nicklevin

It’s really quite disappointing to see how fast just copy/pasting AI responses has proliferated, even into things that don’t benefit the copy/pasters. I’m doing an online course currently that has absolutely no benefit outside of learning the content (i.e. the certificate or whatever you get for completing means nothing) - yet classmates are very clearly just copying/pasting in responses for the exercises. How does that benefit them? More than any slop I’ve experienced thus far, this instance has made me the most worried/sad/pessimistic to see. If even people who are supposedly motivated to learn (why else would you pay for this course?) just revert to the easiest AI slop path, what hope do we have for avoiding it in stuff that more resembles “work”?

The EHR company Epic uses a similar naming scheme for the slimmed down version of their EHR (Sonnet) and mobile app (Haiku). Their Apple Watch app is Limerick.

Saw that coming. Just like Google did with the SEO heist a person bragged about a couple weeks ago, if you make big tech companies look foolish they are going to react quickly.

Pure speculation and just trying to connect dots... I wonder if they realized they are losing a lot of money on ChatGPT Plus subscriptions. Sam tweeted about pausing sign-ups just a few days ago: https://twitter.com/sama/status/1724626002595471740

Lots more signups recently + OpenAI losing $X for each user = Accelerating losses the board wasn't aware of ?

Arc moves the header bar stuff (URL) to the left pane, so the entire right pane is site content.

Simply a great example of using engineering and inventiveness to solve the problem they had. I remember noticing the screen having a flash, but because it coincided with the 'bang!' noise it felt natural.

Matt Levine frequently discusses the fact that losing billions of dollars means at some point, someone trusted you with billions of dollars. Which new employers (especially banks) look on favorably.

Totally agree. I give out my password to friends because their current tiers (4k also means you get 4 screens) make it seem like I’m paying extra to be able to do so.

There were seemingly odd hoops for customers to jump through to have their purchases qualify for Smile. Didn't type in the 'smile' URL? Too bad, no donation for you. One of the oddest was requiring push notifications in order for mobile purchases to count (https://news.ycombinator.com/item?id=21233815).

I also see 3 blocks in PA of 25k+ votes with 90%+ Biden (one is 97%) - can someone better at statistics tell me how likely that is due to chance given the other values (assuming the data is reporting accurately)?

There’s a row in PA listing Biden with 104.3% of the vote in a block with 1285 votes (the differential increased by more than the total number of votes). Data error?