| So the article isn't very good but the vibe coding debate is pretty interesting. This is how I'm thinking about it: in a scenario with increased opportunity and risk... You've gotta know where you stand. First question is how much is more software actually worth to you." This is one with a lot of self deception. Software development is expensive. The companies have to do lists and wishlist and road maps. They have an A/B testing system and a productivity mindset. But... If Linkedin, Salesforce or any whatnot really did have ways of producing software to make money... they would have done it already. Remaining opportunities follow a diminishing marginal value curve/cliff. Imo, software development isn't necessarily a bottleneck. So... opportunity is limited and risk is the bigger deal. The opportunity is at the upstart trying to bootstrap feature parity with Salesforce. If you have no customers yet... you can unfettter the vibe and see if it works. Imo companies need to revisit google's early days. Let a thousand flowers bloom. 20% time. If you unleash capable people and give them tokens .. That's a good way of searching for opportunities. The thousand flowers died at Google because they had reached a point where opportunities are not everywhere. The best ideas had been discovered and also... the markets big enough to move Google's dial are few. There aren't many $100bn markets. There's no way to do vibe coding safely, at scale, currently. |
A really misunderstood vibe coding task, especially in more corperate settings, is code removal and refactorings.
I think this is the the fundamental misunderstanding about agentic development: people only see it as a tool to add code.