Are you building something?
If yes, you need to check this out. I just shipped 100 top platforms and directories to get your startup listed to.
Any feedback will be appreciated, and if I missed any important one, do let me know
Hi,
I just shipped Ask AI functionality on all the Aura++ project's pages. Now, you can get info about any project directly from LLMs improving the project's understanding and engagement on the site overall.
Any feedback, will be appreciated.
One thing I’m curious about is where the real bottleneck ended up: generating optimization candidates or validating that they were actually improvements.
Validation is trivial, I simply used BenchmarkDotNet test and had AI flip git between latest HEAD state and current state.
Optimization is where the real work was - but also fun because I would just think of something, and AI would make it happen. It also had great ideas of its own. Feedback mechanism is crucial.
Yeah, that's ok.
But if I am completely unaware about the code, I don't know what to search in cscope, or anything about the flow of the code. How do we manage in that situation?
Do you have a specific task you need to do in this code? If so, search for something related to whatever that is - text from the UX, data, fieldnames, etc. Once you find one spot in the code related to your task, follow the code back and forth to figure out that feature.
And if you have no task you are aiming at, find the main entry point of the code and start tracing where it goes.
But either way, I would not actually start with the code. The code glues together the data and the UX, so go figure those out first. Use the app. Browse the data schema. Figure out some connections. Once you understand that "datum A" is used in "UX widget B", the code just becomes a middle point to step through to connect them. And stepping through it once for one data field will likely teach you the patterns used in the codebase, so you can apply that learning to everything else.
And don't forget to read documentation. It isn't always kept up to date, but something is better than nothing.
If the underlying workflow is noisy, ambiguous, or overloaded with coordination overhead, faster generation just produces more low-context output to review and reconcile.
It’s interesting how AI monetization often favors firms with existing workflow embedding rather than pure model innovation. Distribution and operational integration seem to matter more than raw model capability.
This feels even more relevant with AI-assisted coding. Generating code is getting cheaper; the bottleneck is increasingly understanding and validating what already exists.
I think that for reading and understanding code you first need to have experience writing it and making mistakes. Are we bound to have a generation of coders that have never written any code at all?
True. Reading code needs some experience in how coding actually works. I feel confident enough to read code in programming languages I've never coded in as long as they are imperative because the the pattern will be similar so that an abstract static analysis is possible.
I don't mind the usage of AI agents but review the code, understand and improve it. Every other prompt I don't even need to run the code to see that it won't work and I'm tired of the "oh sorry, you're right! Here's a fixed version" excuses.
I've already noticed, how AI usage makes people not think anymore. There's a problem? Ask the AI. Run whatever the AI returns. It declares solutions as dangerous and explains why but I saw enough people solving their problems with --force until it caused real damage or failed.
And the explanation is filling with indignation: "Just use --force" - "why?" - "It does not work otherwise" - "why?" - "I don't know". It was a npm telling, it cannot resolve the dependency tree anymore.
One thing that often gets overlooked is how much failure and constraint shape better leadership. It seems like the NeXT years gave Jobs the space to rethink product focus in a way that likely wouldn’t have happened if Apple had kept succeeding uninterrupted.
reply