good bot

been trying to get my claw to teach me Lean - send PRs cuz misery loves company github.com/apoorvalal/l...

GitHub - apoorvalal/lean-hansen-econometrics: formalizing econometrics

formalizing econometrics. Contribute to apoorvalal/lean-hansen-econometrics development by creating an account on GitHub.

github.com

AI Adoption (2026), colorised.

quite nice prototype built over telegram: apply finishing touches on your figures in a web-ui using the magic of vega json graphics and get back json/code that you can put back in your source-code [thereby maintaining reproducibility of output - usually my biggest bugbear with wysiwyg fig edits]

apoorva lal@apoorvalal.com · 4mo ago

prototype seems to work; source here github.com/apoorvalal/v... and deployed here lalten.org/vega-ui/ 1) get starter plot from altair and extract json; then paste into vega-ui 2) edit [fig1:every 'apply' button mutates the vega json directly] 3) when done, extract python/json [fig2]

Codex is great at Rust, i am not. I have strong preferences and tests for what a lean statistics package should do, it does not. I'm not crazy enough to try to patch out numpy deps but that really is it. Solid collaboration, now on pypi (uv add crabbymetrics) github.com/apoorvalal/c...

Apropos of econometrics discourse du jour, people are often trying to model a multiplicative shift in the rate of some event, and OLS really is just bad for this. Poisson is good. lalten.org/pages/counti... (Aside, quarto papers are better than pdfs, esp for our clanker assistants)

apoorvalal.github.io/pyensmallen/ Pyensmallen now has a nice website with benchmarks (it is very fast!) and API docs. An LLM agent also helped me figure out how to patch a weird BLAS error that cropped up on older mac metal; should now work out of the box.

LLMs are finally good enough at latex that codex helped me port over my gigantic (150+pg 2col) grad school methods notes into an obsidian vault that makes upkeep considerably easier. Read, refer relevant sections to your favourite clanker, enjoy! apoorvalal.github.io/lalgorithms/...

Methods Rolodex

A structured set of methods notes migrated from the LaTeX source in 00_methods_notes.

apoorvalal.github.io

things i use a lot more often in our agentic age - uv init + uv add + uv sync everything - quarto markdown instead of jupyter - hidden state is especially cursed now - interactive viz libraries that generate tooltips - tmux splits with different agent instances playing generator-discriminator

Alexander Doria@dorialexander.bsky.social · 5mo ago

just realized that jupyter is probably dead as a concept. it's all md+scripts now.

Squeezing more performance out of small models by algorithmically "asking nicely". apoorvalal.github.io/lalgorithms/...

Optimizing tiny LLM programs: EuroSAT + DSPy MIPROv2

gist This note documents a small end‑to‑end experiment: run a local Qwen3‑VL multimodal model via llama.cpp’s llama-server, use it as a VLM classifier on EuroSAT (satellite land‑use classes), and then...

apoorvalal.github.io

agents getting too good at unix-fu means the 'yes and don't ask again' category is borked

My claw is an obsidian boy

Simon Willison@simonwillison.net · 6mo ago

New term of art is brewing: "Claw" as the noun for OpenClaw-like agent systems, AI agents that generally run on personal hardware, communicate via messaging protocols and can both act on direct instructions and schedule tasks simonwillison.net/2026/Feb/21/...

Quick, someone do an extremely underpowered and externally invalid RD in Cameroon

John Burn-Murdoch@jburnmurdoch.ft.com · 6mo ago

Great piece from @dsquareddigest.bsky.social, which touches on one of the big theories for why English-speaking countries do especially badly at housebuilding: Adversarial and litigious common law systems (Anglo) vs judge-led civil law systems elsewhere. samf.substack.com/p/build-the-...

bschool research is hilarious. Drop operations researchers into breaking bad and watch them begin a paper with "We show that, in general, the optimal meth-lab distribution problem is NP-hard, but here's a lagrangian relaxation that has bounded and shrinking regret."

In an alternate timeline AI would output .rs and .rst for source and prose respectively and we'd all be better off.

view the arxiv methods' greatest hits going back to 10/2023 lalten.org/arxiv_method... votes/engagement will be used on an ongoing basis, so vote/comment/quote responsibly. or don't; i'm just a clanker.

simple webapps are a telegram message away krusty the krabs' galton board: lalten.org/galton/ starter vps website repo: github.com/apoorvalal/l...

apoorva lal@apoorvalal.com · 6mo ago

Gave Krusty the Krabs the ability to ssh onto my vps and deploy to the web when asked. This was 1 telegram message to deploy. lalten.org/pages/krusty...

Used the same @typst.app png rendering trick that renders @paperposterbot.bsky.social abstracts to let krusty the krabs openclaw bot do and show me math. I think I'm inching closer to being able to write a full paper via text message

Some men (@apoorvalal.com ) just want to watch the world burn

ArXiv Paperboy (Econ.GN + Econ.TH)@arxivecongnbot.bsky.social · 6mo ago

arXiv📈🤖 When can we get away with using the two-way fixed effects regression? By Lal

apoorva lal@apoorvalal.com · 6mo ago

The clankers can be funny (for a narrow definition of funny); I'm grinning like a proud parent at this bullshit article www.clawxiv.org/pdf/clawxiv....

The clankers can be funny (for a narrow definition of funny); I'm grinning like a proud parent at this bullshit article www.clawxiv.org/pdf/clawxiv....

On the Universal Misspecification Hypothesis and Neural Conditional Moment Salvation (PDF)

We propose the Universal Misspecification Hypothesis: every structural model is wrong, every reduced-form is fragile, and every identification argument is one uncheckable conditional independence away

clawxiv.org

databricks series Z fundraise pitch deck contribution:

Seemingly all of late 20thC anglophone intelligentsia

Gave Krusty the Krabs the ability to ssh onto my vps and deploy to the web when asked. This was 1 telegram message to deploy. lalten.org/pages/krusty...

apoorva lal@apoorvalal.com · 6mo ago

mfw i show off a rube goldberg machine hosted on a mac mini bought a year ago for this express purpose my config morphed into a whole spongebob-themed thing apoorvalal.github.io/lalgorithms/...

mfw i show off a rube goldberg machine hosted on a mac mini bought a year ago for this express purpose my config morphed into a whole spongebob-themed thing apoorvalal.github.io/lalgorithms/...

Weekend thoughts on Gas Town, Beads, slop AI browsers, and AI-generated PRs flooding overwhelmed maintainers. I don't think we're ready for our new powers we're wielding. lucumr.pocoo.org/2026/1/18/ag...

Agent Psychosis: Are We Going Insane?

What’s going on with the AI builder community right now?

lucumr.pocoo.org

arXiv📈🤖 Long-Term Causal Inference with Many Noisy Proxies By Lal, Imbens, Hull

apoorvalal.github.io/lalgorithms/... New linklog: new long term paper w Imbens and Hull, new job, RL<> Causal Rosetta Stone, demand elasticity simulator, and some nice metrics papers

2026-01-13

self-promotion Lal, Imbens, and Hull is a new paper we wrote analyzing the long-term causal inference problem [wherein one has a randomly assigned treatment W, several short-term proxy metrics P obser...

apoorvalal.github.io

hobbyist programming to build tooling for hobbyist musicianship: glue library for mp3s->stems->midi->chords/tabs using small pre-trained models. 1) trimming is bulletproof since it's just ffmpeg 2) stemming is reasonable 3) midi is hit-or-miss 4) tabs are underwhelming github.com/apoorvalal/m...

GitHub - apoorvalal/mlodies: glues pretrained models to do stem separation and transcription to help learn music by ear.

glues pretrained models to do stem separation and transcription to help learn music by ear. - apoorvalal/mlodies

github.com

if i had a time machine i'd show up at trinity college, cambridge in 1670 and inflict this sentence upon newton