synalx

Imo, the article misses the main point of code review. It's not about finding bugs or spreading around knowledge, but about maximizing value vs maintenance costs. Code is expensive, not to produce but to maintain. Every line of code added to a codebase must be read and understood many times over its lifetime, and therefore imposes a burden on future maintainers. We review code in order to weigh its value against that high cost of ownership. High quality, maintainable code is code which maximizes that value delivered while minimizing the cost of its future maintenance.

AI is changing the game here not by increasing (or decreasing) the value of code, but by reducing its cost of ownership. When it's significantly cheaper to understand, modify, and replace code, the balance point shifts significantly. It's the definition of "maintainable" that's changing.

One thing that annoys me to no end is when a mobile app can't even load without an internet connection. There's a plague of apps that are thin veneers over a web view. I get that this is easier to deploy and manage, but I find it a really poor experience compared to a well-designed mobile app.

I have this problem even in codebases. Claude will work on something, and add detailed comments in which it extrapolates the from the design and confidently states intentions and decisions which aren't actually grounded in reality. Then, later sessions suffer when it reads back those hallucinations and treats them as canonical.

With a soldering iron and USB-CD PD trigger boards, I've successfully converted a few of this kind of thing.

Mine is obsessed with "planes". Data plane, control plane, management plane. Everything is a plane :)

I interpreted this as a visual bug with the sprites.

As a sailor, I love seeing ads or other random imagery of sailboats that's clearly photoshopped, with sails in nonsensical orientations or completely backwards with things like flags or hair blowing in the wind.

That might be challenging, I hear people are pretty short on tokens these days.

Antigravity's coding agent is worlds apart from Gemini CLI, though.

This stood out to me too - there's an underlying assumption that private entities _can_ say no to governments, but that's only true to a point. If the government decides it needs AI-powered killbots as a matter of national security, it can and will nationalize whatever entities it needs to build them.

Vibe biology?

For 3D modeling (assuming you want to do CAD), Onshape is a fantastic alternative to native applications.

In that sense, AI behaves like a human assistant you hire who happens to be incredibly susceptible to social engineering.

The problem with "the language tooling is already a build system" is that cross-language dependency chains are a thing. The moment you need a Rust or Zig file to be regenerated and recompiled when a JSON schema or .proto file is updated, you're outside what most of those language-specific toolchains can support. This is where Bazel absolutely shines.

One major disadvantage here is the lack of training data on a "new" language, even if it's more efficient. At least in the short term, this means needing to teach the LLM your language in the context window.

I've spent a good bit of time exploring this space in the context of web frameworks and templating languages. One technique that's been highly effective is starting with a _very_ minimal language with only the most basic concepts. Describe that to the LLM, ask it to solve a small scale problem (which the language is likely not yet capable of doing), and see what kinds of APIs or syntax it hallucinates. Then add that to your language, and repeat. Obviously there's room for adjustment along the way, but we've found this process is able to cut many many lines from the system prompts that are otherwise needed to explain new syntax styles to the LLM.

And a major consumer of space, because they have to accommodate bigger ports even if they're only doing USB-C pass-through.