In this post we will see how to extend reverse mode automatic differentiation to a language with first class function types, function application and lambda-abstraction. This method is not new, but we will give a new derivation of it by showing how it arises universally from noticing that the category of “additive lenses” is cartesian closed. In the end we will see that this idea sounds like it should revolutionise machine learning, but then doesn’t.

Some interesting ideas, although I won’t claim that I understand them all.

use the expensive models for planning/design and the cheaper models for implementation

stick with small/tightly scoped requests

clear the context window often and let the AGENTS.md files control the basics

Most software engineers are mercenaries typing for cash, but think of themselves as artists

The research-plan-implement workflow I use to build software with Claude Code, and why I never let it write code until I’ve approved a written plan.

I advocate for a yearly (or even twice a year) AI Lent, 40 days of no AI tools

If you’re building a new CI system/IaC platform/Make replacement: please just let me write code to dynamically create the workflow/infrastructure/build graph.

Visual guides for developers

Also why universal healthcare matters. Gotta maintain all parts of a system!

via: https://mit.edu/6.1800/www/readings/01-wrong.shtml

The first reason, then, that we care about low-level is that it allows us to make better choices. We can make better software by starting in the right place, with the right frame and the right stack. Low-level programming allows us to build trucks instead of Trucklas.

I love Ben’s work, but he has a bad habit of not giving credit where it’s due, and makes him come off as smug in his writing. Yes, I too care about performance in apps. But this should not be the sole decision-maker when picking the technology stack.

I don’t mind the ways in which my job is dysfunctional, because it matches the ways in which I myself am dysfunctional: specifically, my addiction to being useful.

Talk from the Roguelike Celebration 2018

Very nice software architecture talk. Patterns discussed: composition over inheritance, type objects, commands/actions.

How we handle 92 million compilations a year without everything catching fire

Exploring five critical questions shaping software engineering through 2026, with contrasting scenarios for each. These lenses help prepare for the evolving …