David Gasquez
Data @ Protocol Labs.
Open Data, Open Source, Open Protocols.
Walks taker. Progressive Metal enjoyer.
davidgasquez.com
- When people are pressured to meet a target value there are three ways they can proceed: 1) They can work to improve the system 2) They can distort the system 3) Or they can distort the data — Donald Wheeler commoncog.com/goodharts-la...
- Reposted by David Gasquez[Not loaded yet]
- Reposted by David GasquezI can’t use ChatGPT to write because it doesn’t know what the fuzzy, half-formed ideas in my head are that I’m trying to crystallise into words. Until we hook LLMs up to telepathy machines it’s not much use. But it’s a good rubber duck to talk to about those ideas as a preliminary step.
- Agent Mode in @zed.dev is here! Time to give it another go. zed.dev/blog/fastest...
- Seems this is even worse these days. Cline, Codex, Aider, Cursor, GitHub, Roo, Windsurf, Goose, Claude, ... All expect rules/instructions in different places. 😅
- Dreaming of a world where I can use @marimo.io alongside Cursor's AI features seamlessy. Does that exists? Using the extension is not the same. 😔 Been thinking about having Cursor's features on its internal browser or something like that but I don't think that is possible.
- Reposted by David Gasquez[Not loaded yet]
- I think could be done with `llm` and some YAML frontmatter! 🤔
- Is there an standard or something like that to represent a conversation with an LLM in Markdow? System Prompt, Model, Config, User Input, Model Output, ... I'd love to have an easy way continue a conversation stored on a Markdown file (that can be updated / modified).
- My stack ended up being: - @pola.rs for data wrangling - @skrub-data.bsky.social for feature engineering! 😍 - FLAML and TabPFN for modeling - Scikit Learn for Pipelining/Model Selection/... Very impressed with Skrub! They have lots of amazing functions like AggTarget, Vectorizer, ...
