miguel_martin

Everything is obvious and "not surprising" in hindsight...

and can be shared. I know what Bento is. which from what I can is what Hex provides. I left before agents was integrated into Bento (if it was).

Is this basically a public version of Meta’s bento notebooks?

I'd be interested in seeing the inverse, i.e. a font that only AI can read

To be fair to the commenter, you didn't provide any details in the example you provided. For all we (the readers) know, Visual Block mode would have worked in the situation you are describing. I have to assume not, but it's not clear from the article.

Vim macros are not that hard to write, but I do agree with you, visual feedback is better. It can be annoying to have to re-record a macro, that's why I tend to use :s (search & replace) or Visual Block mode over using a macro (most of the time I don't need a macro).

I understand you state vim is "just an example", but you use this example as the main backbone of your article. Add more detail. You're arguing with vague anecdotes which requires the reader to read between the lines.

An aside: visual feedback, multi-cursor support and sane defaults is likely what led to new modal editors being created, such as Helix and Kakoune.

fwiw, I think a lot of people just clone odin and build it via `./build_odin.sh release` on MacOS (at least, I do). The compiler builds in ~20s

I disagree.

1. Constructing a "deterministic algorithmic approaches with complex pipelines" is an ML approach. You're simply changing how you optimize (e.g. gradient descent with human constructed rules) and what you are optimizing (i.e. the model from a statistical one to a set of rules similar to a decision tree)

2. "Models can be intentionally poisoned or tricked" this is adversarial examples. Your deterministic and complex pipeline will have attack vectors, but just of a different distribution compared to an LLM (or neural net in general). Adversarial examples are likely unavoidable, you will always have a set of inputs that will cause your model to mis-classify examples. You can aim to minimize the size of this distribution/set, but for language: the set of possible inputs is so large that you will never fully be able to train or test on them all, and thus you will always have a back & forth between finding new attack vectors vs. defending against them: "deterministic" or not.

To expand on 1:

How do you construct a complex pipeline? Hopefully, by following roughly standard ML principles.

That is, you have a train set that you observe and find patterns/rules in. Then you iteratively construct your complex pipeline until you've minimized the error for a train set. Hopefully after this initial version is constructed you evaluate it on your (independent) val set. Then you iteratively improve your complex pipeline until your validation numbers improve. In the end, since you've optimized a val set, you need to use a third independent test set to ensure that you haven't overfit to your val&train sets. This is standard ML practice.

In other words, this process is what an "ML approach" is, just manually performed by a human possibly using some data analysis. Again, you've just replaced the optimization process (e.g. from gradient descent) and the underlying ML model (e.g. an LLM with differentiable parameters) with a more "deterministic approach" similar to a decision tree.

Yes you could automate this process to construct the rules and chain them, in which case your process and your complex pipeline will likely look similar to a decision tree (e.g. xgboost), but you're simply closer to the thing you think you are trying to avoid.

It’s unfortunate that they didn’t eval using subagents/orchestration for such a complex set of tasks (from what I can tell), e.g. analyze program to produce initial spec -> code -> review and rinse&repeat with each of those steps being a separate subagent allocated

I would be interested to see if there’s a significant quantifiable difference.