I've never heard of Google Meet, but I usually see the successful devs who get Sherlocked are the ones who jump to a bigger pond.
As for platform dependency you want to have all your code internal and an API layer negotiating between the platforms' API (DOM/browser/rest/firebase/aws etc anything that isn't code you wrote) and your code
Then after accepting that the google specific stuff is a write off you can see how much work you are doing for yourself vs for google. Once you have that distinction you can make a velocity/ownership, feature orchestration/feature providing tradeoff
Following that you could quickly swap over to using Google's or Zooms or whoevers APIs (Meet add-ons SDK)
This seems like the worst of both worlds. If you don't care about the social/interpersonal aspect of reading then just use the internet. If you do, go to a real library.
Seems to be entirely a boredom-fueled bird-watching project for the maintainer. It's not a "nice thing" at all it's an out of place paternalistic nudge that makes teens feel unsafe in their neighborhood.
The only nice things that comes from them are teens "blowing them up", artists vandalizing them or homeless/vagrants selling the books.
More accurate to say that Humans are biased towards valuing "human" activity.
Logically text is text, and the value is independent of the creator. News flash people are racist/bigoted, they love to create arbitrary in/out groups and believe in patterns that don't exist.
Odd that sociologists have known this for hundreds (thousands?) of years but they still do pretend science, arbitrarily chopping up people into Men/Women/Children/Blacks/Whites what have you. Now we are chopping text into Human/AI/Spam/Ham. Were does it end? I'll give you a hint, the most efficient arrangement for all points in time is null. Material harm is much harder to fake than categorical harm.
The claim is always that this is somehow useful (yet to see a lick of evidence) but the reality is always oppression, inefficiency and abuse. The cure is worse than the disease.
Like OOP this ends with someone working full time to enforce segregation and do the required book keeping. Just the other day on hacker news there was a friend-foe extension. Somebody thought that was worthwhile, spent their time to create it, because segregation creates valueless work. Strong typing is just a human forcing the compiler to throw a tantrum because a purely arbitrary category was violated
In 2016 Musk was saying self driving cars where here now, he'll be saying the same in 2036. And Power from Humans (PfH) isn't a joke Manoj Bhargava is already working on it [1]. Of course the high quality human stock from Stanford consider this promising [2]
Usability is the wrong metric, paint by numbers is more "usable" (sic accessible) than a canvas but you'd be depressed watching your son graduate art school and that's all he can do.
If you do want to optimize for usability you have to make sure you aren't making the system more consumptive at the same time. The prime example from the article is trading a moment where the user must take initiative with a menu. More useable less useful. Lower the floor not the ceiling etc. Windows (and iOS) did make genuine improvements to OSs but because of decisions like these most users are locked out of enjoying them.
>bought “Yes” shares in the “U.S. strikes Iran by February 28, 2026?” market just hours before explosions were reported in Tehran and other cities.
If these sites were serious about their promise they would freeze betting so people have a reasonable window of time to change their behavior given new information. Otherwise insiders will wait too long, especially on bets like these where as the dates approach more people rationally vote no incentive's insiders to withhold as long as possible.
Or at least put this warning on every bet with this structure. Fallacious reasoning cannot produce value, you are just taking money from people reasoning poorly hence gambling
It hasn't even passed the original turning test, depending on the question. There are an unlimited number of questions that cause LLMs to give inhuman looking answers.
As for writing in general slop score is still higher than a human baseline for all models[1], so all a human tester has to do is grade it and make the human write a bunch, the interrogator is allowed to submit an arbitrarily long list of questions.
Google has enough momentum at this point to make the startups technically irrelevant, and they've only just started speeding up. Arguably they've already leapfrogged Anthropic and OpenAI in every field that matters, infrastructure, distribution, price, verification, all google, enterprise and actual model intelligence, currently a toss up but google is improving the fastest, SOTA in so many categories eventually this will coincide with the FOTM.
As for platform dependency you want to have all your code internal and an API layer negotiating between the platforms' API (DOM/browser/rest/firebase/aws etc anything that isn't code you wrote) and your code
Then after accepting that the google specific stuff is a write off you can see how much work you are doing for yourself vs for google. Once you have that distinction you can make a velocity/ownership, feature orchestration/feature providing tradeoff
Following that you could quickly swap over to using Google's or Zooms or whoevers APIs (Meet add-ons SDK)
reply