I wrote a blog post about why I built turri: https://t.co/qLmdqXdaW6
I stopped knowing my own software.
AI coding agents started producing so much code that keeping up with reviewing it became impossible. Eventually, I stopped trying. And somewhere along the way, I stopped really knowing the software I was building and lost my sense of engineering.
That feeling led me to build turri - the first open source coding agent with the power of regenerative software.
The idea is simple: let AI handle the implementation, but keep humans in control of the intent, decisions, and understanding behind the software.
Code is becoming disposable. Understanding shouldn't be.
I wrote a blog post about why I built turri: https://t.co/qLmdqXdaW6
I stopped knowing my own software.
AI coding agents started producing so much code that keeping up with reviewing it became impossible. Eventually, I stopped trying. And somewhere along the way, I stopped really knowing the software I was building and lost my sense of engineering.
That feeling led me to build turri - the first open source coding agent with the power of regenerative software.
The idea is simple: let AI handle the implementation, but keep humans in control of the intent, decisions, and understanding behind the software.
Code is becoming disposable. Understanding shouldn't be.
Nothing really - at least not anything that hasnt been yet adopted by the industry.
What is kinda special is that tasks are broken down into small reasonable chunks (through using ***plain) so that each agent focuses on one task and build on top of what others have built.
And agents can always regenerate the code due to the fact that intent is stored in ***plain. But as far as coding agents go and when it comes to them coding, they adopt the SOTA approaches to that.
@DMAE977 Glad you like it. I akin it to moving away from assembly and into higher level programming languages. Sort of another layer of abstraction higher.
@TheEaglegon This is built directly into turri through a lot of testing. And the user / developer can also specify explicit testing scenarios which validate that the code does what ***plain says it should :)
@iamluoyanjun Yes. I think this is a crucial next step. And understand the software that is being maintained is crucial and that will never go away with AI.
@jacewbuilds I went away from code completely and into ***plain, which is a way of describing durable system knowledge that survives every regeneration.
Sort of how we went from machine code into higher level programming languages.
Using coding agents I noticed that I lost control over the software that I was building. I dont believe the human mind was designed to ingest thousands of lines of code every few minutes. And so I was accountable for something that I didn't really know how it works. That was a scary place for me to be in.
For that reason I built turri, the first open source coding agent with the power of regenerative software and today I am happy to announce that it is officially public! The main idea of regenerative software is that the real value of a software system lies in its behavior, architecture, interfaces, and purpose, not in the code itself - and turri has that built in through ***plain, the language guardrails that we designed at ∗codeplain.
https://t.co/JfKihnD56c
https://t.co/EA72thnz92