What are you doing this week?

If I have time, continue working on what I got to a mostly-feature-complete state during the weekend. I now need to work on the CLI client + FreeBSD ports/rc.d integration for trivial setup.

The idea is to have a remote service that stores disk encryption keys so that my NAS can auto-unlock its encrypted pool and so that the key is kept "unlocked" unless the server loses power for more than N minutes. (The scenario being: if someone were to steal my server, the key would become inaccessible before they could re-plug the machine elsewhere and look at the data.)

Reposting the POC videos!

https://x.com/jmmv/status/2083935858324042185 https://mastodon.online/@jmmv/117026590887907200

We have a bi-modal distribution of users for https://thetix.ai where one group likes and uses the app, while the other signs up but bounces within a week. We have been working on adding features for the latter group of people. So plan this week Is to finish testing the changes to the dashboard and deploy.

Studying k a little bit (reading this: https://xpqz.github.io/kbook), exercising

At work I'm finishing up a couple projects, writing some documentation, and then on Friday I'm going on vacation

Week three at the new gig and I'm doing my best to ship something before I head down to the Smoky Mountains for Miatas at the Gap on Thursday. A couple weeks ago I managed to new sway bars on the NC and get it aligned. I should be ready to take on The Tail of the Dragon with ease!

I do data infra for work and was going to finish a personal PoC local data platform that I decided to revamp. It’s currently a dagster, Kafka, spark and duckdb stack and I’m switching dbs to clickhouse, adding dbt, upping the load the pipelines put on the infra so they’re not kiddie problems and am considering making it deployable on eks. Seems like spark can handle more load than I’m giving it but alas.

Will also try to make more progress on a passion project, which is a performance engineering cli that is currently a harness making LLM api calls while trying to have safety/correctness guarantees. The idea was to install it, configure your (single gpu!) ml scripts/kernels to it, define benchmarks, tests, say what the LLM can and can’t touch and it should be able to optimize it by getting feedback from your hardware metrics. It’s currently buggy, expensive and super broad but I’ve read a fair bit about different performance topics, so it’s a good time. May also wire in an agent system but focusing on the safety aspect first