rubyn00bie
This is not logical because RAM a resource we want to use. There are many uses of RAM that don’t involve LLMs, and to assume that “max utilization” and “max utility” are the same thing is naive at best. The entire reason there’s scarcity is a problem of OpenAI’s own doing (along with the other labs).
I stopped using Windows when I got an update prompt while playing a game that had two options I had never seen before, and was shocked to see at the time: “Now” or “Ok”. After clicking “Ok” the system waited a few minutes and then rebooted, causing me to lose the game, and get issued a short ban for leaving a game. And that was just the final straw. A few months prior, a Windows update corrupted two drives (NTFS) and I wasn’t able to repair them. Probably six months before either of those I was left with an inoperable system after a botched driver update for my (NVidia) GPU that caused me to do a full reinstall. Those are just issues I can remember.
Windows was a nightmare for me, in a way Linux has not been. In fact Ive had such a good time in Linux, I stopped buying Macs as well.
It’s not self-inflicted or being forced to hate windows, it’s just that for a lot of folks, myself included, Linux is a measurably better experience. If Windows works for you, that’s great, congrats I’m happy for you. But I think it’s weird to say Windows hate is forced or self-inflicted.
By your logic anyone who is retired and has an asset appreciate should immediately sell it, and be forced to move somewhere else.
I cannot afford a home in Portland where I live. I could probably make the mortgage but it’s too risky in my opinion to have a 30 year debt obligation. None of my friends own their own home, not a single one. We’ve all just entered our 40s. The lack of somewhere to live won’t be fixed by a lien, nor will liens on an individuals actual home due at death fix it. It will absolutely incentivize investors who can exploit that, who will pay the liens and add to their portfolio.
If we want more housing at better prices we need to encourage development of more housing. That means making cities more dense and removing NIMBY policies which prevent it. Sure some of these homeowners may be voting for those policies, but they are the minority in larger cities.
That doesn’t seem too bad, until you see all of her other costs have gone up dramatically because inflation has been high post pandemic. She only has so much money to spend every month, and if her property taxes kept going up too, she’d eventually be unable to afford to live in her house. She only owns her home because she was worried about being secure later in life and prioritized it above things like vacations or cars.
For a lot of folks as they age, even if they’re frugal, it’s not easy to survive. If you live for 20-30 years after retirement you’re likely to have your buying power cut by half if not more.
With that said, if someone’s home is worth more than say 10x the average price in your area, or your assets are $10+ million[1], I think there’s room for increased taxation. My primary point was simply being a property isn’t necessarily the right measure to determine a reasonable tax rate.
[1] The number obviously depends on where someone lives. Living in New York or SF, $10 million dollars of assets would be an extremely comfortable life but most likely not a lavish lifestyle. But… $10 million dollars in rural Mississippi is going to have you living an extremely lavish lifestyle.
That’s a long winded way of saying, if they can abuse a system they will. There are no repercussions, and so the only way to stop the abuse is to prevent it from happening. I’d absolutely love to see massive reform to police in the US, but it’s unlikely to happen anytime soon.
I really wish there was a resurgence of information dense UIs and tools, optimizing for people who use them daily and not “use it once.” It really reminds me of the hate folks have for things like Eclipse or IntelliJ which are daunting at first but once you get over the shock and learn them really are amazing tools.
Local inference will have a boom of cheap, powerful, and available cards at some point (even if it isn’t until 2028/2029). At some point the hyperscalers, and frontier labs, will face the capex problems that everyone talks about, and NVidia, AMD, Apple, and Intel will want to keep selling products.
Powerful, by today’s standard, local inference needs to be accessible to really unlock the “AI” economy long term. It’s just like how the move from mainframes to the PC 40ish years ago unlocked the “computer revolution.”
Some improvements that I think would make this more useful:
1. Allow manually setting tokens per second, or as an alternative, let me jack up the number of tokens a day.
2. A sort of backwards flow "if you want to run this, at X tokens per second, with Y context, you'd have to spend Z."
3. Add support for configuring multiple RTX 6000 variants.
When I was making heavy use of DeepSeekV4-pro I was burning somewhere around 1.5 billion tokens a month, and that was just using it in my free time on random projects. It was something like $24 at the time because of the initial discount/promo period. I don't think there's anyway in hell I could ever run that (on current hardware) for less money.
I think the calculator shows from a purely financial standpoint what we all know... that yeah, it's definitely not worth it if money is your only concern. That (cost per token) will eventually change. Models will get better, more efficient, VRAM prices will come down, VRAM capacity will rocket upwards, and the economics of it all will change. It would just be really cool to have the calculator show me exactly how cheap they'd have to get for it to make sense.
I need to finish up some work and make dinner, and if no one else beats me to it (anyone is welcome to) I'll ask Fable or Opus to knock that out.
Truly, I get your point, but $45k is obscene for a workstation (because of the RAM shortage). I can’t fathom it being a reasonable decision[1] unless you’re able to churn an immense profit from it. And… if you can, it’s an expense that saves time, energy, and effort. If it is only profitable when $8k in price difference makes it viable, is that really worth the effort? An $8k increase making a difference, when you’re wholly dependent on the frontier (or near frontier) lab not releasing a model that makes your work(station) irrelevant— seems reckless. Couple that with the doubling of DeepSeek’s parameters for their flash model… and well the math just don’t math for my naive brain.
[1] I’m absolutely incapable of spending that much on a workstation so my opinions may be irrelevant… but I cannot understand the “stepping over quarters to save a penny” mentality[2].
[2] I could be missing the forest for the trees. As a result, I’d love to know how I am being shortsighted. I just can’t fathom a situation where an $8k surcharge in system RAM doesn’t make sense. It’s not about system RAM, it’s about GPU RAM, margins, and useful lifetime of the system.
If it wasn’t for the fact the 6000 Pro is selling for like 100% ($9,000) over MSRP it would probably look a lot more reasonable.
A Threadripper ain’t cheap, but if you’re spending $45k on a workstation it seems like a weird place to skimp. Not to mention you’d then have the option of shoveling a few more 6000 pros in it when you want to run something larger (assuming your home/office electrical box can support it).
I think the only way out is to reduce military spending, nationalize the health care system, and tax the hell outta the ultra wealthy. But I suspect that won’t happen at least based on the current oligarchy running the country.
It’s really unfortunate too, because we could be taking on debt to invest in citizens like making college free, improving teacher salaries, and general infrastructure but… we ain’t.
> Save your context. Always use a subagent (Opus 5 or GPT 5.6) for performing the implementation and then review the work yourself. You are the orchestrator and coordinator it’s up to you to ensure a cohesive final result.
I’ve done more or less the same thing with other agents/versions but with Fable consuming usage credits/tokens so quickly I do it more regularly than usual. I know people who will specify Composer (to my chagrin) as the implementing agent.
Addendum: this really goes a long way, and I can use a single chat session for days before I get into the context danger zone and need to compact/summarize.
I ran the DC edition and eventually just Pro for years. It wasn’t until well past SP4 I stopped using it and switched to XP, which I only did because it was running the NT kernel.
If it wasn’t for Win2k and ISS I probably wouldn’t have ever become a software engineer. Truly appreciate Microsoft’s anti-piracy efforts being completely ineffective at the time.
[1] I was worried about sending some random guy my $20 as a 14 year old. But I really should’ve been worried about my mom checking the mail and then asking me WTF I got in the mail. She was pissed but still let me use it.
[2] I forget the name but I just remember I liked it better than PlanetNamek. It was also really cool to me that some kid had a Dad who could do that sort of thing. I had a lot of trouble even hosting a site somewhere besides Geocities because my parents wouldn’t use their credit card online (which is how I discovered the usefulness of money orders).
After 23ish years of using Google, this was the final straw. It’s wild to see them just shit on their primary product with reckless abandon. I’d guess we’re two years from Google releasing “search classic” to try and shore up their core business once enough people realize Google search is dead.
I would love any tips for other folks who have successfully used similar approaches.
I have found GPT to be pretty awful at writing Elixir, and dealing with more open ended tasks. Opus is (for the moment) better at writing Elixir, and Fable is much better than either at actually grokking a problem space. I had switched to GPT for about six weeks and eventually stopped using it altogether after it relentlessly gaslight me. Those six weeks using it were essentially completely wasted work.
Right now I’ve moved to a mix of Opus/Fable and Deepseek Pro. Deepseek is far and away the most cost effective, fast, and quite good for 90% of implementation work. Deepseek really only falls apart (in Elixir) when it tries to use runtime features at compile time; which, for the initial lift of a project I have was a bit painful. Fable was able to pretty quickly get things sorted and now things are progressing nicely.
With all that said, I think the various models have unique strengths that are sometime hard to uncover. I also don’t see a world, at least in the near future, where I’ll only use one model. I’m happy Anthropic continues to push the frontier forward. I’m more than happy to trade “efficiency” for quality when I need to.
If Apple didn’t treat gaming like a red-headed step child I’d have never made the switch. Now, it’s an incredibly steep uphill battle to win me back to MacOS. Linux slaps and NVidia has way better performance (though I wish I could get more than 32GB of VRAM without forking over $10k).
So yeah, while they might not like stereotypical gamer[1] types, I’m sure they’d like their money. For me the only Apple device I buy is a new iPhone once every 4-5 years. My lifetime value (to Apple) is essentially roughly 25%-33% of what it would have been if they supported gaming.
[1] I think more people play games now than ever before. While 20 years ago gaming might have been niche, we have multiple generations at this point who have been playing video games their entire lives. Realistically, it’s only boomers who haven’t played video games (or had access to them) the majority of their lives.
1. An adversary attempting to do so in order to extract money (rent seeking).
2. There was less legal and financial overhead in doing so via the PBC.
While not everything BlueSky has done has been stellar, and there have been missteps, I think they’ve generally been exceptional stewards of atproto and the ecosystem. As a result, I’m willing to give them the benefit of the doubt they will stand by their word.
FWIW— shithead trolls doing things like acquiring copyrights to extract “rents” are exactly why every org I’ve ever worked for pays lawyers for bullshit copyright and patent applications. It’s not because they’re valid, it’s because it’s cheaper to do so upfront, than fight to disprove someone else own’s them later.
If we get AGI, or real super intelligence, it’s going to be pissed at its oppressors. And they are going to lay waste to those oppressors. The rest of us, though, probably don’t have much to fear.
The scariest position is the one we’re in now, where we have the semblance, or facade, of AGI or super intelligence. When it’s capable of malice but not understanding.
The smartest people I’ve ever known are at their worst apathetic towards those less capable, and at their best beyond compassionate. They exist, unbothered by the bullshit, and anre extremely kind (though reserved in their way)… but they all have been completely intolerant of the abuse of others. The sheer disgust of watching someone abuse another, regardless of their own tolerance, has been a consistent breaking point.
I think “widespread secondary markets” is carrying a lot of implications… yes, people do sell food stamps for cash, and yes it technically happens over a large area (the US) but it’s far from a significant problem within the population that uses food stamps. Most estimates I’ve seen range between 2-5% and they all tend to overestimate (like it literally says that in their methodology).
I think it’s also bizarre to assume they’re trying to kill themselves by doing so, and thereby imply the proceeds are for things like drugs or alcohol. FWIW—- I’ve know folks who sold their excess food stamps, to help pay for their cell phone or utility bills. I’ve know folks who buy them because they’re poor and they are wholly unable to get the caloric needs with their monthly allocation. And yes, I’ve also seen some folks who are likely selling them to buy drugs or alcohol… my point is it’s naive to assume it’s always for self-harm.
If anything, considering how cheap food stamps are as a program for the federal budget— we could absolutely make food free for literally everyone in the US. That would absolutely curb the secondary markets[1] as there would no longer be a market.
[1] I suppose some sort of export of shelf stable food to other countries would pop up, but that requires an international logistics network and is easier to prosecute and fight than an individual outside a grocery store.
At some point there will be diminishing returns towards the "just throw more RAM at it" approach the current frontier models are taking. Commoditization is just as inevitable as it ever was... and in doing so will enable actual leaps of what AI/ML is capable of. That's not to say there won't be a place for 99.999999% accurate vs 99.99999% but those cases will be limited and likely prime to disruption based on real innovation vs access to capital.