loveyourclanker.orgbeta

Full Vibing with Agent

If you don't understand code at all or for whatever reason, you don't want to put hands on code for this project, you can use your coding agent to do everything and just not touch the code at all.

The flow

Software Engineer
Agent
01
The software engineer starts by prompting a system endpoint. "A mobile app that counts things", "A website where horses can meet other horses"
02
The Agent evaluates app type, and prompts for Gaps. "should the app run on both IOS and Android?". "Should we include geldings or only breeding horses?"
03
The software engineer answers questions.
04
The agent builds the system and supplies relevant commands to run the system in some sort of preview mode for review.
05
The software engineer tests the system and prompts bugs or changes
  1. Repeat untill done
  2. The interaction ends.

The ratings, in words

Speed

High

Control

Low

You see the code but you might not understand it and you definitely won't know where to look when it breaks. If you need to do real work on it or it gets stuck at 80% complete, you will discover a huge codebase you have had no hand in creating #fun.

Quality

Low

Depends on how well you prompt. Improves if you use context engineering to prompt DRY principles, architecture choices, , multi agent cross checking systems etc etc... but this is entirely up to you.

Safety

Low to Medium

Again, depends on how well you prompt.

Brainrot

High

Near complete handoff. You are not learning anything.

Token use

Highest

You are asking the agent to do everything and if you want to save tokens, you need to be smart with prompting, session length, memory etc otherwise it's gonna blow out.

Best for

  • Prototyping
  • Hobby projects that you may want to pick up properly later

Tools

Any modern harness + frontier model. Claude Code + Opus or Fable, Cursor, Antigravity etc.....