3/
For example —
You have a large, complex code base you are unfamiliar with.
You get the LLM to look at it. It gives you a technical summary of it. And, you can ask it questions about the code base.
Note that for this activity, the AI coding tool hasn't actually done any coding — no Vibe Coding, or any other (non- Vibe Coding) coding.
…
2/
The first thing I did is — I started by paying more attention to how other people are using AI coding tools.
…
One thing I noticed is —
Some users are Vibe Coding —
https://mastodon.social/@reiver/115639766305963429
But, not everyone who uses AI coding tools is Vibe Coding!
Especially the software engineers I looked at who are using AI coding tools. They don't seem to be Vibe Coding — they seem to be using them differently.
…
3/
For example —
You have a large, complex code base you are unfamiliar with.
You get the LLM to look at it. It gives you a technical summary of it. And, you can ask it questions about the code base.
Note that for this activity, the AI coding tool hasn't actually done any coding — no Vibe Coding, or any other (non- Vibe Coding) coding.
…
4/
I tried this myself on a large, complex code base I was unfamiliar with — to try to get that first-hand experience so that I could have an informed opinion. I actually found that use-case useful.
It sped up hours or days of tedious work.
…
I can see how software-engineers would find this particular activity useful — to ask an LLM questions about large, complex, unfamiliar code base.
5/
Another interesting thing is — MCP (Model Context Protocol).
Where software-engineers are creating a type of server. Sort of like a web application server — but it speaks MCP rather than HTTP.
(MCP is just JSON-RPC 2.0 with certain methods communicated over STDIN and STDOUT.)
And, the MCP developer uses an LLM as a type of front-end for their MCP server.
…
That use is NOT Vibe Coding either.
@reiver 💯 I absolutely agree. It annoys when people learn you're using an AI coding assistant and label it automatically as "vibe coding". Totally different. I know what I'm doing, I give the AI lots of design and best practice guidance, and override it when it overruns its context and starts making bad decisions. As an experienced tech team lead, I'd do the same thing with junior developers.
