Ask a bare LLM to multiply two big numbers and it'll confidently get it wrong. But ask an LLM on the web and it gets it right. What's the difference? In my latest post, I dig further into the harness, what it can do and why it matters: open.substack.com/pub/tidydesi...
Who's Who of AI
Hadley Wickham
Why they matter
Tracked through public AI activity and peer connections inside the directory.
- AI signals
- 5 past 30d
- Sources
- 2 distinct domains
- Discussions
- 6 past 30d
- Latest signal
- 1d ago
R, data, 🐕, 🍸, 🌈. He/him.
What they're sharing
Your LLM can’t math (and that’s ok) tidydesign.substack.com
Your LLM can’t math (and that’s ok) tidydesign.substack.com
cuda.ml 0.4.0: GPU-accelerated machine learning from R opensource.posit.co
View on Bluesky →
Introducing debrief: profiling summaries for AI agents opensource.posit.co
View on Bluesky →
AI outside of work tidydesign.substack.com
y code when ai? tidydesign.substack.com
y code when ai? tidydesign.substack.com
Kimi K3 and GLM 5.2 are now in Posit AI opensource.posit.co
View on Bluesky →
Auto mode knows it's dangerous (and does it anyway) tidydesign.substack.com
Help! My coding agent can run code tidydesign.substack.com
A no bullshit guide to LLMs tidydesign.substack.com
AI Newsletter: You probably don't want to fine-tune opensource.posit.co
View on Bluesky →
Articles & links
Ask a bare LLM to multiply two big numbers and it'll confidently get it wrong. But ask an LLM on the web and it gets it right. What's the difference? In my latest post, I dig further into the harness, what it can do and why it matters: open.substack.com/pub/tidydesi...
Hadley Wickham reposted
{cuda.ml} v0.4 is out! It brings GPU-accelerated machine learning with NVIDIA RAPIDS cuML to R. It now handles CUDA/RAPIDS setup automatically—no manual CUDA or RAPIDS installation needed. It also adds first-class tidymodels support. opensource.posit.co/blog/2026-08...
Hadley Wickham reposted
Happy to finally announce the CRAN release of {debrief}, a new package to let AI agents better interface with the output of the {profvis} package Making it so you can work with you AI agent more efficiently to profile and write faster R code! opensource.posit.co/blog/2026-06..…
I'm leaving for vacation next week, so I thought it might be fun to share some of my non-work uses of AI. See my weird and wacky uses at: tidydesign.substack.com/p/ai-outside...
This week on my substack: a write up of my talk "y code when ai" (my thoughts on why knowing how to code is still relevant), along with a brief description of how I turned a video of my talk into this post. Read it at open.substack.com/pub/tidydesi... !
This week on my substack: a write up of my talk "y code when ai" (my thoughts on why knowing how to code is still relevant), along with a brief description of how I turned a video of my talk into this post. Read it at open.substack.com/pub/tidydesi... !
Hadley Wickham reposted
I've been really impressed with Kimi K3 and GLM 5.2 the last couple weeks. I'm excited to share that we've made the models available in Posit AI with strong privacy guarantees and a steep discount compared to the Claude models currently available in the service! opensource.pos…
Ask a coding agent to run a curl-pipe-sh install command and it will — even if it notices the danger. I thought auto mode would protect you from this sort of danger but it doesn't: both Claude and Codex deliberately disable the sandbox. Read more at: tidydesign.substack.com/p/…
A coding agent is only useful if it can run whatever code the task demands, but a tool that runs arbitrary code is just as dangerous as it sounds! So how do you stop an agent from making mistakes you can't easily undo? Read more at: tidydesign.substack.com/p/help-my-co...
I wrote up the talk I gave for userR! 2025: tidydesign.substack.com/p/a-no-bulls.... A lot has changed in the last 12 months, but the bulk of this talk still resonates with me.
Hadley Wickham reposted
New on the @posit.co blog: In my experience, fine tuning LLMs is more difficult and more expensive than I've seen it made out to be. In most cases, I'd recommend focusing your efforts on experimenting with a bunch of prompts and existing models. opensource.posit.co/blog/2026-0…
Their own posts
Recent commentary
I've been thinking that not only did I learn to code before AI, I learned to code before StackOverflow, and even before the internet. Does anyone else remember implementing computer games by typing code out of an Usborne book?
Their network
In Hadley Wickham's orbit
Center = Hadley Wickham. Left = members they follow (green edges). Right = members who follow them (blue edges). Top = mutual follows (orange edges, slightly larger). Drag any node to reposition; click to open that profile.
Are you Hadley Wickham? Show it.
Add the Who’s Who of AI badge to your site or bio. It links back to this profile.
Markdown: [](https://aiweekly.co/whos-who/person/hadley-nz)