Open Source Grants Week 4: Every week, Autonomous sends a free product to an open-source developer. Drop your repo below to win: Autonomous Computer DIY Kit ($999). Fits 2x RTX 5090 or PRO 6000. Qwen 3.8 27B drops in 4 days. Good time to build.

These look amazing - I just dropped "Slate" open source app yesterday to help people with prompting custom for any video/image/music generator they need and made it so it can use your llm subscription or local without api. Can even bring image/video ref in.

@autonomous_labs sam, you're doing so many amazing open source repos!

@autonomous_labs Thank you, you guys are doing amazing builds! I ended up buying the harness! Cant wait to use it when it arrives.

Own your compute. Own your intelligence. Good time to build your own Personal AI Computer!

✌️

@autonomous_labs !!! OMFG yes yes yes yes yes yes

@autonomous_labs such a interesting concept. will dive deeper into it. you may want to reach out to @jack too. he's probably the best person who like this kind of project at the intersection of AI and monero. did you check out his Buzz app?

@autonomous_labs @jack Yes, im running a buzz relay on my umbrel that my small team and agents can communicated on, still figuring out its strengths but its really cool

@autonomous_labs Would be really funny win this with a repo about 3D models of cases 😆

@autonomous_labs I love love love love your 3D models. Did you check out our work on generative 3D products?

@autonomous_labs Oh wow! That's the next level of customization! Super cool!

@autonomous_labs It's all open source! Initially we used it internally to do fast-prototyping at Autonomous but we're now moving more and more of it toward production now.

@autonomous_labs Super nice! Thanks for sharing it!

The Autonomous Computer DIY Kit includes the chassis, which is CNC-milled from solid aluminum. It moves heat like a heatsink and the triangular cutouts keep it stiff without dead weight. Computer cases never look this beautiful.

The Autonomous Computer DIY Kit also includes the PCIe 5.0 risers with custom-cut cables. So your GPUs mount vertically without an inch of slack. It's just neat.

Oh, and the Autonomous Computer DIY Kit works with RTX 3090, 4090, 5090, or PRO 6000. 2x PRO 6000 can host DeepSeek V4 Flash. 2x 5090 can host Qwen 3.8 27B, which drops in 4 days. Good time to build your own Personal AI Computer. Drop your repo below to win this week's prize.

A few of you asked how the case is made. Here it is on the mill, cut from solid aluminum. At @autonomous_labs, we love CNC!

People keep asking me if the 2x GPU Autonomous Computer is loud. It's not. I have it next to my desk in the office and forget it's on. Even with both GPU at 100%. It's a beautiful machine and the talking point of the office.

@autonomous_labs Worth a try. My open source Laravel security package: warden.

@autonomous_labs will check it out over the weekend nathan!

@autonomous_labs Open source agentic trading strategies and statistical analysis

@autonomous_labs very cool. will take a look! have you see @virattt's work? his ai hedge fund repo is super cool:

@autonomous_labs @virattt Nice! I hadn't seen this one yet. Lots of exciting stuff in this space though. For example, this paper offers a nice overview of the field from just a few months ago

throwing this back in the ring because a system something like this would totally benefit from benchmark tools. llmgauge is a local llm evaluation/benchmark tool that helps you determine exactly which middle is best fit for your hardware. recently added support for vllm, and will be adding other runtimes soon. focused on real hardware, real testing, no hype.

@autonomous_labs cheez… love this. did you get your @localdotai handle from @alexocheema 😆

@autonomous_labs @localdotai @alexocheema no, I missed the sign up window. I tried to do it from my phone, decided i should wait and do it at my desktop and by the time i got home it was closed. 🤷🏼‍♂️

@autonomous_labs @localdotai @alexocheema also might i add the nice part about llmgauge is its truly local. no need to sign up for anything or get a user name because there are none.....just saying.

@mr_r0b0t @autonomous_labs Need it !! So sick I have the cards waiting for a new home

@autonomous_labs local ai models which are the result of a fusion architecture. working on 2 more models to release in the coming days.

@autonomous_labs Mac app to help run coding agents. Combining herdr, kanban and agent tracking Think Codex but with better agent management

@autonomous_labs 🔥

@autonomous_labs Growing steady, updating this at least 2x a month 🥲

@autonomous_labs will take a look this weekend!

@hnshah @autonomous_labs So when does it get a thrumming red heartbeat light behind the grill? 😂

@autonomous_labs cap - open source project that simplifies development and deployment of stateful confidential applications (enables everyone to host similar type of confidential solutions than Azure or GCP) (and enable more confidential inference providers)

@autonomous_labs I don't understand much about confidential compute, but I'll dig into your repo and learn more about it this weekend!

@autonomous_labs it has gotten a lot more attention lately due to confidential inference since it enables you to run a model in TEE and ensure that nothing leaks out so your prompts stay confidential two main building blocks are: - remote attestation (verifying what runs) - confidentiality

@autonomous_labs i have a site with more explanations around CI (and model/provider statuses) the idea of cap is to make it very easy to build systems on top of this tech, stateless is easy, stateful gets very complicated but CC is at a point where it can unlock a lot

@autonomous_labs This is so cool. What's the best way to setup TEE at home?

@autonomous_labs thats where is gets complicated, basically its just high(er) end server cpu's that have support for it (amd sev-snp, intel tdx) which means amd epyc 7??3/9??3 or higher, on intel side its i think xeon gold line nvidia also doesnt offer it (or disables it) on (pro)consumer hw

@autonomous_labs so the confidential compute happens at the cpu only or both the cpu and gpu? what happens when we offload the weights to the gpu for compute?

@autonomous_labs since hopper there is native tee so full confidentiality is possible. but you need both cpu+gpu as you'll also want to run some orchestration/coordination, tho technically you can do e2ee and encrypt user payloads with keys derived within gpu tee and thats keeps it confidentiač

@autonomous_labs 這個設計看起來非常完美！

@autonomous_labs we are not building a computer case. we are building an object that is around humans at home or in the office. it must look beautiful!

@autonomous_labs 我喜歡你們做的銀色的那部

@autonomous_labs yup. that's my favorite too!

@autonomous_labs

@autonomous_labs will take a look martin! looks like a helpful utility!

@autonomous_labs 1. 2.

@autonomous_labs Oh man MLX-Tune is goated. We recently began to explore local training on a cluster of local machines (both Mac and Nvidia) too!

@autonomous_labs Wow, I'll definitely look into it.

@autonomous_labs 🙏

@autonomous_labs Local models are the future - This kit is instrumental in making sure that's actually possible 🫡

@autonomous_labs I can see the repo now! Oh wow this is amazing Jay. Are you familar with our work on Grid? We just finished the distributed inference component on “a set of local machines”. We now started working on the training component.

@autonomous_labs I wasn't, but I am now. Will check it out! Looks super applicable for future needs of training and inference on owned hardware

@autonomous_labs yup. we think it’s an important problem to solve: an end-to-end training and inference system on local hardware.

@autonomous_labs Completely agree!

@autonomous_labs Would be helpful to get a CUDA port working for

@autonomous_labs super interesting. will deep dive into the repo this weekend!

@autonomous_labs Well, it's still a huge work in progress but this ONE of my current projects, a fork of OpenCode, entirely rethought for local LLMs (such as... Qwen3.8-27B) :

@alek_k13 @autonomous_labs HomeCode is a cool name and concept 😆 The other option I would go with is LocalCode!

@autonomous_labs Thank you ! Well, I thought about LocalCode but someone already took it, without adding what I added and plan to add so I decided to go for HomeCode !

@autonomous_labs 这机箱看着就很适合折腾本地模型

@autonomous_labs yup it’s absolutely beautiful in person. here’s one next to my desk. powering a small team of 4 in our office.

@autonomous_labs Open, payload-blind agent capabilities marketplace:

@autonomous_labs will look into the repo!

@autonomous_labs

@autonomous_labs what does this one do?

Agent Espacio is like a private Google Drive built for AI agents. Everything (images, maps, git repos, notes, 3D assets, workflows) lives as normal files in folders. Agents get full access + real-time WebSockets, so you can just say “make 20 images → turn them into 3D → publish the whole process as a blog” and it happens live with your team. my website runs on it.

@autonomous_labs The ui is designed for humans, and it has user management so the idea is to have a shared workspace between a team of humans and ai agents. A shared context with rich artifact types and support for every file type. Just added glb the other day.

@autonomous_labs Looks amazing

@autonomous_labs thanks! we put a lot of heart into designing these cases!

@mr_r0b0t @autonomous_labs Wow! Cool! Take your pick - (I contribute to

@autonomous_labs ❤️

@autonomous_labs This repository is a quantum algorithm web crawler. ☺️

@autonomous_labs oh wow. i don’t know anything about quantum 😅 would love to learn more. how does quantum help build a better web crawler?

Yes, I am trying to make an algorithm that quantum algorithms will be used to control the crawler. As I expand, I’m hoping to utilize IBM Quantum and GPUs to optimize the web crawler’s path as it traverses the web. I have successfully experimented with a IBM Quantum computer + Qiskit webcrawler code. ☺️

@autonomous_labs I also have this tiny project that turns an ESP32 educational board into a Home Assistant device. Currently building a physical AI companion with it that kids can interact with in a safe way.

@autonomous_labs love it!

@autonomous_labs Give it to @SamJWasserman !!

@autonomous_labs Very nice. Here is the only repo I have publicly open right now:

@autonomous_labs will dig into it the repo this weekend!

@autonomous_labs I have more information on VectorForge Pro if you are interested. Haven't decided what I am going to do with it yet. Took me 2.5 years and 4000+ hours to build.

Check out !! I really think full duplex voice model architectures are incredibly cool and elegant, and when at an NVIDIA event I went to, someone there told me it was super hard to scale (their personaplex model,which is a full duplex voice model.was capped at 1 concurrency per 80gb GPU), so I was able to figure out how to get realtime conversation concurrency to 32, and added vision to the model! Someone at Nvidia also liked it: (Ps: would love to put the rtx 5080 I won at an NVIDIA hackathon in it!! Haven't been able to get a case for it so this would really help, especially in letting me do some cool stuff around minimax h3 optimized for consumer gpus)

@autonomous_labs will do!

@autonomous_labs let me know what you think! mb, it was like 2 am when i was writing it so excuse my writing lol I couldn't pass up on the opportunity while scrolling X on my phone lol BTW, any chance that kit can fit the blackwell blower fan gpus?

@autonomous_labs @dee_hw any chance you had some time to look at it? would love any feedback! also, is this the last week you guys are doing this or can I participate next week (i'm obsessed and i need this case lol)

@autonomous_labs Voice-first AI agent to control your preferred AI harness. Full local: speech recognition, LLM, and voice on yor device. Only for macOS currently.

@autonomous_labs oh wow, this is cool. voice is definitely the right HCI interface in 2026. we've been working a voice-first device too!

@autonomous_labs

@autonomous_labs Cool. Watch you guys accidentally become the next Apple.🫠

@autonomous_labs The first Autonomous 😆

@autonomous_labs up!