I've been disliking the recent trend of LLMs releasing with very promising and high benchmarks, only to disappoint by failing simple tasks or sanity checks. Plus, benchmarks don't really tell you about the LLMs' stylistic choices.

So I made 'Model Style not-a-Bench' to be able to quickly eyeball various LLMs, most preferably at 4-bit quants, under 100GB weights (excluding storage offloads like Qwen Flash Next's approach).

https://msnab.pages.dev

'Simple' is subjective, but it's just like phones in many ways.

Interesting! Thanks for the share

So interesting, including the website

This just sounds like a prompting problem to me, not a technical peer review one.

A redditor adapted this idea for the Samsung Frame https://old.reddit.com/r/TheFrame/comments/1wkg3fq/listen_fo...

Seems like a nice adjacent project to the magic mirror!

What are some good alternatives to Stripe for people to pay for your stuff?

Unsure how much I trust my searches for this kind of stuff on the web lately

Adyen is similar to Stripe, and it's an European company. But they select their customers, so you need to contact them, you can't sign-up "diy".

But there is literature concensus that at least in the US, non-US born CEOs are the best, and there is little relation between founders -> CEOs and good company performance on multiple metrics.

They can be the best at bringing in shareholder value, which would have a great correlation with enshittification. Having the ability to reduce the quality of something so that customers barely stay (or have no where else to go because they have been trapped in a monopoly or a walled garden) is not a skill. I'd argue it is a trait out of cartoon villains.

We watched very different cartoons.

Scooby-Doo > * when it comes to accurately depicting the predictable nature of capitalist villainy.

Eindhoven?