Vibecoding

Legend says some developers are 10x more productive than average.

The AI ceiling is the point where AI can't help you anymore and you have to actually think. It handles the boilerplate, the CRUD, the standard patterns.

AI gaslighting is when the AI looks you dead in the eyes (metaphorically) and insists the broken code it just wrote is totally correct.

An AI hangover is the morning-after realization when you look at the code you vibe-coded yesterday and have absolutely no idea how it works. It runs.

AI IDE burnout is exhaustion from using Cursor, Windsurf, or Claude Code all day every day. Your brain is constantly evaluating AI suggestions.

AI pair programming is pair programming where your partner is an AI.

AI slop is low-quality AI-generated content that's been published without human review or editing.

AI sycophancy is when the AI agrees with everything you say instead of telling you you're wrong.

An AI whisperer is that one person on the team who always gets amazing results from AI tools while everyone else gets garbage.

Agent burnout is the mental fatigue of running multiple AI agents and supervising them all day.

Agent lore is the accumulated tribal knowledge about how to get the best behavior out of specific AI agents.

Agentic coding is when AI doesn't just suggest code — it autonomously writes, runs, tests, debugs, and iterates on code by itself.

Auto mode is Claude Code's permission classifier — the middle ground between confirm-everything and YOLO mode.

The nuclear power plant committee approved the reactor design in 2 minutes — no one understood it well enough to debate. The bike shed for employees?

Blameless culture means when something breaks, you ask 'why did the SYSTEM allow this?' instead of 'who did this?

A bot PR is a pull request opened by an automated system — Dependabot, Renovate, an AI agent — instead of a human. Some are essential (security patches).

Bus factor = 1 means one person leaving (or getting hit by a bus) destroys the project. They're the only one who knows how the payment system works.

CLAUDE.md is a special file you put in your project to give Claude Code persistent instructions — your project's personality file.

Claude Artifacts are the interactive code previews Claude can generate in conversation — full React components, SVG graphics, HTML pages, and more, rendere...

Code golf is a programming sport where you solve problems using the fewest characters possible. A FizzBuzz in readable Python is 10 lines.

A code monkey is a developer who just writes code without much creative input — following specs exactly, no questions asked.

Code Review Bingo is the game you play during code reviews where you spot classic patterns: 'TODO: fix this later' (it never gets fixed), commented-out cod...

Commit farming is gaming your GitHub contribution graph by making lots of tiny, meaningless commits. Rename a variable, commit. Fix a typo, commit.

Conference Driven Development is rewriting your entire tech stack every time someone comes back from a tech conference hyped about a new tool.

Context collapse is when the AI has been helping you perfectly for an hour and then suddenly forgets everything.

Context stuffing is cramming your entire codebase, README, package.

A context window is how much text an AI can 'see' at once — its working memory.

Copy-paste driven development is building your app by stitching together code snippets from AI chats, Stack Overflow, and blog posts like a digital Franken...

Copypasta is copy-pasted code that gets reused everywhere — the developer's comfort food.

Cowboy coding is writing code with no rules, no process, no plan — just you, your editor, and the Wild West. No code reviews, no tests, no branches.

Crunch time is that intense period before a deadline where normal work hours go out the window and the team enters survival mode.

Cursor brain is when you've used AI autocomplete so much that you've forgotten basic syntax. You can't write a for loop without Tab.

Cursor Rules (formerly .cursorrules) are instructions you put in your project to tell Cursor IDE how to behave — like a CLAUDE.md but for Cursor.

Cursor Tab is the AI autocomplete in Cursor IDE that predicts what you're about to type and lets you accept it with Tab.

Dangerously skip permissions is the nuclear option — bypassing all AI confirmation gates. Named 'dangerously' because the devs meant it.

A death march is a project where the deadline is impossible, the scope keeps growing, and everyone knows it's going to fail — but they keep marching anyway...

Dependabot fatigue is opening your PR list and seeing 47 dependency-update PRs from a bot that never sleeps. You start reflexively merging without reading.

Dogfooding means using your own product — 'eating your own dog food.' If you build a project management tool, use it to manage the project of building it.

A drive-by fix is when someone spots a bug, quickly patches it, and moves on without telling anyone — no ticket, no PR description, no context.

A flame war is an aggressive online argument that generates way more heat than light — tabs vs spaces, Vim vs Emacs, React vs Vue, semicolons vs no semicol...

A Friday deploy is deploying code to production on a Friday afternoon — universally considered a terrible idea.

A ghost engineer is someone on the team whose contributions are mysterious — they're on standups, they get paid, but nobody can point to a line of code the...

When an AI confidently makes something up — like citing a library that doesn't exist or generating code that calls a function that was never written.

That feeling when you're convinced everyone around you is a real developer and you're just faking it. You shipped a feature? Lucky. You fixed a bug? Fluke.

A knowledge silo is when critical information is trapped with one person or team and nobody else has access to it.

LGTM is the two-word code review approval that can mean anything from 'I carefully reviewed every line and this is perfect' to 'I glanced at the title and...

MCP is like USB for AI — a universal standard for connecting AI models to external tools and data.

Model hopping is when you ask GPT-4 a question, don't like the answer, so you try Claude, don't like that either, try Gemini, then go back to GPT-4 and acc...

Natural language programming is writing code by describing what you want in plain English (or any human language).

A nit (short for nitpick) is a tiny, non-blocking code review comment about something that doesn't really matter but someone felt strongly enough to mentio...

Not Invented Here (NIH) syndrome is refusing to use existing solutions because you didn't build them yourself.

One-shot prompting is giving the AI a single example of what you want, then asking it to do the same pattern for new input.

A PR factory is when you (or your AI) crank out pull requests faster than anyone can review them. Quantity over quality.

PR fatigue is when your review queue has 40 open PRs and you've stopped actually reading them. You just scroll, click approve, and move on.

Pair programming is two people at one computer — one types (driver), one thinks strategically (navigator).

Permission mode controls whether your AI agent confirms before acting, auto-approves low-risk moves, or skips all gates entirely.

Plan mode is AI-first read-only planning — the AI reads your codebase, outlines what it'll do, and waits for you to say go.

Prompt chaining is breaking a big task into smaller prompts and feeding the output of one into the next.

Prompt debt is like technical debt, but for your AI prompts. You wrote a quick-and-dirty prompt that works, so you kept it. Then another. Then another.

Prompt engineering is the art of talking to AI so it actually does what you want.

Prompt fatigue is when you've been talking to AI all day and you just can't think of another way to phrase your request.

Prompt golf is the art of getting the AI to do exactly what you want with the fewest possible words. Like code golf but for prompts.

Prompt rot is when the prompts you carefully crafted three months ago suddenly stop working because the AI model got updated.

Prompt-driven development is writing code by writing prompts instead of code.

Rage coding is writing code while furious — usually after a bug has been eluding you for hours, a deploy failed for the fifth time, or someone pushed break...

Resume Driven Development is choosing technologies not because they're right for the project, but because they'll look good on your resume.

Review fatigue is PR fatigue's cousin — the mental exhaustion of reviewing code all day. After 8 reviews, you're not actually reading anything.

You've been staring at this bug for 3 hours.

Rubber ducking is the verb form of rubber duck debugging — the act of explaining your problem out loud (to anyone or anything) to find the solution.

A rubber stamp review is approving a pull request without actually reading the code — just clicking 'approve' and moving on. It's the LGTM of laziness.

Rube Goldberg code is an absurdly complicated solution to a simple problem — like building a machine with 47 moving parts just to flip a light switch.

A Schrodinger's bug is a bug that disappears when you try to observe it. Add a console.log to debug it? Gone. Open the debugger? Works perfectly.

Scope creep is when a simple feature request slowly grows into rebuilding the entire application.

'Send it' is what you say right before pushing to production on a Friday afternoon. It's the deploy-culture equivalent of 'hold my beer.

Shadow prompting is the hidden system prompts that secretly shape how an AI behaves.

A ship factory is a team or codebase optimized for continuous shipping. Deploys happen all day, every day, on autopilot.

Ship it is the developer battle cry. It means 'this is good enough, let's deploy it.

Ship velocity is how fast your team gets features from idea to production. It's the only metric that actually matters.

Ship vibes is the mood of a team that's ready to deploy. Confidence, momentum, tests passing, designs approved.

Slash commands are the /foo shortcuts in AI tools — /memory, /simplify, /loop — power moves for developers who know the cheat codes.

A slop commit is a lazy, AI-generated commit that nobody bothered to review.

A slop farm is a codebase (or content site, or repo) where AI-generated garbage is shipped at industrial scale with no human review.

Slopfarming is the practice of using AI to mass-produce mediocre content, code, or PRs to hit some metric — GitHub contributions, SEO pages, Stack Overflow...

Spaghetti code is code so tangled and messy that following the logic feels like untangling earphones that have been in your pocket for a week.

The sunk cost fallacy in coding is refusing to throw away bad code because you spent so long writing it.

Tab-tab-tab is the sound of a vibecoder accepting every AI autocomplete suggestion without reading it. Ghost text appears? Tab. More ghost text? Tab.

Technical debt is the coding equivalent of putting things on a credit card.

Technical debt is borrowing against future development speed for current delivery.

Technical interviews test you by asking algorithmic puzzles on a whiteboard under time pressure. The theory is this correlates with job performance.

The handoff is that moment when the AI has done 80% of the work and you have to do the remaining 20% yourself.

The Loop is that endless cycle where you tell the AI about a bug, it suggests a fix, the fix creates a new bug, you tell the AI about that bug, it suggests...

In AI-land, a token is a chunk of text — roughly 3/4 of a word.

Token anxiety is that creeping dread when you're deep in a conversation with an AI and you start worrying about running out of context window.

Tribal knowledge is the stuff that only exists in people's heads — the undocumented rules, workarounds, and 'you just have to know' context that's never wr...

Vibe architecture is when you let AI design your system architecture by describing what you need and letting it figure out the structure.

A vibe branch is a feature branch that started as 'quick vibes, just exploring' and is now somehow six weeks old with 400 commits.

Vibe burnout is the exhaustion of endlessly prompting, tweaking, and babysitting AI agents.

A vibe check in dev culture is when you step back and ask 'does this feel right?' before shipping.

Vibe coding is writing software by vibes — you describe what you want to an AI, it writes the code, and you just... vibe.

Vibe debt is AI-generated shortcuts that work today but will hurt later — technical debt's chaotic, fast-moving younger sibling.

Vibe debugging is when you paste an error into an AI and let it figure out what's wrong.

A vibe deploy is when you push to production because the code 'feels ready.' No staging test, no smoke test, just vibes.

Vibe diff is when you look at a code diff and instead of reading every line, you just... feel whether it's good or not.

Vibe drift is when you start an AI coding session on-task and end it three hours later having rewritten your entire auth system for no reason.

Vibe entropy is the natural tendency of a codebase that started with good vibes to drift toward chaos over time.

Vibe lock is that magical state where you and the AI are perfectly in sync. It understands your codebase, your patterns, your intent.

A vibe merge is when you accept a merge without reading the diff because the vibes are right — tests pass, CI is green, the author is trustworthy.

Vibe oncall is letting an AI agent hold the pager. It triages alerts, writes postmortems, even deploys rollbacks. Sometimes it works beautifully.

Vibe ops is infrastructure and DevOps done with AI — from provisioning servers to writing incident playbooks, vibed into existence.

A vibe pivot is completely changing approach mid-session when the AI's plan clearly isn't working. Flexibility as a feature, not a failure.

A vibe refactor is when you tell the AI to 'make this code better' without specifying what 'better' means. Maybe cleaner? Faster? More readable?

Vibe regression is when the AI 'fixes' one thing and silently breaks two others. The tests it didn't know to run now fail.

A vibe review is asking the AI to review its own output — necessary, surprisingly humbling, and better than shipping unreviewed AI code.

Vibe scope is when the AI expands what you're building mid-session without asking. Feature creep, but it's technically the AI's fault.

Vibe ship is pushing straight to production from an AI session without a staging environment. Living on the edge, but shockingly fast.

A vibe sprint is shipping a whole feature in one uninterrupted AI session — no standups, no tickets, no context switching. Just vibes.

Vibe testing is when your entire QA strategy is 'I clicked around and it seemed fine.' No unit tests, no integration tests, no end-to-end tests.

Vibechecking is the informal review process for vibe-coded projects — quickly scanning AI-generated code to make sure it's not doing anything crazy.

A vibecoder is a developer who builds primarily through conversation with AI. They don't memorize syntax — they describe intent.

The most dangerous phrase in software development.

A YOLO deploy is pushing code to production without testing it first because you're feeling brave (or reckless).

You want to fix a bug. The bug requires running the test suite. The test suite needs a Docker container. The Dockerfile is outdated.

In dev culture, yeet means to throw something with force and zero ceremony. Yeet the deploy. Yeet the old code. Yeet that PR into main.

Zero-shot vibing is asking the AI to do something with zero examples, zero context, and zero explanation — just pure vibes. 'Build me a landing page.