Pardon the Mess: January 2026 Monthly Digest

Winter sun, winter trees (Brooklyn, NY — January, 2025)

Hello everyone,

Happy new year! This is the monthly digest you signed up for by Max Forbes — indie developer, AI researcher, mediocre essayist and photographer, and freshly never-graduated (as they call it) Recurser.

Today’s digest covers new website material for 3 months: Oct, Nov, and Dec of 2025.

Daily Blog

During Recurse, I asked Nick (a co-founder) about the much beloved blog of Julia Evans. He said that when Julia joined, they decided to publish a blog post every day.

This blew my mind. Daily!? I usually struggled to publish a few times a month. But I decided to try it out.

To my further surprise, publishing daily was easy. In fact, the harder thing was to rein in how much I was writing every day. Whereas I thought I wouldn’t be able to fill a post, it turns out I can easily write pages.

This phenomenon never ceased surprising me, as follows: if I stopped publishing daily, I’d think, “surely I can’t write a blog post today, I won’t have done enough by the end of the day.” Then, if I wrote a blog post that day, I discovered an enormous amount of stuff to write about. This same doubt reappeared every time I stopped publishing.

So, I really recommend publishing daily if you want to write and struggle to do it.

The only minor problem is that this threw a wrench in my website’s structure, which is designed more around meatier topic-based posts, and less around time-based blog posts. In particular the “recency” layout of the Studio page is a mess. (And, this digest will be too.)

Three Highlights

So, in lieu of figuring out the new website structure (hopefully soon), I want to draw attention to a three small projects I enjoyed working on in this period. (They also appear below in the full list.)

1. Mean and Median Surprises

Click to visit the interactive plot.

Say you have a set of numbers.

Consider this question: what is the number that minimizes the sum of squared distances between itself and your set of numbers?

Nobody I asked this knew the answer: it’s the mean of those numbers!

This post contains interactive plots to play with, plus a slight variation that produces the median. (link to post)

2. Maximum Likelihood Estimation

I will keep shouting about this table until it blows somebody else’s mind.

I somehow missed the unifying beauty of the framing of Maximum Likelihood Estimation (MLE) from machine learning classes. My suspicion is that’s because they dull you by first introducing MLE as a way of performing obvious tasks, like estimating the bias of a coin.

But listen. First, establish a very generic setup: write down how likely any data is under any model. Then, MLE does an amazing thing: ask just two questions (1) what do our outputs y look like? (2) what kind of randomness do we assume? And from this, the everyday loss functions like mean squared error and cross entropy just fall out.

This blog posts shows how this happens and has a nice little table. (link to post)

3. Embedded Spaced Repetition System

My favorite touch is the bloom (glow) around the review indicators in the top right.

I built a tiny embedded flash card system into my website. Just finished the first draft this weekend.

The idea is to integrate flash cards into the reading experience, with a ‘browse by default’ mentality: it’s ok if you stop reviewing, and you can always make reviewed cards visible. It works by loading a set of questions from a single file on my server, and saves all your reviews locally to your browser.

I am excited to start using it! (links to: manifesto, early development, live demo)

OK, them’s the highlights. Without further ado, here is the full list of posts that are new from the last three months.

Posts

Many daily blogs!

Playing Cartpole By Hand

Published	Jan 16, 2026
Tags	blog
Outbound	Mean and Median Surprises Scheming An Embedded SRS Embedded SRS Development Maximum Likelihood Estimation Flashcard Test Page

Pardon the Mess

Daily Blog#

Three Highlights#

1. Mean and Median Surprises#

2. Maximum Likelihood Estimation#

3. Embedded Spaced Repetition System#

Posts#

Garage & Microblog#