SleeplessCityLights

joined 11 months ago
[–] SleeplessCityLights@programming.dev 7 points 2 months ago (4 children)

Have you used Opus 4.8 at API costs? Without using agents, I can burn $20 an hour no problem. I use Kimi K2.6 and GLM 5.2 these days.

That article is confusing, are they talking about the RAM chips themselves? Or the packaging(sticks)? Or both? Also without an ad blocker on a phone, that article is herpes. Why would anyone voluntarily read an article from that site.

I am surprised that anyone who can make decisions realized the June Ai bill was going to be crazy. I was betting it would take until July, when the crazy June bill is charged, for companies to realize they can't afford frontier models.

[–] SleeplessCityLights@programming.dev -1 points 2 months ago (2 children)

I had to move to Kimi K2.6 because Claude models are too expensive to use. BYK from OpenRouter. I have not tested GLM models.

As they stop subsidizing, everything is going to get more and more expensive until they price everyone out. If they are actually "selling shovels for a gold rush," they are cooked. A lot of companies can't afford to use Ai heavily anymore. Customers are going to dry up.

It is Japanese copyright court, so it's a clusterfuck and loosely makes sense. Nintendo won't lose and PocketPair won't win. It will be some shit decision in the middle.

They are mostly getting hit because they seeded, which is hilarious that even Meta couldn't risk get banned for hit and run on a tracker.

High security data centers are usually invisible, unless you know that some random building is a datacenter.

I am going to give you one guess on who wrote the bill and is pushing the police lobby. All LE in Canada has an existing relationship with this company.

Its worse. The expected life span of the compute server hardware is 2 years. The processors cannot be re-purposed and the memory needs to be re-packaged in a non-novel way. They will most likely send it too ewaste processing to get the precious and rare earth metals back.

If people ever wondered who wrote and paid for Bill C-22 and the incoming online identification laws. Now you know.

I have found the opposite. Codex spits back mostly useless code that is twice the length it needs to be with a bunch of unessesary stuff and Claude is the only thing I get useful output from.

I don't understand how they don't know that as soon as you make a metric, people will try as hard as they can to game it, making the metric useless.

view more: ‹ prev next ›