It's already like the dot-com boom, with everyone talking about how amazing AI is and how disruptive it is. How easily it can generate any media files and text with click bits.
But let me be human, humble and open, sharing my own thoughts. And putting a belief that it will bring the results.
Exploring AI models reveals their immense potential, yet they lack real-time feedback. As software developers, we can manipulate the code and create everything required for software product: server, client, infrastructure, metrics, alarms, dashboards. This ability makes us wonder why we can't use AI to instantly build startups.
Working with AI in code generation, I've realized it's complexity. Humans have a contextual 'Mind Map', unlike AI, which predicts the next token based on probability, not logic and implications. Therefore, AI struggles with understanding interconnected logic in software development.
I have a deep right feeling of using Event Driven Architecture (EDA) for AI-generated code. You're already a profound user of EDA by sending an email. You send an event (text, subject, recipients) into SMTP system, which sends you results if email was delivered or not later, without blocking you to wait. This architecture, scalable and resilient. And it can be generated and tested by AI. Why? it has a defined scope, independent testing, reduced complexity, clear feedback.
While EDA isn't a magic solution, it holds potential as a good starting point. Some view that approach as a step towards AGI, where software can self-improve. I'll leave this thought here.
This approach allows us to focus not only on outsourcing AI-generated code, but also on building powerful startups in various fields such as healthcare, renewable energy, education, entertainment, logistic in a just a day.
Right now the main idea is to build that building system that builds other software based on EDA.