> They didn’t even get to that point. Because my friend didn’t have the vocabulary to ask the AI to write code.
What harness did you use?
In e.g. claude, there are two modes:
1. Spit out code
2. Draft a plan, ask questions, GOTO 1
You literally have to go out of your way to get it NOT to write code. I keep mine on a tight-ish leash because it modify code way too happily even when there's no intention or instruction to do so
"rawdogging" - as the youth says - weights on local comodore cluster?
They are either using some generic web frontend, ala chatgpt, some local app like claude desktop, or programming app like cursor.
Each of those will detect that you are "building an app" and will spit out code in one form or another. You have to try really hard and be very explicit that you want the output in some other format than code.
How would an absolute newbie know what a harness is? What an IDE is? That code is represented in a set of files? That some of those files are actually metadata? What metadata is? What GOTO means? That you're suggesting a modified SDLC?
None of this is related to intelligence, desire or potential. It's about context and experience. The vast majority of people use their computers as consumption devices, like a TV. If I asked you to "make a movie" where would you start?
I let her do all of it without influencing her choices. She choose the web interface of ChatGPT because she already had an account and that's the tool she was familiar with.
While I agree with you it is not an optimal choice, web chatgpt can solve the problem. I just asked it now to do it (in my own words) and it spit out the code in one go.
I'm pretty sure the idea of the question "what harness did she use" was to ask the expert recounting this anecdote to include more detail, not suggest that he relay the question to his example friend.
That's the terminology they have chosen because they believe they are "in control". It's a purely psychological thing, not different from calling different DB servers "master" and "slave".
When I explain things to people, I say: the LLM is a big mouth, and it is a big mouth without hands. It can only talk. The agent is what gives some hands to the LLM, and then it can do some work.
Harness refers to the tooling that allows you to interact with an LLM. Web chat interface is a harness, CLI coding tools (claude code, codex, opencode, pi, etc.) are harnesses, agent systems (openclaw, hermes, etc.) are harnesses. They present different capabilities to the underlying LLM. Codex, for example, is more likely to write code if you say "I want to build an app" than ChatGPT on web will.
Personally I even refuse to learn what it is in the AI context. Feels like a waste of energy given new ~~~best ways to use YOUR tokens~~~ are discovered every month and things you learn now will be useless next month.
I am doing perfectly fine with the web UI version of these tools... They seem to also not make tokens dissappear as fast as using claude cli tool to automate implementations. Makes my work day more tolerable as well as I actually have something to do over waiting until some implementation can be read through...
What harness did you use?
In e.g. claude, there are two modes:
1. Spit out code 2. Draft a plan, ask questions, GOTO 1
You literally have to go out of your way to get it NOT to write code. I keep mine on a tight-ish leash because it modify code way too happily even when there's no intention or instruction to do so