.

Some solutions I can think of are 1) choose a fixed model to test against but they become deprecated over time or 2) perhaps fine-tuning might help.

Naming things

But new episodes were coming out weekly after the first few seasons were released on Netflix.

Wikipedia: "It had fair viewership in its first three seasons, but the fourth and fifth seasons saw a moderate rise in viewership when it was made available on Netflix just before the fourth season premiere."

"Breaking Bad's viewership grew greatly as viewers binged the series on Netflix, helping to assure that a fifth season could be made. The fifth-season premiere had more than double the viewership compared to the fourth season premiere, attributed to the Netflix availability"

Set the temperature to zero to make it more deterministic.

They call it a monolith and the way they describe it I think they are using that term for the application rather than the repo.

And Flappy Bird and 2048

>The language models are a useful tool, they're not your job. It's as stupid as someone claiming to be an "IDE engineer" or a "stackoverflow engineer".

Agree if you're using ChatGPT to learn/generate/debug code for you.

But if you use an LLM to generate, summarize, or classify text from within an app then I think it is a skill just like being able to work with the Stripe API. Since the API takes freeform text as input and the output varies based on how things are worded, there is some extra nuance needed to use it correctly but I think that's basically it. Also, there are chances that the skill will grow stale quickly as LLMs improve.

You don't have to send the customer's name, etc. Perhaps can obfuscate some of the data before you send it, too, which might involve sending data to other models.

Not sure why that article went with centaurs; I think cyborg is a more accurate term. Just asked ChatGPT for a term for half human / half AI and it suggested cyborg as well.

Could look into how to show a webpage as the screensaver.

From a Google search: https://osxdaily.com/2016/06/05/web-site-as-screen-saver-mac...

Yeah - it needs to be clear to investors if the tech will scale as the business grows and if the tech has a good chance of improving if trained on a larger dataset or ML techniques improve generally.

Consumers should care about if a product is able to solve an AI-like problem that normally requires domain knowledge. Shouldn't care if done by ML, rules-based systems, or people. (Except perhaps may want assurance the product will continue to be able to support them as the customer scales.) Also should care about how the decision-making works.

Not sure why you're quoting the false advertisement text; it doesn't pertain to the argument of what is considered to be AI.

At least rules-based expert systems have always been a part of artificial intelligence. They shouldn't be called machine learning, though.

if/then logic is really just an implementation detail of rules-based expert systems. I think these rules-based systems can be differentiated from normal code in that there is a lot of compact domain knowledge encoded for classifications or decisions.

But if those things are considered AI then why would it matters if you market them as AI?

> Is XML better than JSON? For longterm stability sure.

Don't get that.

>to understand XML you kind of have to know how to work with trees.

JSON is trees, too.

Someone addressed this on Quora: https://www.quora.com/Why-do-jobs-start-at-Level-59-at-Micro...

Says that levels start around 50 but those lower numbers are for non-engineers. Not clear to me why it doesn't just start at 1 though.

A lot of Chrome development is public; check out https://www.chromium.org

One could pretty easily build a browser extension that does this automatically all the time for some sites - in either direction

Sort of related: https://asdf.com/asdfemail.html is still going strong.

"You would never guess how many people use asdf@asdf.com (or some variant form) as a filler email address across the Internet. Opening up @asdf.com lets in hundreds of email messages a day pertaining to porn site registrations, real audio downloading, mailing lists dealing with Windows NT ... the list goes on and on! If you would really like to send us an email, you can do so here."

I found reading books after "doing" for awhile to be useful to just improve how I organize things in my head. I usually find a few (but not many if a focal area of experience) things I didn't know beforehand, and it improves awareness of how much of it I do know.