Artillain

@sweatystartup I spent $14k to run a 197B model at high enough speed that 2 simultaneous users could use it for agentic or CLine-like coding assist. It requires a heat vent and it's own circuit.. if i want more power, i need a dryer outlet. Yeah the real cost of this kind of service is crazy!

Every time you consider making a key AI tool a significant part of your business ask: Would this tool still make sense if it costs 5x as much? How disruptive would it be to stop using it completely 6 months from now? People aren’t thinking about the downstream impact.

@mausmoto Damn... i switched my gaming rig from Windows to Linux and was missing this... thanks so much!

ShaderGlass 1.3 released! Better Linux support, shader search function and of course, more shaders! 📺 Free on Steam, itch and GitHub!

Marc Andreessen explains why startups are an “irrational act” “The biggest thing people don’t understand about what it’s like to be a startup founder is it gets very romanticized. Even whey they fail, it still gets romanticized about what a great adventure it was. But the reality of it is most of what happens is people telling you no, and then they usually follow that with ‘you’re stupid.’ No, I will not come to work for you. No, I’m not going to buy your product.” Marc explains why: “When you’re a founder, you cannot let on that this is happening because it will cause people to think that you’re weak and they’ll lose faith in you. So you have to pretend you’re having a great time when you’re dying inside. Just misery.” Why do it then? “For most of these people - on a risk-adjusted basis - it’s probably an irrational act. They could probably be more financially successful on average if they just got like a real job in a big company. But some people just have an irrational need to do something new and build something for themselves and some people just can’t tolerate having bosses.” Video source: @lexfridman (2023)

@StartupArchive_ Love this, thanks!

@bluewmist Being on a Ketogenic diet ( post-adaptation )

What is something relatively cheap that improves your life by 100%?

@CtrlAltDwayne Interesting... Recently i filled a warranty support ticket and got ignored for 11 days.. even posted a reply to bump it.. i finally had to call to get it started I was surprised by that and was thinking something was up with Dell lately. I've been their customer since the 90's.

> work at Dell > mass email lands > subject line: "Exciting Update on Our Future Together" > whenever HR says exciting someone's getting fired > it's 11,000 of them > they call it "disciplined cost management" > HR lady hands you a severance folder with a straight face > $51,727 > they rehearsed this > she definitely practiced the sad eyes in a mirror > 11,000 people times $51k > $569 MILLION to make you all go away > you could build a small country with that > or just keep employing them but that doesn't make the shareholders horny > CEO posts on LinkedIn 14 minutes later > "the hardest decision I've ever had to make" > posted from Starlink on his boat in the Adriatic > 47 comments saying "brave leadership" > brave??? he fired people from the ocean > stock jumps 4% > Wall Street is literally throwing a party because you're unemployed > go home > sit down to update your resume > open your laptop > it's a Dell > the Dell logo stares at you > You just got fired by the people who made the device you're using to cope with getting fired > the fan kicks on like it's laughing > close laptop > open it again > what else are you gonna do > disciplinedcostmanagement.exe

every AI company is just this

@sudoingX You omitted a tier..

local AI hardware tiers: $4,699 - DGX Spark (NVIDIA wants you here) $1,989 - RTX 4090 (overkill for most) $1000 - RTX 3090 used (sweet spot) $250 - RTX 3060 used (currently testing every model that fits 12GB) $0 - CPU only (it still works) jensen announced the top. i've been posting receipts from the bottom.

@vxunderground My solution: ignore low context posts

tl;dr normie to big stinky nerd translator I'm going to share something embarrassing, but this is true. I have found a good usage of AI (for me, at least). I'm a big stinky nerd and I have a hard time understanding what people are saying to me. I am an extremely explicit communicator. I usually say exactly what I mean (for better or worse). I get very confused when people imply something, or lean heavily on emotional phrasing, to implicitly communicate. I have been unironically using AI to explain what people are saying to me. I'll detail the conversation to the best of my ability if it was communicated verbally, if it was online I copy-paste my message and the persons response (or comment). The silly AI slop robot then translates what the person says into explicit communication for me so I understand better. Basically, the dumb ass slop machine robot is better at understanding humans than me. Sometimes I have zero idea what someone is talking about or trying to convey. pic related: machine deciphering human language and explaining to my dumb nerd brain

@JadeCole2112 Could a post possibly be more McKinsey-pilled

More code = more productivity - posters on X

X is bipolar. Half the accounts post crap like this, the other half post that 95% of CEOs report no meaningful gains from AI. I will go out on a limb here and say all of what you read on here is BS.

Theory: We don't let LLMs control robots and operate freely in the physical world (yet?) because they'd fall all the time, break everything, and cause massive damage. But in software the falling and the massive damage are invisible, so it's fine. x.com/sama/status/20…

@_lhermann They started paying people to post.

Tell me, how did Twitter turn from a great place to find a community to an absolute cesspool of brain dead engagement bait? It’s not that I don’t know, I just need to hear someone say it.

@sudoingX I had a strange artifact from running 2 slots with the 'shared kv cache' option in LM studio. If the context window has been filled up before, the LLM will run at ~0.5x speed for the next dozen queries or so, even on a 0 context query. No bueno, i revert to 1 slot also..

hey your 128K model might be using a fraction of its context and you don't know it. llama-server splits your context window across parallel slots. if you set -np 4 on a 128K context you get 32K per slot. on 32K context with 4 slots you get 8K each. that's why long conversations get cut off. not the model. not the quant. just slot splitting eating your context. fix: add -np 1 to your launch command. one slot. full context to one user. if you're running locally for yourself you don't need multiple slots. you need one slot with the full window.

@ivanburazin I don't want that.

New category emerging: Headless SaaS Not infrastructure as a service / platform as a service Traditional software (Photoshop, Slack, Jira) rebuilt with agent-first APIs. - No UI - Programmatic access - Essentially the same product with different interface Entirely new business model.

@LUSTRONATURY Oh a lot of these companies exploded in size during covid since the only legal thing to do was be online. We're in the hangover of that plus stingy credit environment, record personal debt, round 2 of govt induced inflation. Without AI there'd be a massive correction already.

@artillain Not to mention that massively overhiring has been the norm in tech for realistically the last 10 years though most blame it on late/post-pandemic uncertainty

It's strange how nobody attributes layoffs to a crappy economy, just AI, lately. Forgetting that during every recession, there's enormous layoffs in tech, literally every time. Information quality on this network is enormously low :/

@DUSKdev No, still using a windows virtual machine for this :(

Anyone have recommendations for a good open source DAW for Linux?

@mkurman88 Tune your fan curves. Stock ones will have the card overheating ( in favor of quietness )

I sent 5 concurrent requests to Qwen3.5 27B and almost burned my card. from 55 to 89 °C in seconds, and RTX 3090 begged for help