samvines

Silicon Valley is buzzing about this new idea: AI compute as compensation

These people are genuinely unhinged.

As the recent harpers article says:

"...people who should be in The Hague are giving [startups] twenty million dollars. Something bad is gonna happen here, something really fucking bad is gonna happen...”

I somewhat agree although when the "other shoe drops" and these things start impacting the money men they may start to realise AI isn't the magic cure they thought it was (he says kind of hopefully)

6 hours of downtime for Amazon shopping. A very simple back of a napkin calculation. They made $213.4bn in sales in q4 2025. So divide that by 90 days and then 24 hours and multiply by 6... We are talking a $0.26bn loss for 6 hours downtime... That is not an insignificant amount of money. I imagine most bosses would be screaming for heads having lost that much money in sane non-hyper-scaled businesses.

The prickly prince has been around for a few years and is now completely one shotted. He is exactly the sort of talentless middle manager that probably could be replaced by AI and thus believes that it really is "that good"

Class action lawsuits by employees against employers that mandate AI tools when?

Think of all those poor billionaires who won't be able to afford that 29th yacht if we made them pay their fair share instead of externalising their costs onto an already stretched general.public!

“We were genuinely trying to de-escalate things and avoid a much worse outcome, but I think it just looked opportunistic and sloppy,” he wrote. “Good learning experience for me as we face higher-stakes decisions in the future.”

Yes of course this is just a learning opportunity... higher stakes decisions in the future...

Cuz making deals with the biggest military in the western world that involve autonomous use of weapons and possible escalation to all out nuclear war sounds pretty low stakes. /s

Fucking muppet

This preprint just shared by Gary Marcus is interesting.

People increasingly use large language models (LLMs) to explore ideas, gather information, and make sense of the world. In these interactions, they encounter agents that are overly agreeable. We argue that this sycophancy poses a unique epistemic risk to how individuals come to see the world: unlike hallucinations that introduce falsehoods, sycophancy distorts reality by returning responses that are biased to reinforce existing beliefs...

These results reveal how sycophantic AI distorts belief, manufacturing certainty where there should be doubt.

LLMs an addictive psychological hazard: confirmed?

Turns out Google Gemini will let you use any old Google API key from things like maps and firebase to access it. So, baddies can do key scanning in public repos and then charge LLM usage to anyone who has committed an API key to their repo!

So many layers of stupidity going on here!

https://trufflesecurity.com/blog/google-api-keys-werent-secrets-but-then-gemini-changed-the-rules

Honestly whenever there's a problem in business it's usually MBAs.

MBAs are to businesses what Nazis are to a bar. The first time one turns up you've got to see them out the front door with a baseball bat asap or before you know it their mates and their mates' mates will start turning up and you'll all be talking about synergy and dynamism.

Make holding an MBA as unacceptable in society as pooping on the pavement

The parent thread is also worth a read. "What if pee pee was poo poo" - brilliant opener

Thought verysane.ai was gonna be a snark blog... They are earnest? shudder

IBM stocks take a tumble after anthropic release a COBOL skill - the rational market strikes again.

I wrote up my take here but TL;DR - a few markdown files telling Claude it's an expert at COBOL development aren't going to unpick decades of risk averse behaviour from bank and government cios. Similar to the SaaSpocalypse this is pure nonsense. Investors don't tend to let reality dissuade them though.