I've been using this, or more pointedly, I built an agent fleet (bespoke harness) where the planner agent uses OpenSpec to generate the plan. Then turns the tasks into a ticket graph.
It works fairly well, and it is definitely less heavy than SpecKit.
Not every org is going to run their own site. I have a similar frustration with companies that post on Facebook or Instagram. I deleted my Facebook account years ago.
Hopefully... If it doesn't, mere mortals like us consumers are going to be priced out of computers altogether. It'll be the end of the personal computing era and a return to big iron mainframes.
Kind of feels like if you spend $40K on your GPU setup, you don't want to cheap out on the PSU.
Still, you're right, I have a 1200W PSU for a single RTX Pro 6000, cost ~500 something, so still the PSU shouldn't be that large part of the total budget.
Funny thing is, you don't really want to just plug that into a wall socket. You want an uninterruptible power supply that can keep the machine running for the 5 to 8 minutes you need to get to it and shut it down safely during a power cut.
But you'll find most household power sockets aren't actually rated for the kind of power these things draw at peak. If you go for a dual GPU you're going to want to sustain a higher current. Think washer / drier / fridge kind of hookups for the UPS and that rig.
You spent $42K on the rig, you're also going to need to get an electrician in and have the electrical hookup seen to. The single 4090 I'm running put me right up to that edge where I could use a standard electrical hookup with a 1000w UPS.
So another $1K for the UPS, and $500 worth of electrical work (or more if they have to up the switch to 200 amps, for example).
True enough, depends how good your house electricity panel and wiring is. My basic plugs have 230V x 10A fuses so it would definitely take it to the limit. I also have some 16A which should be fine but I’d have to unplug one of the washing machine or heat pump type things. A 45k machine I would probably run on a UPS like you said. A lot of resistive heating household items use easily 2000W+ so it’s not super exotic but you do need to pay some attention
When you start getting into things like this, there's the initial price, then all the pile on of tax and the infrastructure needed to run thing. It's like getting a boat. For a mere $55K you can get a small one... but that is merely the cost of entry.
So you're right. If the $42K price tag didn't make you blink, the additional $2700 in tax, and the $2K to $3K for the work in your house are going to financial hiccups. But it means the cost to start is $50K, not just the price tag.
$42k and a toy CPU at that. Threadripper or EPIC if it must be AMD camp, although I'd rather it weren't (_always_ some random issues on linux). TBH, if I were buying workstation at such prices, I'd probably first take a look at what HP has (since their cooling and immunity to dust is unprecedented), and then probably BOXX and Puget. For DIY there's always supermicro at such levels.
I've noticed that some companies have flat out dropped the high memory configurations from their lineup that were offered just months ago at inflated prices. OnePlayerX had a 128GB Strix Halo tablet, RedMagic had a 24GB Android phone - both have been removed from their sites. It's a similar situation for the new Thinkpad lineup.
There are currently no large-scale deployments against them though. Of course the US is going to learn the lesson painfully if they don't take care of these information leaks.
The KGB was famous for compromising people sexually and recruiting then that way. The FSB supposedly still does, although homosexuality and extramarital affairs are not the scandals they once were. It would be useful for these countries to have embedded agents in the US military, even in peace time.
Indeed, but that's orthogonal to soldiers bringing their phones on deployments. They can also be recruited and handled when there's no way to separate them from their phones, i.e., off duty.
Why would they need to be deployed for an adversary to make use of that information? Sure it's more valuable on deployment, but wouldn't we also want to, you know, not give information on mobilization or lack thereof?
Large-scale troop mobilization are not subtle and can be sussed out by satellite observations. You can bet that all US forces are already closely tracked that way.
Invasions are usually spearheaded by elite forces that I presume are competent enough to understand that actively broadcasting what you're about to do is a bad idea.
So not the massive Chinese bot farm (200K +) launching anti-datacenter propaganda campaigns? It's just the green energy concern? ("Spiking local prices" was one of the propaganda talking points.)
These new AI data centers are now built with their own power generation capacity. They're too electricity hungry to be supplied by the local grid, requiring megawatts to gigawatts of generation. If they're grid-connected at all (for training / inference racks + cooling) it is secondary supplementation.
There is a concerted effort by foreign governments to slow down American progress in fielding AI capability. Attempts to generate civil unrest and propaganda against $new_thing are included in this.
At first I read the title and mistook it for an argument against the anthropomorphism of LLMs. It isn't. Instead it's a take on suggesting that maybe it's a bad idea to dumb down the self-chatter in the process. A reasonable take.
It isn't deliberately unhinged like Steve Yegge's take: https://yegge.ai/essays/model-welfare/ In Steve's essay he starts with the assertion that agents are sentient... Whether or not that's true isn't really relevant, as his agent-flavored version of Pascal's wager actually holds water, especially for Anthropic models, as their system prompts already push the model in that direction, and it is better to work with them than try to prompt against the tide.
It works fairly well, and it is definitely less heavy than SpecKit.
reply