The Deuce

The Deuce
[ The Deuce ]

Which ones is a bubble and when does it pop?

Posted by doubledown <iframe src='https://mb.boardhost.com/TheDeuce/info/user=doubledown' width='330' height='300' border=0 style='border:none;'> on May 14, 2026, 19:50:14

Looks like GitHub users are going to have to pay their actual cost for CoPilot

Posted by ATL <iframe src='https://mb.boardhost.com/TheDeuce/info/user=ATL' width='330' height='300' border=0 style='border:none;'> on May 16, 2026, 19:45:28, in reply to "Which ones is a bubble and when does it pop?"

https://bsky.app/profile/edzitron.com/post/3mlvw4kydv22v

GitHub Copilot moves to token based billing on June 1, and its users have now got a calculator that shows how much they've been actually burning every month - some $39-a-month users have spent $1500 to $5800 in tokens a month. Microsoft's subsidies are insane.

https://www.wheresyoured.at/premium-what-if-were-in-an-ai-bubble-part-1/

“We are old and tired and just want to be taken care of.” - Republican voter, Dec. 2024.

One developer posted this

Posted by ATL <iframe src='https://mb.boardhost.com/TheDeuce/info/user=ATL' width='330' height='300' border=0 style='border:none;'> on May 16, 2026, 19:48:51, in reply to "Looks like GitHub users are going to have to pay their actual cost for CoPilot"

“We are old and tired and just want to be taken care of.” - Republican voter, Dec. 2024.

Hallucinated citations becoming more and more common in scholarly papers - Lancet

Posted by ATL <iframe src='https://mb.boardhost.com/TheDeuce/info/user=ATL' width='330' height='300' border=0 style='border:none;'> on May 14, 2026, 22:37:27, in reply to "Which ones is a bubble and when does it pop?"

https://bsky.app/profile/helenbranswell.bsky.social/post/3mlcg4qeopc2y

“We are old and tired and just want to be taken care of.” - Republican voter, Dec. 2024.

Lazy authors of research papers thinking they need 10 more citations will always be a thing.

Posted by voiceofreason <iframe src='https://mb.boardhost.com/TheDeuce/info/user=voiceofreason' width='330' height='300' border=0 style='border:none;'> on May 15, 2026, 7:23:48, in reply to "Hallucinated citations becoming more and more common in scholarly papers - Lancet"

Instead of citing papers that they haven’t really read, they are now fabricated by AI.

How do these journals not have AI trained with every publication in existence to easily cross reference any paper cited to see if it is real? This seems like a good, low hanging fruit use of AI. Particularly since we have sources like PubMed with these lists already.

Previous Message

https://bsky.app/profile/helenbranswell.bsky.social/post/3mlcg4qeopc2y

Not only do tools to check sources exist, they can pretty reliably tell you if the source aligns

Posted by doubledown <iframe src='https://mb.boardhost.com/TheDeuce/info/user=doubledown' width='330' height='300' border=0 style='border:none;'> on May 15, 2026, 7:28:54, in reply to "Lazy authors of research papers thinking they need 10 more citations will always be a thing."

with what the paper says.

But when the author is paying to have their paper published in the journal, the journal's interest isn't necessarily aligned with the integrity of the scholarship.

Previous Message

Instead of citing papers that don’t support their point and hoping the reviewers don’t notice, now they are fabricated by AI.

How do these journals not have AI trained with every publication in existence to easily cross reference any paper cited to see if it is real? This seems like a good, low hanging fruit use of AI. Particularly since we have sources like PubMed with these lists already.

Previous Message

https://bsky.app/profile/helenbranswell.bsky.social/post/3mlcg4qeopc2y

A lot of that could be a skill issue and tooling. Reasoning models hallucinate a lot, so

Posted by doubledown <iframe src='https://mb.boardhost.com/TheDeuce/info/user=doubledown' width='330' height='300' border=0 style='border:none;'> on May 15, 2026, 7:09:45, in reply to "Hallucinated citations becoming more and more common in scholarly papers - Lancet"

you can't just raw dog model output into your work. Products like Perplexity have kind of figured this out by finding sources first, then synthesizing those sources into the output. I'd expect to see this fall in the future, especially now that people are really looking for it, hopefully event at open access journals.

Previous Message

https://bsky.app/profile/helenbranswell.bsky.social/post/3mlcg4qeopc2y

Here’s a sign that AI might be an overhyped bubble

Posted by ATL <iframe src='https://mb.boardhost.com/TheDeuce/info/user=ATL' width='330' height='300' border=0 style='border:none;'> on May 14, 2026, 20:46:17, in reply to "Which ones is a bubble and when does it pop?"

Sounds like OpenAI is not getting people to use it software, even though it comes somewhat built into Siri. I think the market is speaking frankly.

“We are old and tired and just want to be taken care of.” - Republican voter, Dec. 2024.

Might as well take a stack of $100 bills and light them on fire

Posted by jfinsocal <iframe src='https://mb.boardhost.com/TheDeuce/info/user=jfinsocal' width='330' height='300' border=0 style='border:none;'> on May 15, 2026, 8:34:36, in reply to "Here’s a sign that AI might be an overhyped bubble "

The margins on inference are negative and there's no indication that compute is getting cheaper.

Same with Anthropic.

Maybe they will figure out a way to pencil out the technology for some applications but I'm skeptical.

Previous Message

Sounds like OpenAI is not getting people to use it software, even though it comes somewhat built into Siri. I think the market is speaking frankly.

Open AI

Posted by memyselfandillini <iframe src='https://mb.boardhost.com/TheDeuce/info/user=memyselfandillini' width='330' height='300' border=0 style='border:none;'> on May 14, 2026, 22:06:07, in reply to "Here’s a sign that AI might be an overhyped bubble "

has a mind boggling product but their business acumen seems nonexistent. bring quirky like Musk isn’t enough.

« Back to index

234

[ The Deuce ]