weekend ai reads for 2026-01-23

It’s almost too easy to make new software, in fact, and that can be exhausting. One project idea would lead to another, and I was soon spending eight hours a day during my winter vacation shepherding about 15 Claude Code projects at once. That’s too much split attention for good results, but the novelty of seeing my ideas come to life was addictive.

The moment I went to post a job and saw the predetermined field choices I knew the system wasn’t flexible enough. I hadn’t designed the system properly and the AI filled in the gaps for me.

You can prompt it not to do this.

You can tell it to introduce global settings, environment variables, or config tables.

But you have to already know that it shouldn’t be hardcoded. That knowledge doesn’t come from vibe coding. It comes from experience (which I have, but wasn’t using at that moment!)

Ralph is a development methodology based on continuous AI agent loops. As Geoffrey Huntley describes it: "Ralph is a Bash loop" - a simple while true that repeatedly feeds an AI agent a prompt file, allowing it to iteratively improve its work until completion.

The technique is named after Ralph Wiggum from The Simpsons, embodying the philosophy of persistent iteration despite setbacks.

Because tools ranging from Claude Code to Lovable typically don’t require robust coding knowledge just to get to a functional app, we are witnessing the early rise of micro apps. These are apps that are extremely context-specific, address niche needs, and then “disappear when the need is no longer present,” Legand L. Burge III, a professor of computer science at Howard University, said.

Friction-maxxing is not simply a matter of reducing your screen time, or whatever. It’s the process of building up tolerance for “inconvenience” (which is usually not inconvenience at all but just the vagaries of being a person living with other people in spaces that are impossible to completely control) — and then reaching even toward enjoyment.

I don’t think it’s very likely that it’s gonna be able to write anything meaningful, or that it’s going to be making movies from whole cloth, like Tilly Norwood. That’s bull.

But there’s something AI cannot do: tell you whether something should exist.

That requires taste: judgment formed by long exposure to the best work humans have done, and by living with the consequences of decisions over time. In the AI era, personal taste is the moat.

AI assistants won’t be exempt from these dynamics just because they’re conversational rather than algorithmic; rather, the opposite is true: the degree to which there will be tier-dependent user experiences is going to be larger than on any other software category. And given that people nowadays use chatbots for everything, the degree to which these differences will impact their lives will also be greater.

To reduce the spread of low quality AI content, we’re actively building on our established systems that have been very successful in combatting spam and clickbait, and reducing the spread of low quality, repetitive content.

The result is that the current generation of LLMs is far more gullible than people. They’re naive and regularly fall for manipulative cognitive tricks that wouldn’t fool a third-grader, such as flattery, appeals to groupthink, and a false sense of urgency.

A minimal free version of this course will be offered online, simultaneous to the CMU offering, starting on 1/26 (with a two-week delay from the CMU course). By this, we mean that anyone will be able to watch lecture videos for the course, and submit (autograded) assignments (though not quizzes or midterms/final).

AI markets feel chaotic because they are not markets at all. They are proto-markets: evolutionary precursors where demand exists but selection and speciation have yet to take hold.

Kande attributed this tension not to the technology itself, but to a lack of foundational rigor. “Somehow AI moves so fast … that people forgot that the adoption of technology, you have to go to the basics,” he explained, citing the need for clean data, solid business processes, and governance. PwC is finding that the companies that are seeing benefits from AI are “putting the foundations in place.” It’s about execution, not technology, he argued, and that comes down to good management and leadership.

So, for all the K-12 Systems Administrators reading this, maybe think twice about the tools you’re paying thousands of dollars a year for. Are they the best options — not even from the students’ perspectives, but from the teachers’ perspectives?

This year, both a single reader and AI will give scores for each essay question, and if there’s a discrepancy, an additional human reader will also give a score. That’s so far saved an estimated 8,000 hours, according to Juan Espinoza, vice provost of enrollment management.

For the most part I enjoy my job. It is interesting and challenging in the right ways. Yes, there can sometimes be tedious bits to it but even those are enjoyable in a meditative way and I don’t think ridding myself of them would make me a better developer.

MCP servers are not thin wrappers around your existing API. A Good REST API is not a good MCP server. We assume just because LLMs are “smart”, they can use APIs as a human would. Thats wrong.

On Tuesday, January 13, University of Alaska Fairbanks undergraduate student Graham Granger was detained after he had been found “ripping artwork off the walls and eating it in a reported protest,” according to the UAF police department.

On Saturday, tech entrepreneur Siqi Chen released an open source plugin for Anthropic’s Claude Code AI assistant that instructs the AI model to stop writing like an AI model. Called “Humanizer,” the simple prompt plugin feeds Claude a list of 24 language and formatting patterns that Wikipedia editors have listed as chatbot giveaways.

My mom recently broke a bone in her back and she’s laid up, and on the 12th day, things weren’t getting better and it was getting worse, so she was using ChatGPT, and she asked it, “I was really doing good and this is my 12th day and it’s really bad”, and it gave her this detailed, “You used to be acute, now you’re subacute”, all these terms, and said, “But this is normal on the 12th day”. That’s BS, that’s not right. So I got on a different device and said, “On my 12th day, I felt remarkably better”, and the same system told me, “That’s normal, on the 12th day, you get remarkably better”, it is designed to tell you what you want to hear, not what you need to hear.