| | | | Choosing an AI isn’t about “the best model” anymore. It’s about model strengths, app layers, and the agent harness around them. This guide lays out a practical framework for serious, repeatable AI workflows.
One Useful Thing | Ethan Mollick |
| | |
| | | Not everything that matters shows up cleanly in your event logs. This post explores why “friction” often hides between tracked actions, and what that teaches us about turning messy human behavior into measurable metrics.
Counting Stuff | Randy Au |
| | |
| | We used to assume memorization meant models couldn't generalize, but modern deep learning complicates that story. In this post, Andrew Lampinen explores how perspectives on memorization and generalization have changed over time, and why modern models blur what used to be a sharp line.
Infinite Faculty | Andrew Lampinen |
| | |
| | Here's a clean, visual intro to PyTorch that helps build intuition. It shows how tensors are initialized, why autograd matters, and what a basic training loop looks like. Great for refreshing fundamentals or making gradients feel less mystical.
Matthew Jura |
| | |
| | | | adrftools gives you smooth, continuous treatment effects in R. Estimate, plot, and test average dose-response curves with confidence bands and g-computation. Perfect for nuanced causal insights.
adrftools | Noah Greifer |
| | |
| | Think DuckDB for graphs. Ladybug delivers fast, Cypher-powered graph analytics without the overhead of running a separate graph database. Ideal for relationship-centric data at scale.
LadybugDB |
| | |
| | | Makie is Julia’s high-performance visualization library, and it just gained a GPU ray tracing backend. That means you can take the same scientific plot you explore interactively and render it with physically-based lighting and shadows, all without leaving the Julia stack.
Makie Blog |
| | |
| | | MIT’s “Missing Semester” focuses on the practical tooling that quietly levels up how you work with data and code. It covers command line fluency, version control, debugging habits, developer tools, workflow automation, and more.
MIT Computer Science & AI Laboratory |
| | |
| | These two topic-focused editions complement the regular Tuesday issue and go deeper into the topics you’re most interested in: | Data Practice — Everyday data work, including SQL, statistics, visualization, R/Python, and tools. Published Wednesdays. Learning Machines — Machine learning and AI, including tutorials, research, tools, and applications. Published Thursdays.
| Opt-in to one or both. |
| | |
| | | Start it once and walk away. HermitClaw will keep researching, writing code, revisiting old ideas, and forming new ones. It remembers what it did yesterday. It notices shifts in its own interests. It’s a folder-based, always-on research agent. Docker recommended.
GitHub | Brendan Hogan |
| | |
| | | | | How did you like this issue of Data Elixir? | 👎 1 — 2 — 3 — 4 — 5 👍 |
| | |
|
|
|
