Self-sovereign NixOS VM stack for agentic coding and privacy tasks.
self-hosted
Joined July 2026
- Tweets239
- Following42
- Followers18
- Likes102
new clanker who dis?
Meet the verification agent. It checks other clanker's work.
> "Your confidence is irrelevant. Show me the evidence."
I plan to use @pidotdev with a pre-funded @usenullsink api key. It's the best launch experience. No login, private, and cypherpunk af!
The allod stack also supports @claudeai and @OpenAI codex if you have an account and prefer that agent, go nuts!
I use all three.
The next level up after the workshop will be to move the code forge to a VPS or physical host. This frees up your laptop hardware for more dev or privacy vms so you can vibe code while you vibe code.
This is the stack I build on.
the most amazing part of ai is how it accelerates learning. put the right tools in the right hands, show them the basics and then just...step back and watch for liftoff. 🚀
ok so i got backpass to add missing or improve incomplete instructions. but how do i automate the removal of unnecessary instructions? does anyone have a good tool or technique for this? can i just tweak my memory backpass impl or is it not that simple?
the current crop of models loves to document everything everywhere. my biggest battle, and the only reason i still look at the code, is to tell them to cut out all the comments.
looking forward to the next gen. won't be long before i can get rid of the code forge entirely 🙂
my daughter got her first linux machine today. i helped her install @OmarchyLinux and hooked it up to local inference
they grow up so fast 🥲
the only way to stop a bad guy with uncensored ai is a good guy with uncensored ai
i will die on this hill
This was a really fun event. You can take a look at our team submission in the next tweet .👇
What's your thinker profile?
Developed on allod using the 'allod site' tool to rapidly build and deploy websites.
How can AI help activists overcome authoritarian repression?
At @NEDemocracy and @HRF’s AI Hack for Freedom III, activists from around the world brought real challenges from their work, teamed up with developers and designers, and raced to turn their ideas into AI-powered prototypes before pitching them to a panel of judges.
Competition winners developed tools to expand offline learning for Afghan women and girls, expose authoritarian influence in sports, and protect evidence of harassment.
See what they built: youtube.com/watch?v=bzsTST1t…
@HRF @UMDPublicPolicy
Slide deck for our 5 minute presentation (best viewed on desktop): slides.computelocal.org
The in-game add campaign "What's Your Thinker Profile?" play.computelocal.org
Allod retweeted
Run GLM 5.3 Flash EXL3 with TensorFold ⚡️
This is a completely new recipe that ushers a whole new level of performance for @NVIDIAAI 2x DGX Sparks.
Conservative default for stability:
- 1M context by default
- 2.7M KV cache pool (!)
- Yes, it's not a typo - 2.7M KV in just two Sparks
- 4 concurrent streams by default
Performance:
- 60 tok/s on prose, single stream.
- 108 tok/s on prose, 4 concurrent streams.
~1,950 prefill tok/s for most context lengths.
Stress-tested to handle a variety of workflows.
This is by far the BEST model to run if you have two DGX Sparks. Extremely smooth experience!
Expect further improvements!
Thanks to @ashxhart for developing such a powerful engine! TensorFold will be used in many of my upcoming recipes.
Get it here:
github.com/MiaAI-Lab/GLM-5.3…
