Daniel Jackson on WYSIWID: Rethinking software structure to enable LLM coding

For an archeologist that travelled the entire galaxy and prided himself on using dozens of languages, this post has a few too many typos and errors, and is at times hard to parse...

That being said, it's interesting to see professors jump on the opportunity to use LLMs and LLM assisted coding to teach! That's a breath of fresh air. In the past two years all we've heard is how screwed everyone is, how students use LLMs to cheat and so on. To see this used in classrooms is great! Kudos to the professor.

On the concept itself, I'm not convinced if it'll get enough traction. It's an interesting way of looking at the problem, but there are two things going against it - the plethora of training data in "the classical way" of software design, and ever increasing capabilities in both context usage and agentic e2e task completion.

It does, however, bring up an interesting point. With more and more LLM assisted code generation, we might see a resurgence of micro/mini services. As long as the problem you're solving for can be solved by a series of microservices, the LLMs might just be the perfect thing to build them and, importantly, maintain them. You could choose the best language for the job, the spread of responsibilities is manageable, and the scope of work should fit into the current capabilities quite nicely.

It is probably the latest in design, but, ... can the text be lightened up? Or provide a CSS toggle to use a different template? it is uncomfortable to trying to read whatever it is you wrote.