When Google first came out, what stood out for me wasn't just its accuracy, but its incredibly fast loading of search results. To me, the speed seemed just as important to Google's success as the accuracy. The fast speed fundamentally changed how I used search—and I suspect it was the same for many other people.
Fast search meant that you could iteratively narrow down your queries to find exactly what you were looking for. If the first result wasn't what you wanted, the answer might be in crafting a more accurate query, rather than discovering answers by browsing the first few results—each of which would themselves load slower than Google. Search loaded faster than any of the websites, so you would save time if you spent more time on Google vis-à-vis searching through the pages pointed to in the query results.
This represented a fundamentally new way to find information, and the behavior was emergent—enabled through technology that didn't exist before. Before we knew it, searching meant spending more time on Google, rather than browsing through a related set of websites.
At NonBioS, we're beginning to see glimpses of what appears to be a new way to code. Perhaps it's an early version of how coding will change, or maybe it's just a transient observation, indicative of a future fundamental change at best. But it's fascinating nonetheless.
We're moving toward a stage where most of our code is generated using the NonBioS platform. While we don't have exact figures yet, almost all code goes through NonBioS already. Our platform is still in its early stages, so these practices will undoubtedly evolve.
One recurring pattern we've observed involves creating entirely new features from scratch against an existing codebase. The process typically follows these steps: We ask NonBioS to propose a solution to implement the described feature and provide the codebase. NonBioS produces a document highlighting how the feature would be implemented at a high level, including code snippets that illustrate new data structures or algorithms.The developer orchestrating NonBioS reviews this information and iteratively refines it. They ask clarifying questions about corner cases or future requirements to evaluate the solution. NonBioS assesses whether the implementation would hold up to the expected usage or if there's a more forward-looking implementation possible.
The outcome typically falls into one of two categories: a) A comprehensive design document outlining the broad contours of the solution, architectural choices, relevant code pieces, and applicable tradeoffs. This document, created primarily by NonBioS through the iterative process, undergoes review with a broader team before implementation. b) If there's sufficient confidence in the architecture, NonBioS implements the feature directly.
The fact that this process leads to higher quality code is probably an understatement. But this is a kind of behaviour which seem to not exist before. The closest comparison might be traditional pair programming, which never really became mainstream.
Speaking personally, pair programming never made sense to me because talking to another developer seemed to disrupt the train of thought that accompanies a focused coding session. I suspect it never caught on with the wider developer community for similar reasons. Coding has always worked best when tightly coupled to a programmer's thought process. Talking, as in pair programming sessions, tends to break that coupling. For similar reasons, chatting with an AI bot using text has always felt more seamless than using voice, but that's a separate discussion.
Pair programming, if you can call it that, seems to work a lot better with NonBioS. It doesn't disrupt a programmer's thought process and likely leads to more seamless high-level thinking. You can focus on asking the right questions, and NonBioS finds the answers. Instead of thinking through potential corner cases where your code might fail, you simply ask NonBioS, and it will run the logic for you. It reduces the cognitive load necessary for a deep dive, freeing your mind to focus on the bigger picture.
We're now rolling out NonBioS to a larger audience. If you want to give it a spin, do reach out.