Patrick McEwen

The original thought the above article was actually about situations like this. Will hopefully get around to writing about that, but I think dealing with tagging data about links is far more complex

bsky.app/profile/niti...

AI seems to be like alchemy, get rich quick investment seminars, etc in that if they really had the good stuff, they wouldn’t be offering to sell it to you at any price because they would be too busy getting rich using it themselves

On Using ATProto to Improve Social Media Habits Is there a way to encourage sharing original content rather than uncredited reposts?

First @leaflet.pub post! I wrote about @noturtlesoup17.bsky.social having her content reposted and if there anything we can do with #ATProto to try to improve the situation

I think it’s just the difference between vibe coding done by someone who already knew how to write code without LLMs and vibe coding by someone who has never written any code themselves ever

Also need agreement on some sort of reform to ensure it won’t just all happen again right after the budget passes

Twitter always knew they had something with real time sports discussions and tried so many things to improve upon it, including streaming games on the site, that never really caught on other than maybe official team hashtags.

Could integration of the data and conversations lead to something better?

Probably true at pro levels, but not at the HS level where a great coach can set up the infrastructure to truly develop the talent they need to at least be very good ever year

A huge difference is the right doesn’t ever skip that. It’s mostly completely disingenuous and manufactured, but they do it constantly. If you want the media and elected Democrats to treat something like an impeachable offense, there has to be someone out there saying it’s bad and merits impeachment

I remain convinced that a major reason for the asymmetry in politics is that everyone one the broad left of center has started just skipping over the getting upset part and right to complaining about the hypocrisy, lack of action, asymmetry, etc, but skips right over the initial reaction

I haven’t had the time recently to really start digging into it, so perhaps something is out there I don’t know about, but again that is part of the problem

I think ATProto has a ton of potential for sports, but part of that is the ability to use shared data (like schedules, scores, stats) and discussions. I don’t have a specific vision for any of that yet, so I would like a way to explore what anyone else is working on in that space

Because I also have an idea but I don’t want to start from scratch if someone has already done something, but there isn’t a good way to confirm no one has done anything similar

Build the solution to your discoverability problem

Yes, I think that's spot on. To me, the people who are in trouble are all but the best people with pure code skill sets. The big winners are people with non-software expertise who can write some code. As writing code gets easier, knowing what code to write becomes the key skill

In this context, serverless doesn't mean the servers don't exist, but just that they manage the servers so that to the user, it's like the individual servers don't exist

Maybe better way of saying this is that software people are so used to being the ones that automate away other people's jobs that they weren't mentally prepared to process tech that could automate away their jobs, hence the freak out and all the predictions that AGI is right around the corner

Software is so new that maybe this is the first innovation that increases productivity so that fewer devs are needed. Previously most things expanded the number of use cases of software and thus they could hire more devs. It feels like an apocalypse because now software is like everything else

100%. I also think people under appreciate how much long run 2% productivity growth means tools that completely change how many people work and automate away some jobs has been happening regularly for hundreds of years. It can be ground breaking/awesome/revolutionary in that same sense

Also, people say there are no good discussions of AI on this app, but this one has been wonderful

The business deals OpenAI has made mean they are toast if they don't develop AGI, so there being an underlying tech reason for that as well at least makes the business decisions they have been making rational. Which I don't know is true for all the other big tech companies

I believe there are plenty of uses for it. Just none of them are likely to make any profits. The closest thing to current LLMs from before the last few years is translation apps. They are great and people use them all the time, but the business model is as a loss leader app for selling phones

Yeah, you're right. Could be decent autocomplete models are just so cheap to run users don't even bothering to optimize local vs cloud.

That dynamic puts an interesting constraint on the AI industry. They need models, GPUs, etc cheaper and more efficient, but not so much that we can do it locally

This conversation has made me realize that likely the reason this isn't happening already is that I get unlimited autocompletes for $10 a month. When they stop incinerating VC cash to acquire users by subsidizing my use, it will become far more price competitive to do it locally

Exactly. If I have to provide guidance and double check all the output anyways, that is the way to go. If it doesn't pick up on what I want, starting to write the code until it catches on via the autocomplete is way more efficient than typing out more explicit natural language

I have found I like to go either very big picture and think through a problem by asking for several options of approaches and the pros and cons of each to then implement myself or very small, namely autocomplete.

The stuff in the middle that is too much code to review quickly that causes problems

And given the percentage of the paying power users that use it for coding, we are potentially one innovation away from a local version with access to the entire code base all the time becoming better than the cloud version

Not sure I follow. So when inflation is low we should expect below inflation increases in electric rates and when it's high there should be higher than inflation increases in electric rates? Why?

Does that chart end in 2020? And if so, doesn't that mean it shows massive increases in exports during a time of below inflation increases in electric rates?

What is your theory as to why electric rates were flat from 2016 to 2022 and thus falling in real terms, but then rising at rates greater than inflation since then if it's not data centers?

It's great if the underlying tech is good, but it can actually be completely made up bullshit as long as the expert witnesses are convincing enough to judges and juries to get convictions.

For example, read up on some of @radleybalko.bsky.social's writings about bite mark analysis