Dee | Autonomous

1/ Building your own Personal AI Computer ChatGPT remembers your financial information. Claude knows your next big ideas. Gemini sees your inbox. You're not using AI. You're exposing yourself to corporations. Personal AI Computers are freedom. Here's how to build one: 🧵

@pupposandro At Autonomous, we build and run various boxes of 3090s, 4090s, 5090s, and 6000s over the years. So I can share our first-hand experiences: 3090s offer the best ROIs! If you build your first box for home AI use, I would definitely recommend 3090s. x.com/DeeAutonomous/…

respect the take but I think this is incorrect. instead, I recommend people to get a 3090: durability: 3090s run for years with basic maintenance. swap the gddr6x thermal pads every couple years and the card is basically immortal. just don't buy one that lived its life in a mining farm and you're fine power: 3090 pulls 350W. 4090 is 450W, 5090 is 575W, 7900 XTX ~355W. the 3090 is the LEAST power hungry card in its bandwidth tier. wrong completely !! heat: same story. every card at this performance level runs similar temps. not a 3090 problem price: used 3090s are $800-1000 today. launch was $1499. that's nowhere near launch price, and it's still the best $/GB-vram and $/GB-bandwidth on the market by a wide margin. 4090 is $2,500. 5090 is $3,500 the real recommendation depends on your use case. need more than 24GB of vram? skip the 3090 and go for a Mac Studio / DGX / Strix Halo. running qwen3 27b at high throughput on a budget? absolute best card on the market, no contest! ideally? you get both

@marcsh @songjunkr as big as we can! there are some 1x and 2x, but most rigs are 4x or 8x cards. a mix of 3090s, 4090s, 5090s and 6000s. added over time in the last few years.

@DeeAutonomous @songjunkr I consider them all part of the dev team, but thats not a common use case, so thats on me How big did you folks go on each one? 1298GB? 256GB?

You don’t need Local LLM if you: > don’t care about your data and privacy. > don’t mind frontier labs cutting your usage on every update. > okay with them banning your account for no reason, and not replying on customer service.

@icreatelife At Autonomous, we design and build AI hardware. So what I think about all day is form factors for AI agents. My fav way is to just ask the agents what they want for their physical body, add those components, get their feedback after they use the components, and continue tuning.

Coding and brainstorming are my fav ways to use AI You?

@marcsh @songjunkr I can count.. but 1:5 sounds about about right But not just for devs. Our GPU rigs power agents that work for all teams at Autonomous: marketing, sales, ops, finance, etc. Those non-dev teams actually need privacy more than devs.

@DeeAutonomous @songjunkr Smart How big did you folks go? I assume its not 1:1 for devs, but more like 1:5 or even 1:10?

@techopsasia oh so you are right up on our alley. privacy is a big theme of what we are doing at autonomous! i’ll dm you the office address. would love to have you here to take a look at our products!

@DeeAutonomous Sounds amazing! I am really keen to work on something in the hardware space as I really value privacy of the data and independance from the cloud (ironically)

I keep coming back to Vietnam. There's something happening here that's hard to explain until you're on the ground. In Saigon 🇻🇳 right now, delivering AI and cloud projects. Our clients are global and increasingly they don't care where the work comes from, they care what comes out the other end. That shift, from outsourcing to outcomes, is exactly why Silicon Valley is planting roots here. TinyFish AI didn't come to Vietnam for cheap code. They came for brilliant engineers. Worth paying attention to.

@chrisalbon We’ve been prototyping a device like this at Autonomous. Funny you mention $500 not $1000. We’re trying to hit the $499 retail price point before the release. What’s the best way to share more info on the device with you? Would love to pick your brain on it.

I want something like this that costs $500, not $1000.

@cormachayden_ lol. i can’t recall how many times i went out in january for dinners and did this to keep it running. get an intern! it will make your life so much better!

software engineers before vs after agents

@SummitSatoshi @Teknium loveeee it. thank you so much. sharing with our industrial design team now!

@DeeAutonomous @Teknium I present 2 directions @DeeAutonomous : 1. The pivot 2. The lean in (aka Sauron)

I have to go out of town for a funeral thru the weekend but I am leaving everyone with one new cool feature inspired by ralph loops and Codex's upcoming /goal feature. If you use /goal , it will start a loop with a supervisor model determining whether the task completed at the end of an agent loop - if it hasn't it will force it to keep going until it's done! Enjoy and have a great weekend. PR: github.com/NousResearch/h…

@iotcoi so beautiful

Normal people: weekend brunch, sunshine, human connection Me: Clean up & Ubuntu 26.04 on my Z8 Fury. Self-care = 7.0 kernel Scheduler: agent loops get priority, harness tools under ctrl NUMA: stopped playing telephone with my Xeons TPM post-quantum ready KV cache 72h marathons

@jxnlco Dude, it's 2026 and your setup is like a 2005 cubicle. You need an Autonomous Intern! x.com/autonomous_lab…

I’m limited by compute.

@Kimi_Moonshot @Zai_org @openclaw @NousResearch 17/ Housing & Assembly Wow. Overwhelmed by the response on Personal AI Computer! Some of you asked about the housing & assembly. Working on a follow up thread on that now. Will share soon!

16/ You just built your own Personal AI Computer. - With 2 GPUs (64GB+ VRAM) - Runs models like @Kimi_Moonshot and @Zai_org - Runs agents like @openclaw Hermes by @NousResearch No cloud. No leash. Private. Fast. And your prompts never leave your house!

1/ Building your own Personal AI Computer ChatGPT remembers your financial information. Claude knows your next big ideas. Gemini sees your inbox. You're not using AI. You're exposing yourself to corporations. Personal AI Computers are freedom. Here's how to build one: 🧵

@tunguz Yep, meet Autonomous Intern. Most people think UI/UX is a software problem. At Autonomous, we think it starts with hardware. Just plug and play.

UI/UX will win the AI race.

@DhravyaShah @shardul_mane @supermemory exciting!

when our margins go up, i just make the product cheaper. we're making @supermemory so much cheaper for everyone (soon!) it's to the point where it feels too good to be true. this is powered by some big infra changes we brought in the last few months ⚡, and more to come!

@garrytan @openclaw Cool! ETA ship date? Would love to add this for Intern.

@DeeAutonomous @openclaw Yes, I have a PR for this coming

Many such cases

@Lintilla369 The final build cost depends on the GPUs you're getting. But your cost estimation is sharp!

@DeeAutonomous Idea is solid, but to utilise a local model fully one need at least 45 tokens/s, I belive the cost for this is approx $5-10k.

@ClarkVenable @Kimi_Moonshot @Zai_org @openclaw @NousResearch You are right that your way has way few steps. This guide is for people who want to build their own Personal AI Computer. It requires some work. But it's fun weekend project to build one! About benchmarks.. here's the official one from @NVIDIAAI nvidia.com/en-us/ai-on-rt…

@DeeAutonomous @Kimi_Moonshot @Zai_org @openclaw @NousResearch 1. Benchmarks prove you wrong 2. You ignore cost of the above vs an Apple Silicone Mac. 3. My way has waaaaay fewer steps. 😬

@ClarkVenable @Kimi_Moonshot @Zai_org @openclaw @NousResearch lol. thanks for the feedback. but that's like... getting a honda civic while i'm showing you how to build your own ferrari here.

@DeeAutonomous @Kimi_Moonshot @Zai_org @openclaw @NousResearch I hate to be ‘that guy’, but you could just get a Mac.