@kaia@lain It seems to me that we can’t pay for what we have, can’t pay for what we need. Can’t do that so that means that businesses slow. That means less hiring and well shit rolls downhill.
If I could I’d invest in depression and anxiety. Cause the USA is working on having a lot of it. People could then buy it and put it on their bread line and snort it. Seems about our speed right now.
FYI I don’t get depression or any of that. I’d be depressed if I was homeless like I once was. That’s my benchmark. I’m not homeless so I’m doing ok lol.
Dot com boom + mom being murdered * dad disappears / no family contact = homeless. That boom was awful. The crash let me get out of the homelessness. So I kinda like depressed economics. It’s my place I know well.
@kaia@lain Plenty of RVs just no one wants to hire. I even tried applying to dairy jobs all kinds of things. It’s just not happening.
That and it’s extremely competitive. So many people applying. Just means you’re screwed. I think that some are just trying to make it look like they are hiring too. So it looks like they are growing in this shit economy. It’s so bad that the windstorm that hit here few months ago? Roofs are all fucked up, windows months without repair just plywood. No money to fix them.
@kaia@lain Well I can’t get a job even local mundane job. I’m like an RV expert with all my own tools. I ran a mobile rv service before working in tech. Im extremely capable, just no opportunities.
I dunno what to say. It has been over a year.
When you look at the data it’s like some 34 years or something to replace all the unemployed. At the rate we’ve been hiring people.
@feld Local or hosted models makes no difference, the reality is that the models need to be tuned per task. They need to behave in ways that are not harmful to the user and allow the user to opt for the model to run slower or faster for given tasks.
If you are struggling with the model handling your context from loss of attention or heads? You should be able to increase its accuracy at cost of performance. Run fast when you can, but when it’s actively harmful you should be able to turn up the heat so to speak.
@whitequark That’s why the rebreather runs constant partial pressure of 1.30 when breathing. This is basically a bomb if done wrong. The forces at play in these units with these small tanks. 3 litter and 6 liter at 200 bar for oxygen are enough to burn the boat I’m on to the water line in a flash if you’re not doing it right.
But if you’re diligently doing things and take appropriate precautions it’s not hazardous. It’s a risk but it can be mitigated.
@whitequark Thanks for the kind response. It’s important to realize these pressures are absolutely insane. Yes they are but nature has made what we can’t touch beyond 200 bar safe. In fact medical refuses to even go close to 200 bar at 99.9999%
@whitequark I think you mistake pressure. Anything over 200 bar is inherently hazardous. Believe me I know I boost AVO. It’s not rupturing that’s the hazard it’s auto ignition from Adiabatic effect. The heat from compression explodes it. Any dirt, dust or incompatible lubes or orings is instant death.
Very few tanks are above 200 bar to get 400 bar is insane. My hardcore bank gas tank goes to 400 bar. My compressor goes to 350 bar for a commercial dive compressor. I think you’re mistaking pressures of what’s reasonable for humans to mess with.
Also consider inside the swim bladder there is no condition that leads to contamination that can cause auto ignition. That’s not how that works.
Yes it’s cool what they do but two entirely different processes are at work that generate the hazard.
If you can buy 400 bar AVO. You better run if you open that valve. Not even nasa is that crazy.
@mischievoustomato I do care about privacy if it ever wants to be taken seriously. Not everyone wants to have things assimilated. We are not the Borg.
Like I make private projects all the time. They are in my personal GitLab. I would prefer to know my data is secure and not used. If I wanted it public it would be on GitHub.
When you ask yourself why your agent uses 10000-100000k tokens to do a 5-20 line diff? 20-250 tokens vs that? What?
Why would you need 100k tokens to do 250 tokens of actual code. That’s the pondering question isn’t it? It’s the design of agents is by definition wasteful. The way things are handled you can’t see. The monstrosity that is just lurking.
If you really boil it down to it. If you’re generating these small diffs to iterate code. Why would you need enormous token counts to get it done. That’s the age old question. It’s the question I ask. That’s the problem I’m solving. In that, the batching, streaming, queuing. The relentless pursuit of reducing token counts not increasing them.
If you can reduce token counts to only what’s needed. This means you can then use that. To increase user experience. If the experience is increased, you wind up with a new product all together. One that works with the user for the user. More importantly it doesn’t have waste. That reduces cost, complexity and therefore you can offer the service at an affordable price.
I’ve worked out the math and all of that freemium you paid for that was $20 for unlimited. I bet you miss that where you could just pound away new ideas. Without violent terms of service that said your thoughts and code become their intellectual property. What a violation of trust that is. Ohh let’s send a mass email to everyone saying not only will we charge more. But all your base, code and ideas belong to us.
If you want to opt out of that system, if you want someone who will protect your ideas and code. Cause frankly let’s face it. I don’t have any interest in stealing to train supposed better shit. No that just convolutes things more. Plus I don’t have the time, my interest will be to provide a superior service not steal from you.
Be too busy fixing it, making it better and working with more languages. That’s the ticket that’s important. Not this madness of whoa bro. We want to take all the input and make sense of it is crack talk. What? You’re gonna take broken code and try to train AI on that with its infinite wrong answers? Come on that’s not even reasonable.
What’s more important is to make it so the service works as intended. Reduces cost and reduces the time it takes to use it. Increases automation and productivity. At a time when these providers are seeing push back from corporate for not delivering they pull this shit. Not cool.
Anyway that’s my opinion and promise with AI as this goes forward and I get closer to release. No theft, no abuse, no fucking opt out button. You’re just opted out I don’t want your shit. You may be special may not be special I don’t care. Want me to think it’s cool tell me what you’re doing. I’m not gonna waste my time.
Cause like I have a life that’s lame. If you end up paying for my service? You’ll get to see my wife and I hoisting rackmount shit. See me screaming at the computer in useless YouTube videos? Not so much.
Who the fuck has the TIME to do that.
God help us all if that’s our future. To be headshots and theft. No thanks. These are the only headshots that are actually cool.
Hacking crude software by day.Hacking even cruder software by night.FreeBSD running Bhyve VM's FreeBSD, FreeRTOSElixir and C powered bullshit.All Underwater photos and video are mine unless specified otherwise. Living the Florida Keys Disease. Island living is not for everyone. Idyllic pair of dice, roll your fate.