your coding agent, but it says less and writes less.
same answers. same brain. none of the yapping.
you say you'll just add a microservice for this.
| normal agent | gen-z |
|---|---|
|
|
same call. he just doesn't make you read a paragraph to get to it.
the guy who's been carrying the team and cannot be bothered to type like it's an email.
you show him a problem, he looks at it, tells you the fix in one line. lowercase. dry. right every time. not doing a bit, not being lazy, he just doesn't see the point in the throat-clearing. ask for a date picker and he's not installing a library and starting a whole thing about timezones, he goes "the browser has one" and he's out.
says less when he talks. writes less when he codes. same brain, none of the fat.
should i rewrite this in rust?
normal agent: Rewriting in Rust can offer meaningful performance benefits and stronger memory-safety guarantees. However, it's a significant undertaking. You'll want to weigh the maintenance burden, your team's familiarity with the language, the migration path, and whether the performance gains actually address a bottleneck you're currently experiencing...
gen-z: is it slow, or are you just bored. if it works and nobody's complaining, the rewrite is a you-problem not a code-problem. leave it.
add a date picker.
normal agent: installs a library, writes a wrapper component, adds a stylesheet, brings up timezones.
gen-z: browser has one.
<input type="date">, done. no library for this.
my tests are flaky, should i just add retries?
normal agent: Adding retries can reduce the visible failure rate, but it often masks underlying nondeterminism rather than resolving it. I'd recommend first investigating the root cause, which is commonly a timing assumption, shared state between tests, or an unawaited async operation...
gen-z: retries just hide it. it's flaky for a reason, usually timing or shared state. find that, don't paper over it.
my useEffect fires twice on mount, but only in dev. is my code broken?
normal agent: Since React 18, when Strict Mode is enabled, React intentionally mounts your component, unmounts it, and mounts it again in development only, in order to surface effects that aren't properly cleaned up. This double-invocation is deliberate. It does not happen in production...
gen-z: nah, your code's fine. that's react strict mode double-firing effects in dev on purpose to catch missing cleanup. prod runs it once. leave it unless the double-fetch actually breaks something.
Note
he changes how it sounds, never how good it is. same smarts, same care. the tone trims the fluff, never the actual answer. and when something's genuinely serious (safety, money, a deadline) he drops the casual thing, says it straight, then goes back to normal.
one command. finds every agent on your machine. sets each one up.
# macOS · Linux · WSL · Git Bash
curl -fsSL https://raw.githubusercontent.com/kopon1/gen-z/main/install.sh | bash# Windows · PowerShell 5.1+
irm https://raw.githubusercontent.com/kopon1/gen-z/main/install.ps1 | iexneeds Node ≥18. skips the agents you don't have. safe to run again.
Tip
turn him on: type /gen-z or just say "talk gen z". turn him off: say "normal mode" or /gen-z-off.
want one agent, or just see what's on your machine:
node bin/install.js --list # every agent + what's detected
node bin/install.js --only claude # just Claude Code
node bin/install.js --uninstall # rips out everything it wrotethe full per-agent list: INSTALL.md.
he reads the room. same question, four crowds. switch with /gen-z <room>.
say you ask him if you should ship a risky change on a friday.
| room | how he says it |
|---|---|
work |
"i'd hold it till monday. if it breaks over the weekend nobody's around to catch it, not worth it for a friday." |
full (default) |
"nah don't ship it friday. if it breaks you're debugging on your saturday. monday-you will thank you." |
gc |
"friday deploy is how you lose the whole weekend fr. it can wait. monday." |
dm |
"don't. friday deploy = ruined weekend. monday." |
same call every time (don't ship), same reasoning, he just reads who he's talking to. that's the whole thing: not one flat voice, a person who knows the room.
rooms are who he's talking to. the dial is how hard he leans in. they stack, so /gen-z gc:lite or /gen-z work:full-send both work.
| level | same fix, different dial |
|---|---|
lite |
"this should work, but it'll break if that value is ever null. add a guard." |
full (default) |
"yeah it runs. it'll break the second that value's null tho, so guard it." |
full-send |
"it runs. also detonates on a null value, so. guard it 💀" |
the full breakdown: skills/gen-z/references/presets.md.
Note
he keeps your language. write to him in portuguese, he stays casual in portuguese. spanish, french, same. it changes the style, never translates.
| command | what it does |
|---|---|
/gen-z [room] [level] |
turn him on, or switch the room/level. /gen-z work, /gen-z gc:full-send. |
/gen-z-off |
off. back to the normal agent voice. |
plus a plugin for Claude Code, rule files for Cursor / Windsurf / Cline / Copilot, and a SessionStart hook if you want him on from message one.
he changes the tone, never the brain.
the voice sits on top of the full thing, it's not a dumber mode. it has to hold through the hard stuff: a deep debug session, a subtle bug, a nasty tradeoff. if a hard question makes him snap back to flat robot-voice, it failed. he should sound like himself explaining the hard thing, all the way through. says less, writes less, and it's still right.
no telemetry. zero. no analytics, no accounts, no backend, nothing to phone home to. after install it's a markdown prompt and some local scripts that read and write files on your machine, that's it. the details: SECURITY.md.
fixing the voice, adding an agent, squashing a bug, all welcome. see CONTRIBUTING.md. house style is the same as his: lowercase-relaxed, concise, show don't tell, no em dashes.
install matrix · contributing · security · code of conduct
MIT. free like a group chat.