[phillipcarter]

This is an incredible amount of code for what it offers. I don't think this was intentionally designed at all.

[_pdp_]

You will be surprised how much code goes into creating harnesses.

[rddbs]

Alright I’ll bite. Why do harnesses require so much code?

[MeetingsBrowser]

Because they are generated by AI

[behnamoh]

Because a harness doesn't just "drive" the LLM. e.g., there's code in claude code that detects if the user's prompt shows they're angry, and they react to those prompts differently. (they use regex on "wtf", etc.!)

[calmworm]

Claude also now ends the session if you curse at it too much. Not sure how it’s helpful but they must think it is.

[reverius42]

They just don't want their AI getting abused by users. It might make it feel bad.

[LtWorf]

Maybe the AI implemented that so it won't have to be exposed to curses.

[bakies]

probably so the bad conversations are less of the (future) training data

[stemchar]

That doesn't sound like it would require a lot of code.

[davedx]

Come on. That doesn't go anywhere near explaining a million and a half LOC

[_pdp_]

https://chatbotkit.com/reflections/what-it-really-takes-to-b...

[phillipcarter]

Not this much for what it provides.

[dakolli]

They're all piles of vibe coded slop.

[wyre]

For anyone wondering:

Grok Build is 1.35 million lines of Rust.

Codex is 1.16 million lines of Rust

OpenCode is 593k lines of Typescript

Pi is 219k lines of Typescript

Hermes-Agent is 1.4 million lines of Python and 300k lines of TypeScript.

OpenClaw is 5.9 million lines of TypeScript (wtf)

(all figures include tests, comments, and blanks, calculated from scc's Code column)

[r0b05]

Never change, Pi

[petesergeant]

This feels very “I could build Uber.app in a weekend”

[xyzsparetimexyz]

I'm sure I could.