I just spoke to Apple. They insist it will arrive within a week. I cannot wait to work on that machine. Building Recursant should be fun.

Reposting to my slim number of followers. Give them a DGX Station please, he is working his … off for the OSS AI community.

Is anyone actually going to buy this RTX device and run Windows on it? Never really understood the proposition of doing AI on Windows. It’s too heavy-handed and slow. If I could run Linux on it, then maybe I should cancel my MBP M5 Max order.

As reported in the @WSJ, we have raised a Series B to bring Hermes Agent to new frontiers (and, yes, build a mobile app) Thank you to our investors including @nvidia @M12vc @SamsungNext @robotventures @usv @ycombinator @MenloVentures, among others wsj.com/pro/venture-capital/…

Kolibri-1 is fast on 2x Sparks.

Love having Hermes + my local models negotiating an EV trade-in.

Created new recipe for Kolibri-1, the new European model, optimised for two Sparks, running on TensorFold. ~2x tps (93 tok/sec) vs baseline (to be expected), but some sane optimisations. Publishing full TerminalBench run later today. Repo at github.com/ajensenwaud/Kolib…

Being a dual EU/AU citizen I am quite interested in Kolibri. Currently working on optimising it for 2x Spark.

Ordered maxed out MacBook Pro M5 Max nearly two weeks ago, still shows up as ‘processing’ in the Apple Store app and nothing has been debited from my account. Never seen this happening before. Could they cancel my order, just to increase unit price given the memory shortgage? @Apple

Presenting Recursant: cut the AI agent bill. Keep the secrets. Break nothing. Recursant is an agent-aware model router that means you can operate agents seamlessly across public and private model infrastructure. It is built in C and fast. It is my first major OSS project. It’s free. github.com/ajensenwaud/recur… So far it works with vLLM/TensorFold and OpenRouter, demonstrating 30%+ token savings on coding tasks and complex agentic workflows. Hoping to get some feedback from the X community on this one which is very close to my heart: making private inference infrastructure more accessible and compliant. Of course there is first-class support for @NousResearch Hermes Agent @Teknium

What a generational fumble from Mistral. The French startup was meant to be the sovereign AI alternative in Europe. Instead, it moved away from building models to applications and infrastructure. It decided to move away just at the time when open source models became good enough and cheap enough to be viable. Now Aleph Alpha has swooped in with a cheap, high-performing, open-source model. The German company has raised about 15% of the cash that Mistral has riased.

Posting ‘stop’, an htop style UNIX tool for monitoring DGX Spark clusters with built-in measures for vllm and TensorFold inference engines. Written in pure C, only dependency is ncurses. Download from github.com/ajensenwaud/stop. Run GNU make and run it. There is probably 100 tools like these out there, now there is 101. But I like it.

No, systemd. When I want my machine to reboot, reboot. I have never seen this before! Be a UNIX.

I can quite confidently report that GLM 5.3 Flash is not good for generating ANSI-style graphics for the terminal. I had to pull in Opus 5.5 for that.

This kind of misinformation (or ignorance?) on LinkedIn kills me. Instinct is NOT an open, local workflow. It is just as closed as the others. And what is up with the OpenClaw logo?

The best programming language of all time is C Just C

Inference engineering in nature

Something exciting is coming. I promise.