The future ๐ซก
so many useful products will come from this & especially at the costs enterprises need when integrating agentic products into their organizations. Great work!
Replying to @michellechen
we architected the model to read from inputs in a single forward pass. this makes it incredibly cheap and fast when compared to LLMs, since there is no token generation involved.
in our live demo below, you can see how using a cascaded pipeline with clairic + clef + LLM is 3.5x faster and 77% cheaper than just using an LLM.
clairic-pipeline-demo.workerโฆ
Joshua Waller retweeted
#buck2 #nix #nixpkgs #nixos the integration between these things can be frighteningly good....
What's stopping us?
fzakaria.com/2026/10/17/one-โฆ
Confidential compute for AI is inevitable. Just will take millions in attacks and losses to realize that sub-5% degradation in performance is worth the cost & effort to migrate infra to private compute clouds
Jev for command and control is so nice. I cant believe the progress ive gotten with my tool building tools with jev. With full docs SAD/TAD/ADRs created and reviewed throughout lifecycle. May not be popular but damn, effective engineering
State machines have infiltrated the AI hive mind. Remember AI devs just have to be pushed to blockchain architecture & rediscover a new term to solve their hard problems
Joshua Waller retweeted
Foldkit + Cloudflare Workers docs
developers.cloudflare.com/woโฆ
Anyways...to prove this out, I actually got the goat engineering tool to a state to make decisions then let Jev drive decisions + opus 5.5 to recursively build out itself. All ADRs, reviews, docs, standards references driving the development of itself.
Idk I guess I'll just publish the products & let my friend nobody decide how effective this tool is. Feels empty out here maybe I should vibe carelessly?
How I'm using standards driven development with @typesafeai Jev + LLM. This example is where I leverage all of the public standards from DoD/CISA/NSA/NIST to help shape a product's maturity throughout its lifecycle
Writing up an article on details. Been a fun experiment ๐พ
Joshua Waller retweeted
Replying to @ChaoticBeautys
there's now a half dozen Private Inference providers making basically the same architecture decisions but little effort to promote interopability or consolidate audit surface
Phala/NEAR/Venice share the stack, redpill.ai has verifiers for those+tinfoil+chutes
How I'm using standards driven development with @typesafeai Jev + LLM. This example is where I leverage all of the public standards from DoD/CISA/NSA/NIST to help shape a product's maturity throughout its lifecycle
Writing up an article on details. Been a fun experiment ๐พ
Yall remember when the SLM was controlling the phone and producing meaningful information through chat or image analysis? That was pretty interesting and felt like the next step.
Seems like @typesafeai will make that experience on all devices no matter the system & FAST.
If this is the case next week, I would go tell the world how to integrate jev in every product they EVER built with Typesafe AI. I'm already promoting Jev at Accenture ๐คฃ, but I will start to become annoying & integrate my 6 laptops, 4 phones, flipper zero, gl inet routers, security cameras, etc. just to test if its possible to. I'm so hyped for this week.
Need more Nix experiments. I built nixo 6 months back to orchestrate isolated sandbox environments (network/dev tools/namespace/writable dirs/env). I could use nix flakes to support launching coding agents & agent dev tools from numtide/llm-agents.nix
github.com/HashWarlock/nixo
My ideal experience would be taking my nixos setup & make it managed through a decision model to help manage the nixos config across my devices. I was lacking the skills a year ago when I tried to build that but I think I can reattempt when I get some free time.
๐๏ธ the world hasnโt figured out yet that you can literally just fix everything with a Nix overlay
In this blog post, I go into how I use Nix, why you should use it, and in the tweet below is a runnable demo showing all of the concepts in this post ghuntley.com/nix/
Remove the tech debt of software maintenance. Our strategies today are so outside-in & trying to force ourselves to solve human problems. Not abstract enough. We will be headed to an interesting place once the kernel model + decision model start to become the agentic stack
Rethink how you speak to intelligence. Its no longer telling what to build. It is not a technical problem that will print money next.















