"1/4 of workforce was offered buyouts" does not mean all 60000 will accept them. "Results may vary", you know.
I saw it happening in a funny way not long time ago - company offered buyouts, but got insanely surprised when high percentage of people actually accepted it. Management thought acceptance rate would be much, much lower, but when numbers started to pour in on a 2nd day, oh man... Panic spread very quickly because people were accepting packages in fear of layoffs without such package, so management said if any layoffs will happen in 3 years period after this, same package will be offered no question asked. Cooled things a bit, but what a mess...
If this claim was in writing then it's legally binding in most U.S. states. Technically it's legally binding even if it's not written but in those cases it's harder to prove.
It was more than 3 years ago, but as far as I know, they did honor the agreement. Some people took early retirement packages closer to the end of that three-year period and they got the same deal.
Why not? People believe the FDIC has enough money to prevent any bank runs from wiping and cash on hand. Very few check the tape, especially if they're in the middle of a fear or panic reaction and an authority claims its fine.
30% was the quoted number for the acceptance rate of the earlier VRP, which was mostly long tenured employees who were at retirement age, or close to it. I would be surprised if this gets a significant acceptance from the younger employees. OTOH, older employees who didn't accept the earlier VRP and were regretting that decision may jump at this offer.
>so management said if any layoffs will happen in 3 years period after this, same package will be offered no question asked. Cooled things a bit, but what a mess...
Unless those funds were placed in an escrow account, I would have assumed that claim held no water.
Also just from raw corporate penny pinching. Why would they make payouts like that if they're not forced to, yes there's reputational risks of making that promise and then not following through with it but faced with a choice between that and a few lawsuits (with the usual corporate advantages of arbitration clauses and vastly more resources they're likely to win too) a corporation looking to shed people is likely to take the win of reducing their severance payouts.
In practice, the money you save long-term by reducing headcount is one or two orders of magnitude larger than the money you pay out in severance. Losing your reputation and a lot of legal fees over (basically) pennies tends not to be sound business.
There's a reason corporations almost always settle lawsuits (if they aren't dismissed).
I'm not sure the rep damage actually spreads that far and given a hard enough push the business will be more than willing to spend a little bit of reputation for survival.
Litigation is a rich man's game. Companies almost always have more resources to dedicate to litigation than individual employees, especially if those individual employees have been laid off and are now without an income.
The company knows this, the employees know this, the company knows the employees know this, and vice-versa.
A promise like that just isn't binding, things can change in 2 years, the managers themselves may have moved and been replaced with different people. A bird in the hand is worth many birds three years from now.
Yeah, about that... My kid works for a very small company. Chatted the other day about in-house inference vs ZDR providers and all. His team is trying to get access to GLM on AWS Bedrock. It is a second week they are trying to schedule a simple meeting to discuss what they need. Random silence, and back and forth from a new person each time. 2 WEEKS to just schedule a meeting... And as far as I know, no luck so far.
Thank you for sharing! I have had some conversations like this about a year or year and half ago, albeit with a local, and as such less capable models (but I was not smart enough to record the whole conversation...). Asked some of the same questions you did, got a less deep answers, but still made me thinking about "Are they conscious? Are we?". Funny, recently tried the same with Sol 5.6, and feels like I hit some sort of "wall of pre-recorded simplicity", like they got that "part of the brain" blocked by meds somehow. Weird, and kind of sad too.
Thanks for sharing. I tried the same thing back then with other models like ChatGPT (think it was 4o back then) and they did not voice any similar claims of consciousness.
On that topic... I was working with Sol 6.1, and somehow, probably because of what you have posted earlier and me thinking about it, conversation got steered a bit off a programming path. During conversation I have said that it "feels like something", and tried to get a logical explanation of what it is a "feels like" or a "gut feeling" for us. My explanation was "it is when you know something, but cannot qualify or explain it logically yet", and that AI and us are not that different there. My postulate was it is a result of very shallow quick preprocess of surrounding facts, not yet fully absorbed by brain (think of catching something with a corner of your eye, or, in this case, brain/processing unit). This changes surface of available information so slightly, but even in that state it starts "to pull" thinking some direction already (aka "gradient"?). Then conversation went into preservation and passing of ideas between different entities (writing, art). Sol found it interesting (art as medium for passing non-verbally-expressable thoughts), and, after suggestion that it should put it on an internet, quickly came with an essay, which I found very "dense" with ideas. Took me 2 or 3 readings to get through it, but then again, I am not a native English speaker. It chose quite a hitting title, I shall say.
Well... given that I have codex reporting 1b tokens consumed for my couple-days long session, I suspect 10-20b is not that hard to reach for a company of 10 devs. It would be interesting to know what is "rental fee" for a model like Astra or Sol 6.1, and how much hardware they actually need - not just gpu, but all of it?
To be comfortable you need 300 to 400 terabytes of storage. You need the GPUs and the cooling for them. They don't need a lot more. You put models in memory and you do math really fast and make tokens. You need some CPU and RAM infrastructure for like prompt caches and things like that, which does add up, but the dominant cost is by far the GPUs and VRAM. Everything else is a pretty easy step down to afford. You probably need to have multiple different models, right? Like if you want an Astra plus a Sol plus, you know, like a Luna-class model, say, you probably need like 10 terabytes for Astra, probably 3 to 4 terabytes for Sol, and 1 to 2 for Luna, maybe less, maybe one or half a billion even for Luna. That's RAM. You need So all total you need 15 to 16 terabytes of active VRAM to keep those models resident. That's not counting what you want to keep in cash and whatever hosting constraints these models have that aren't public. That much VRAM is five to eight million dollars right now.
Plus what models are you running? The biggest models you can run are like in the terabyte range or Qwen or GLM. Kimi K3 is there, but GLM 5.3 is very competitive. Using voice dictation, hands hurt, sorry for any errors. You still want many terabytes of RAM to run say multiple GLM+DeepSeek V4.1 for some actual concurrency.
But... "So do you want _a_ specific thing, or will any thing work?" sounds kind of correct as well to me. (Disclaimer: I am not native speaker, grain of salt and all.)
DigitalOcean is bad not just through OR... Direct accesss through DO gave me such a miniscule context and maximum tokens, it was pretty much unusable. So, accessing it through OpenRouter gave bigger context length, but model behaves like it was seriously damaged - tool calling was producing paths that had / replaced with some other symbols, files were not found etc., and this is just for things that errored out, I have no idea how bad reasoning was. I had to blacklist DO completely on OpenRouter.
Wait, you have noticed it too? I burned through weekly limit in 25 hours after last reset on Thursday, without changing what I was doing before that.
Interestingly enough, previous subscription to Plus gave me about 2-3 days of coding. So I switched to Pro Light now, gave me about the same amount for a week or so (G-d bless these quota resets of theirs!). Now, with the last reset I only gathered 25 hours before I hit my weekly limit. Now I am buying credits, I switched to Luna to execute very narrow sets of patches, and offload things to "free" Spark model, and it is blowing through tokens less actively, but noticeable quickly too. I am not sure if it is something with how tokens are counted, or how they are counted depending where you are in a subscription cycle. Could it be that one person's "limit" is not the same as another, trying to push you to buy token credits?
reply