Learning an unfamiliar aspect and doing it be hand will have the same issues. If you're new to Terraform, you are new to Terraform, and are probably going to even insert more footguns than the AI.

At least when the AI does it you can review it.

No, you can not. Without understanding the technology, at best you can "vibe-review" it, and determine that it "kinda sorta looks like it's doing what it's supposed to do, maybe?".

> Learning an unfamiliar aspect and doing it be hand will have the same issues. If you're new to Terraform, you are new to Terraform

Which is why you spend time upfront becoming familiar with whatever it is you need to implement. Otherwise it’s just programming by coincidence [1], which is how amateurs write code.

> and are probably going to even insert more footguns than the AI.

Very unlikely. If I spend time understanding a domain then I tend to make fewer errors when working within that domain.

> At least when the AI does it you can review it.

You can’t review something you don’t understand.

[1] https://dev.to/decoeur_/programming-by-coincidence-dont-do-i...

> Learning an unfamiliar aspect and doing it be hand will have the same issues.

I don't think so. We gain proficiency by doing, not by reading.

If all you are doing is reading, you are not gaining much.

It's a security hole. Rust doesn't prevent you from writing unsafe code that reads it. The bug wasn't that it could be read by a well conforming language, it was that it was handed off uninitialized to use space at all.

MCP servers are really just skills paired with python scripts, it's not really that different, MCP just lets you package them together for distribution.

But then those work only locally - not in the web ui’s, unless you make it a remote MCP, and then it’s back to being something somewhat different.

Skills also have a nicer way of working with the context, by default (and in the main web uis), with their overview-driven lazy loading.

Yes. I find these very useful for enforcing e.g. skills like debugging, committing code, make prs, responding to pr feedback from ai review agents, etc. without constantly polluting the context window.

So when it's time to commit, make sure you run these checks, write a good commit message, etc.

Debugging is especially useful since AI agents can often go off the rails and go into loops rewriting code - so it's in a skill I can push for "read the log messages. Inserting some more useful debug assertions to isolate the failure. Write some more unit tests that are more specific." Etc.

And for more complex linters I find that it can be easy to get the LLM to write most of it itself!!!

I do a lot of LLM work in rust, I find the type system is a huge defense against errors and hallucinations vs JavaScript or even Typescript.

https://xkcd.com/793/ captures the stereotype well.

Especially because those annoying dilettante know-it-all physicists are often right.

Something like Shrimp is useful for at least coordinating different Claude subagents.

you just... file for bankruptcy like any other person or corporation?

Yeah but when you come to bankruptcy court with significantly more assets than debt, they aren't going to let you sell the business for pennies.

I'm asking how you would believe this vehicle would go broke, which is the usual reason to go to bankruptcy.

Corporate bankruptcy happens for a lot of reasons other than being "broke". Chapter 11 is a court-supervised way of restructuring your debt. This has a lot of utility in many situations other than not being able to pay.

It'll go before a judge and creditors would be able to object, so if it's just a ploy to get rid of debt you can be certain it'd be contested.

Meta may have lots of assets, but the LLC may not. The ability to have one wholly owned LLC go bankrupt by itself is one of the main reasons shell corporations exist.

I've seen a lot of uses for SAT solvers, but what do you use them for in data science? I can't find many references to people using them in that context.

Root causing from symptoms is one case where SAT or their ML analogue -- graphical models are quite useful.