Tfw there's a fart on the way and one seeks plausible deniability

academics write normally challenge: impossible

apoorvalal.github.io/lalgorithms/... new linklog - ecological inference, adversarial estimation, bargaining games with The Agent

Bay area people, come see Yin Yin with me! They're fantastic; think Khruangbin but actually good.

We have got exciting (and unconventional) stuff cooking and we are hiring for a strong research engineer on the GDM Game Theory team in London. Consider apply if you are interested in the intersection of game theory, multiagent systems and LLMs! job-boards.greenhouse.io/deepmind/job...

Nothing prepared me for how often newborns look like this

Acquired one of these newfangled learning agents. No API documentation though; time to figure it out by trial and error over decades.

simple little uv-runnable python script to generate interactive diffs for arxiv papers based on the underlying latex source. github.com/apoorvalal/a...

Looking forward to the beer potluck tonight.

the long shadow of the world cup 2006 final youtu.be/N0OBF83rL-8

(Z’X)^{-1}Z’Y

link 📈🤖 Challenges in Statistics: A Dozen Challenges in Causality and Causal Inference (Cinelli, Feller, Imbens et al) Causality and causal inference have emerged as core research areas at the interface of modern statistics and domains including biomedical sciences, social sciences, computer scie

Magic lasso to get you estimates and standard errors for "significant segments" of heterogeneity in your garbage-can DML regression where even the base effect isn't identified:

new linklog - economic model of online platform enshittification, assumption-lean inference with sample splitting, consumer search models apoorvalal.github.io/lalgorithms/...

News you can use (assume all the authors did hands on work)

casual causal inference: 1) metacritic colors badges differently w cutoffs at 50 and 75 2) given how prominently steam features metacritic, one might think this would manifest in higher purchase counts at 75 v 74. Turns out not really. 3) Rating distributions are weird. Check out 80.

How your email finds me

set it up just the way i like it github.com/apoorvalal/a... this is now my favourite computer [this is coming from an already high benchmark of tiling ubuntu via regolith, but the additional bikeshedding that folks have done for omarchy is nuts - free riding is the best].

Why has nobody been nerdsniped into building a bike stash cache on caltrain? The price of anarchy here is probably huge.

arch btw (but actually usable; bless dhh)

Mirifici Logarithmorum Canonis Descriptio

anyone have online radio channel recommendations? I've been making a more concerted effort to listen to the radio lately [thanks in part to this little terminal radio player i built for myself github.com/apoorvalal/t... ] and am on the lookout for stations that stream to a m3u/pls url.

economic models are surprisingly fun devices for expressing strident opinions with no empirical basis. attempt to bait theorists into working on platform design in tech with a caveman model of my own. apoorvalal.github.io/lalgorithms/...

bumping this for no particular reason because the profs are bloviating on llms based on hearsay again

Tossed Salad and Scrambled Eggs: 2/10.

numpy quirk i'm puzzled by - why do array shapes not correspond to axes for operations? array shapes are [rows], [rows, columns], [layers, rows, columns] etc, but operation axes always start at the outermost layer and go inwards.

whenever i come across a piece of math terminology that is smug and feels like an inside joke, it is always these dweebs.

Running an extremely underpowered experiment with no control group to estimate the long-run behavioural effects of in-utero exposure to Aphex Twin

tfw

apoorvalal.github.io/lalgorithms/... new linklog - the proximal pentagram, econometrics in torch, benn jordan wizardry

attn: associate profs in academia and L6+s in tech [getting through to people who once could do things and labour under the misapprehension that they still can is a unique kind of hell. if you've fallen off, just delegate and go quietly into the night; you're a glorified MBA now]

Typical day on the farm

Universities should return to their original mission. Yale should stop making fiddly locks with poor haptic response.

News you can use

5 (V, Five!) CATE estimation+policy learning papers in a single volume. Could probably teach a good class with this issue alone.

The fact that GMM (Generalized Method of Moments) works very poorly for estimating GMMs (Gaussian Mixture Models) is morally wrong. arxiv.org/abs/1508.06958

Increased bandwith turns Bart into Homer

i want my optimizer to have a "n_wolves" argument

Stopped keeping track of chronological index, i guess this is a habit now. apoorvalal.github.io/lalgorithms/...

My rube-goldberg machine: a tailscale vps with a cheapo mac mini as the server, and my phone(termux) or ipad(a-shell) as client, which lets me run local models frugally (bordering on fru-pidly) from anywhere. would be great to get local+cloud interop via `llm` cleanly.

Llm tinkerers, any idea what's happening in this example i cooked up? Multi-step tools + structured generation seems annoyingly error-prone? Kudos to @simonwillison.net for making llm such a useful tool for experimenting with these models github.com/simonw/llm/i...

bsky bots are exactly what you'd expect

One of the worst venn diagrams I've ever seen.

Godaddy has caught on to the "most tedious people in SV naming things after random pots and pans from middle earth" thing

link 📈🤖 Shrinkage-Based Regressions with Many Related Treatments (Dilber, Gray) When using observational causal models, practitioners often want to disentangle the effects of many related, partially-overlapping treatments. Examples include estimating treatment effects of different marketing touch

surprised that da yoot of today don't know the konami code of captive wifi - 192.168.1.1