Chemist in industry working on materials & data science. #rstats, modeling, & sustainability. Owner of many pets. Mastodon: https://nitter.cf/t.co/AGQir8blUj
Midland, MI
Joined July 2009
- Tweets487
- Following2K
- Followers783
- Likes1.9K
Pinned Tweet
My holiday project was building dsprrr, a package for declarative LLM programming in R, inspired by @DSPyOSS. The core idea is to treat LLM workflows as programs you can systematically optimize, not prompt strings you tweak by hand.
This episode of The Test Set is worth your time.
Holding AI enthusiasm and existential angst in the same breath was refreshing to hear. Plus, who doesn't like hearing more about uv, ruff, and ty?
ICYMI, our conversation with @charliermarsh on The Test Set podcast:
youtu.be/NoxoQHhrkoc
70
James Wade retweeted
We are excited to announce Inspect Scout, a tool for in-depth analysis of AI agent transcripts: meridianlabs-ai.github.io/in…
Scout lets you go beyond simple success/failure metrics to detect issues like misconfigured environments, refusals, and evaluation awareness using LLM-based or pattern-based scanners. Scout includes tools for developing scanners interactively, validating rubrics, and exploring scan results visually.
We are especially appreciative of the feedback we got from @AISecurityInst, US CAISI, @METR_Evals, and @apolloresearch during the development of Scout.
Blog post: aisi.gov.uk/blog/a-pipeline-…
Website: meridianlabs-ai.github.io/in…
Coding agents can explore codebases. But you can't optimize them, compose them, or put them in a pipeline.
RLMs can do all of that. They're DSPy modules, not agents. I built an app to understand how they work.
Try it yourself.
You can run your own traces if you clone the app locally.
🔗 jameshwade.github.io/dspy-ex…
📖 github.com/JamesHWade/dspy-e…
I write a lot of code in R, so I also make one for my R port of DPSy called dsprrr
🔗 Interactive app: jameshwade-rlm.share.connect…
📖 How RLMs work: jameshwade.github.io/dsprrr/…
📖 Tutorial exploring a real bslib issue: jameshwade.github.io/dsprrr/…
2
72
I built an R package that turns Shiny apps into UIs that render directly inside Claude Desktop or ChatGPT.
It's called shinymcp. Drop-downs, plots, tables all inline in the chat as MCP Apps.
1
1
230
There are several limitations compared to a fully shiny app, but I'd love to hear ideas where you this might be useful for you.
jameshwade.github.io/shinymc…
1
1
39
You can check out the package here or install it with pak::pak("jameshwade/shinymcp")
github.com/JamesHWade/shinym…
28
James Wade retweeted
Really cool initiative from @JamesHWade! There are now versions of DSPy not just in Python but also in R (see below), Typescript, Rust, Ruby, Elixir, Go, Clojure, and Lisp.
My holiday project was building dsprrr, a package for declarative LLM programming in R, inspired by @DSPyOSS. The core idea is to treat LLM workflows as programs you can systematically optimize, not prompt strings you tweak by hand.
It's still early, but enough pieces are there to play around: >10 module types and optimization strategies (teleprompters). Built in bridges to vitals for evals.
Install with pak::pak("jameshwade/dsprrr")
github.com/jameshwade/dsprrr
1
1
97