There’s nothing preventing closed models from doing the same thing

I can’t really speak directly to your question because I don’t have LLMs write react, but I get great results from svelte.

At this point I don’t think it’s about training examples, once there is sufficient training data it becomes a problem of verifiability, ie how easily can the model check its work for success.

Because they’ve essentially exhausted pre training scaling and are looking to post training to expand capabilities, which is really just optimization via reinforcement learning against specific tasks aka bench maxing.

I generally use sonnet 5 for most coding tasks, a lot of coding is really pretty straightforward.

> the cost to build a modern agentic model is in the 10 figure range and growing

Source? The proliferation of labs building competent models would seem to suggest the opposite.

seems self evident if you read the new. You can Google it yourself, but here's the results from my googling - and this is just for the hardware. Double that to add personnel and corporate infrastructure

"To build or purchase the physical hardware required to store tens of petabytes of data and train a State-of-the-Art (SOTA) frontier AI model, you are looking at a capital expenditure (CapEx) ranging from $320 million to well over $1 billion."

Strongly disagree with your premise.

Programming languages are what you build creative and interesting things from, building in safety into your tools is a good thing.

I wouldn’t want a nail gun that shoots in the wrong direction when you apply the incorrect amount of pressure. Software engineering over indexes on an obsession with the tools themselves over the things you make with them.

Claude comments look like this

//add returns the sum of x and y

//per section 2.1 of addition-implementation-plan.md sum is designed as the seam for user addition interfaces.

//previously sum added numbers, now it adds numbers

def add(x, y):

    return x + y

Not subtraction, not division.

just reading this comments made me despise the fact that I have to read this shit again tomorrow

It's really time to move to OpenAI...

I’m trying to pitch my boss on something like bedrock/foundry/vertex where we can just swap the model and still have it serverless over an api.

Digital ocean particularly looks promising as well.

The corollary here is to have Claude write tests to enforce this. The only thing it is consistently responsive to is test failures.

I’ve been landing on go for a similar reason, which was the realization that LLMs are in many ways just massive cargo culting machines. People will call it “quality training data”, but really LLMs will just reflect the norms and behaviors of whatever ecosystem they’re using.

I like to think go’s lack of magic and standard library will lead to lower maintenance burden over time, but that’s mostly just vibes so far.

As someone who has mostly written Rust and Python over the last 10 years... yeah. Go is actually a really solid choice for Agents IME