pollinations

Maybe the skill could have references to the code. Like if everything else fails it can look at the implementation.

Intuitively it feels like if you need to look at the implementation to understand the library then the library is probably not well documented/structured.

I think the ability to look into the code should exist but shouldn't be necessary for the majority of use cases

They were just giving that as an example that Zed's inline suggestions aren't very good for basic tasks. There are hundreds of othersmall tasks like this that can't be handled by the language server.

Plus executable.xode snippets. I think their actual source code doesn't use context. But feels like function calling packaged.

Whether you believe the the article or not, the point you posted seems orthogonal to what google is saying.

They're not claiming anything about the quality of AI summaries. They are analyzing how traffic to external sites has been affected.

   # Configure NF4 quantization
   quant_config = PipelineQuantizationConfig(
       quant_backend="bitsandbytes_4bit",
       quant_kwargs={"load_in_4bit": True, "bnb_4bit_quant_type": "nf4", "bnb_4bit_compute_dtype": torch.bfloat16},
       components_to_quantize=["transformer", "text_encoder"],
   )

   # Load the pipeline with NF4 quantization
   pipe = DiffusionPipeline.from_pretrained(
       model_name,
       quantization_config=quant_config,
       torch_dtype=torch.bfloat16,
       use_safetensors=True,
       low_cpu_mem_usage=True
   ).to(device)

update: doesn't work well. this approach seems to be recommended: https://github.com/QwenLM/Qwen-Image/pull/6/files

I'd say commit a comprehensive testing system with the prompts.

Prompts are in a sense what higher level programming languages were to assembly. Sure there is a crucial difference which is reproducibility. I could try and write down my thoughts why I think in the long run it won't be so problematic. I could be wrong of course.

I run https://pollinations.ai which servers over 4 million monthly active users quite reliably. It is mostly coded with AI. Since about a year there was no significant human commit. You can check the codebase. It's messy but not more messy than my codebases were pre-LLMs.

I think prompts + tests in code will be the medium-term solution. Humans will be spending more time testing different architecture ideas and be involved in reviewing and larger changes that involve significant changes to the tests.

This reads pretty definitively. If LLMs are intelligently thinking programs is being actively debated in cognitive science and AI research.

Why isn't it debuggable?

I don't think individual examples make sense to solve these kinds of discussions as for me it can vary easily by 6x thinking with exactly the same input and parameters.

I was recently trying to write a relatively simple htmx service with Claude. I was surprised at how much worse it was when it's not React.