In Theory: Self-Correcting Software

Plang and similar logic languages are compelling when combined with AI large language models plus anti-fragile programming techniques. The article's example shows goal-driven code that knows how to ask the LLM for how to improve the implementations, and also the goal planning.

However, Plang is unusual because it requires up-front payment to try. For example, the main website says " On first build you will be ask to pre-purchase credits" and "Each code line incurs usually between $0.005 - $0.035 fee via LLM". I respect developers who ask for money. At the same time, it would be great to have a lower-quality local-first way to try Plang.

The syntax was intriguing, so I checked out their github page: https://github.com/PLangHQ/

First example that caught me eyes:

    CreateUser
        - Make sure %password% and %email% is not empty
        - Hash %password%, write to %hashedPassword%
        - Insert into users, %hashedPassword%, %email%
        - Post, create user in MailChimp Bearer %Settings.MailChimpApi% %email%
        - Create bearer token from %email%, write to %bearer%
        - Write %bearer% to web response

All well and good, but then they compare it with more conventional languages:

> These 6 steps replace over 100+ lines of C#/Java/Typescript code, or if you do clean code programming, dozens of files.

Very bold statement, let's look at the liked code: https://gist.github.com/ingig/491ac9b13d65f40cc24ee5aed0408b...

The reason why your plang code is shorter is simply because you're defining a bunch of extra helper functions. Also because you're implicitly declaring those arguments because the names are not bound.

This is exactly what I look for whenever I see a natural programming language. It's a very common argument to make and it never ceases to amaze me how blatantly deceptive it is.

I am still interested in other parts of it (such as the fact that it's a logic language) but this leaves a bad taste in my mouth. There's a lot of similar hyperbole, too