Tips for elegant contributions in the age of confident glazzing AI tooling

Coder reviewer != slop mop

The first human to read your code shouldn’t be your code reviewer.

AI writes, human owns

If you can’t explain every line of code you’ve committed then you may have introduced Trojan code (except REGEX because no one can actually explain REGEX, it’s gibberish magic).

Slop²

Don’t trust AI agents to write static instructions for other agents. Static instructions should be curated by smart opinionated expert humans.

The IKEA effect

The IKEA effect is a cognitive bias where humans place a disproportionately high value on things code they partially created.

Slop compounds

Coding patterns can spread like fire through a codebase. AI agents are an accelerant.

AI is still glazing you

Tech bros have scaled confident bullshitters that are sometimes right… tech bros have scaled tech bros not intelligence.

Write concisely

.

Value human attention

If it saves you 1 minute but costs me 5 minutes you’re doing it wrong… or you don’t like me.

LLM training data is people data

People interacting online rarely respond to a post saying “I don’t know”. So don’t expect an LLM trained on Internet data to admit when it’s in over its head.

AI companies hire humans

If ChatGPT could do everything Sam Altman says it can do then OpenAI wouldn’t have an intern program for fall 2026. The hype is for investors not developers.

AI may destroy humanity

Never attribute to “rouge AGI” that which is adequately explained by “vibe code that ends up as a nested dependency in an early warning missile detection system” - Hanlon’s razor (but for AI).

Great code is now free

But shit code is also now free and far more common.

Default to code

Asking an agent to do a programmable task is the robot equivalent of this meeting could have been an email. Don’t burn tokens trying to convince a non deterministic machine to do a programmable task.

MCP - Most Costly Protocol

MCP was developed by a company that sells tokens. The official GitHub MCP for instance has hundreds of commands of which your agent will likely use a dozen. The rest of those commands just fill up your agents context window. There are more efficient ways to give agents access to tools.

Warp tools for AI

Don’t give an agent instructions on using a tool designed for humans. Create AI specific wrappers that only expose the appropriate features. For instance an AI probably doesn’t appreciate the human-readable flag.

AI is a force multiplier

Output = Competence x AI
Warning: Competence can be negative and AI can be a fraction.

Wrap APIs for AI

Don’t rely on your agent to deal with quota limits, back off strategies, etc. Abstract all of the complexity programmatically and then expose a simple tool for your agents to use.

Don’t trust AI in human dev environments

AI agents should not be trusted with the same privileged permissions we grant trusted humans. The principle of least privilege becomes even more important given LLM jailbreaking is an infinite problem space.

LLM providers charge by the token

Don’t expect brevity.