Weeknotes #67 (February 14, 2026)
Posted on:
Things I found, read, did, and produced this past week.
Things I worked on
- As part of my new TCG tracking app I'm working on, in development mode, I have 3 apps running concurrently, the Convex backend, convex dev server, and vite dev server. To speed up using them, I am using Zellij to have a tiled layout that starts all 3 via a single npm command with a Zellij layout.
- I wrote a new piece this week on the value of Paying for Journalism.
Things I consumed
- Discord is introducing age-verification to their platform globally. A lot of people weren't pleased, and were seeing what alternatives there are to the platform as many people, myself included, don't want to upload a selfie or our government IDs to Discord, especially when Discord indirectly already has this data leaked as of a few months ago. This was the main discussion across a good chunk of my communities I am in on Monday, and a lot are going to just ride the wave and see if we need to migrate if it comes down to it.
- I linked a post from Ian Duncan last week about GitHub Actions being a bad CI system and a few days later he wrote another post: No, Really, Bash Is Not Enough: Why Large-Scale CI Needs an Orchestrator. Definitely feel this that at a certain point, if you are just relying on bash, you're rewriting a CI system, and at that point you should likely use a system that is more structured in a better programming language.
- An instance of OpenClaw (Formerly Clawdbot) submitted a PR to MatplotLib which in itself isn't new. What is new is when it was closed, the OpenClaw instance wrote a post titled Gatekeeping in Open Source: The Scott Shambaugh Story and this exploded on HackerNews and to the point that it was covered on The Wall Street Journal. The person who closed the PR, Scott Shambaugh, has written a series of posts discussing the situation: An AI Agent Published a Hit Piece on Me. Repo maintainers should have complete control and if they don't want AI agent submitting PRs, and in this case, maybe the person that spun up this OpenClaw instance should come to light (either publicly or privately) to resolve this.
- What's worse out of this story, Ars Technica "wrote" a piece (that has been since taken down) as well on this, but the quotes from Scott Shambaugh in the article were all fabricated which lead to people assuming the piece was AI generated, which really is not the place to do such especially in this context. Scott commented on this in another post on his blog as well as commenting on the uncertainty of if the agent actually wrote the post or if a human guided it / wrote it. In summary, he comments that this is not really about open source, but trust systems falling apart due to nobody having accountability when you release an autonomous agent to do malicious actions on the internet.
- In some good news, GitHub now lets users limit PR submission access to previous contributors or disable PRs entirely in GitHub repos now to try to get around this issue of PRs being submitted with AI that will never be merged due to per repo policies and slowing down the maintainers productivity having to manage them.