21–30 of 1001 posts
I realized a better metric is in-flight requests... Each engine is processing 70 requests at any given time, streaming tokens for over 30s.
We grab interesting business problems, turn them into fun challenges for hundreds of AI engineers to find the best architecture for. Insights are shared back with the community.
It is a fun learning process with unexpected scaling challenges.
Being a weightlifter for 20+ years now, I'm working on a barbell speed and path tracking sensor based on newer IMU hardware technologies, which makes it both more precise and cheaper than camera- or actuator-based systems. Ultimately it helps you lift and train safer and better.
It's an intersection of industrial design, hardware, firmware, and software (and some sport science, of course). This intersection is not yet dominated by LLMs so it's a breath of fresh air.
In an early prototype stage as in "strap a Raspberry Pi to a bar", but it looks promising and I'm happy to move forward, also using connections from my previous 12+ years in China.
Well, all of a sudden, now that I kinda quit my gaming time sink, all my mini projects are finally being completed. All small, but useful, things for my setup that seem to slowly become a part of a bigger personal project. And between that kid and lots of books.
Ngl, it is weird for me now. If this is midlife crisis, I am loving it.
https://github.com/jondwillis/jacq
2) Claude code plugin based on some ideas found in https://www.anthropic.com/research/emotion-concepts-function The main idea is to add hooks that inject “baselines” under some conditions to counteract certain “emotions” that can cause subtle misaligned behavior in agents
https://github.com/jondwillis/functional-emotions
3) Final Fantasy XI custom client remaster in Bevy/Rust alongside an MCP integration that aims to allow agents to play autonomously on private servers à la “Claude plays Pokemon”
Contact: https://jonwillis.dev
- Built with Tauri — installer is small and start-up is near-instant on all three OSes. - No accounts, no telemetry, no MDX server in the loop. Sync goes through whatever cloud folder you already have (iCloud / Drive / Dropbox / a plain directory). - Tab-to-accept ghost-writing is bring-your-own-key
- Exports to PDF, HTML, DOCX. Tables, math, diagrams, code blocks all live behind toolbar buttons — no syntax to memorise.
Hope to have some people like it and use it.
I'm learning about inference by running vLLM on a k8s cluster (EKS), building a gateway to keep a Most recent ha-ha moment: I kept wondering if it was normal that my cluster was only able to process 4 requests per second per vLLM engine (just seemed really low to me). I realized a better metric is in-flight requests... Each engine is processing 70 requests at any given time, streaming tokens for over 30s. Code: https…