1,572 karma · joined April 24, 2015
- WezTerm
- Signal
"Of course our American catastrophe played a role in our final decision."
"My decision was entirely because of the political climate in the United States."
"Both Toronto and the Munk School were and are very attractive places unto themselves... bracketing the American descent into fascism."
I haven't kept super close tabs on it but last year we were hiring for a role to do tech lead stuff and OSS community building for Vector, and yes several of the original Vector employees still work here.
It looks like backend APIs need to be written in JS and and then deployed as separate files. Not quite what I want, but I can appreciate why they went that route.
It'd be neat to have a project like this as a Rust crate, where you can write your own APIs in Rust and compile the whole thing to a single file.
2. Watch Inventing on Principle if you haven't already. Pushing build times into milliseconds or sub-milliseconds enables new capabilities
Bret Victor's Inventing on Principle is probably the best demonstration of why this matters; in the first 10 minutes he shows a very concrete example of an insight that would not have occurred without instant feedback https://www.youtube.com/watch?v=EGqwXt90ZqA
What, specifically, did you try? Like what LLM and what tools around that LLM?
I suspect that most people who express this view haven't tried cutting-edge LLMs and tools recently. I've met a lot of people who tried early versions of ChatGPT and Copilot a year or 2 ago, concluded that the tools aren't impressive or useful, and haven't revisited things since.
It's wild what Sonnet 3.5 can do with a tool like Aider; if you have a commit-sized piece of work in mind there's a very good chance it will get it right on the first try.
It's not magic, but it is very useful. I find that it works best if you have a commit-sized piece of work in mind; something that might take you half an hour but can be described in a few sentences.
But if you want a shell that makes it easy+quick to work with data in all kinds of formats, Nushell wins IMO.
I've recently been using Simon Willison's (hi simonw) excellent `llm` tool that can help with embeddings, and it takes the simplest approach possible: just store embeddings in SQLite with a few UDFs for calculating distance etc.
The simplicity of that approach is very appealing, but presumably at some level of traffic+data an application will outgrow it and need a more specialized database. Does anyone have a good intuition for where that cutoff might be?