Onto retweeted
the smartest engineers I know who use AI to build software at scale are all thinking about ontologies
we're now building 10000x faster but systems are not 10000x more correct. When you have probabilistic models touching your system in a world where code is generated at a much faster rate than humans can (or care to) check, you stop knowing whether your system is correct. Some people's answer is human review, some people say its formal verification, and some people say the answer is actually one level down at the ontology layer: a shared, formal understanding of what things mean in your world that enforces what’s allowed to be true.
more writing on this soon but if you have thoughts on this lets yap
Onto retweeted
If the ideal ratio is 1:1 and every team in an org is a single engineer and pm, ownership and understanding of the ontology and it's ability to pass hands when someone leaves becomes the single most important thing.
As others keep adding layers of abstraction on top of probabilistic machines hoping it can guess what you mean we are taking a different approach.
Onto puts you closer to intent, closer to true grounding in human defined processes, instead of AI as a translation layer, we use it as a discovery layer to refine and extend the words you define. The future of process is multiplayer, auditable, and deterministic
Onto retweeted
Correctness of a system is not a state- it’s an ongoing relationship between requirements, the executable system, and the world. As requirements and the world change, the system stays correct only if it changes with them. Correctness isn’t a property you achieve - it’s a relationship you maintain. This is an important part of what we are trying to solve with @ontoapp
Onto retweeted
Stories Are All You Need - Intent Driven Development without the last mile
blog.onto.space/Stories-Are-…
