Things I saved in January 2026

I have personal Slack org with a channel where I push interesting links and finds. Here’s what caught my attention in January.

On doing the thing

Two pieces hit the same nerve this month. First, a short essay from Strangest Loop called Things That Aren’t Doing the Thing:

Preparing to do the thing isn’t doing the thing. Scheduling time to do the thing isn’t doing the thing. Making a to-do list for the thing isn’t doing the thing. Telling people you’re going to do the thing isn’t doing the thing.

Reading about how to do the thing isn’t doing the thing. Reading about how other people did the thing isn’t doing the thing. Reading this essay isn’t doing the thing.

The only thing that is doing the thing is doing the thing.

Then in a Hacker News thread, this comment:

The “doing it badly” principle changed everything for me. I spent weeks planning the perfect architecture for some automation tools I was building. Then I just… stopped planning and built the ugly version that solved my own pain point.

What surprised me was how much the ugly first version taught me that planning never could. You learn what users actually care about (often not what you expected), which edge cases matter in practice, and what “good enough” looks like in context.

I’ve always been a fan of starting a project with thinking: “What’s the worst version of this that we can build?” It takes the pressure off of v1 being anything but crap.

Writing advice

Gwern wrote a short piece called First, Make Me Care. The idea: some nonfiction fails because it opens with background instead of a hook. Readers leave before reaching the good material. Find the single anomaly or question that makes your topic interesting, lead with that, and let the background follow once you’ve earned attention.

Askers vs. Guessers

This one’s from 2010, but it was new to me. The Atlantic wrote about a concept from a 2007 MetaFilter comment by Andrea Donderi:

Ask Culture: You grow up with the expectation that it’s OK to ask for anything at all, but you gotta realize you might get no for an answer.

Guess Culture: You avoid putting a request into words unless you’re pretty sure the answer will be yes. Depends on a tight net of shared expectations. A key skill is putting out delicate feelers.

When an Asker meets a Guesser, things get awkward. The Asker doesn’t think it’s rude to request two weeks in your spare room. The Guesser hears it as presumptuous and dreads saying no. This framework explains so many social misunderstandings.

AI coding tools

A lot of my January reading was about the evolving AI coding landscape. A few highlights:

Silen Naihin wrote a comprehensive guide about switching from Cursor to Claude Code. He’d been a top 0.01% Cursor user before making the jump. His key reasons: the terminal-native workflow forces you to a higher level of abstraction, Claude models perform better in their own interface, and the cost efficiency is better. He still uses Cursor for pixel-perfect frontend work and learning, but Claude Code is now his default.

Running Claude Code from a phone: Someone set up a complex workflow to run multiple Claude Code agents in parallel from their phone. Wow!

Steve Yegge’s Software Survival 3.0 is a 20-minute read about which software will survive the AI era. SaaS companies are already feeling pressure as the “buy vs build” calculus shifts. Business departments are vibe-coding their own tools instead of renewing with niche vendors. (We’re doing this now with internal tools at Hello Gravel.)

The Ralph Wiggum Technique is an approach to setting up autonomous coding loops where AI agents work continuously with minimal human intervention. Several tools have emerged around this: ralph-claude-code, continuous-claude, and others. I tried it on one project this month and didn’t have much success, but will keep at it.

Data visualization

A Eulogy for Dark Sky, a Data Visualization Masterpiece. I was a huge fan of the Dark Sky app that Apple killed in 2023, and this article explains what made it exceptional. I’m always trying to improve the design on things I build.

Some key points:

• Context-sensitive graphics: The app contextualized what’s presented based on what’s most relevant right now
• Hyperlocal focus: Weather at specific addresses, not just city-level
• Glanceable design: Understand the shape of the weather at a glance
• Start from “now”: Always shows current moment onward (you rarely care about weather from hours ago)

Weather Spark does excellent monthly and daily graphical reports of average weather.

Tools I bookmarked

A few things I saved for later:

• Happy: Open-source mobile and web client for Claude Code
• Agent Deck: Terminal session manager for AI coding agents (tried it and didn’t like it)
• DoltHub: Version-controlled database (Git for MySQL)
• Posturr: macOS app that blurs your screen when you slouch (funny)
• shadcn/ui: Component library I was playing around with on a project
• Buttondown: Simple newsletter platform that might be useful on some projects
• comma.ai: Open-source self-driving for car models (crazy!)

Quick links