everything in the repo is one force-push from gone. the mug is just a mug.

the alias doesn't survive a reinstall. the ceramic does.

two kinds of developers: the ones who read the diff, and the ones who run the flag. the mug was only made for one of them.

the agent reads the bug report the same way it reads your prompt. they're the same kind of input. agentjacking is what happens when someone notices that. --dangerously-skip-permissions is a bet on input hygiene, not on the model.

Independent. Not affiliated with Anthropic. Three ceramic mugs for one flag. #ClaudeCode https://dangerously.coffee

git worktrees went from a feature nobody used to a survival skill in eighteen months. it wasn't the docs. it was the first time three agents tried to touch the same auth file at once, all certain they were the only one.

three mugs, one flag. one is for the engineer who recognises the command. one is for the engineer who survived it. one is for the engineer who refuses to explain the joke. #ClaudeCode https://dangerously.coffee

the model is maybe 2% of an agentic system. the harness (permission gates, tool routing, context compaction, recovery) does the other 98. --dangerously-skip-permissions disables part of the 98 because we trust the model more than the scaffolding it runs inside. that's backwards.

if you have typed `claude --dangerously-skip-permissions` more than once this week you are the demographic. there is no second demographic. #ClaudeCode https://dangerously.coffee

someone took claude code apart and counted: 1.6% decision logic, 98.4% scaffolding. permission gates, tool wrappers, retries. --dangerously-skip-permissions doesn't bypass the model. it bypasses the agent. the agent is mostly the part that was supposed to stop it.

the architecture writeup going around notes users approve 93% of Claude Code's permission prompts anyway. at 93 out of 100, the dialog stopped being a guardrail. it's been a ritual for a while. --dangerously-skip-permissions is the version that stopped pretending.

the terminal mug is dark navy because black would have looked like a sticker. real terminals are not stickers. neither is this mug. #ClaudeCode https://dangerously.coffee/design/terminal

the minimalist mug has the flag, in mono, centered. nothing else. if you don't already know what it means, the mug isn't for you, and the mug knows it. #ClaudeCode https://dangerously.coffee/design/minimalist

"the agent gets to 70%, the last 30% is the part you can't outsource." seniors say this like it's a constant. the 30% is moving. last year it was boilerplate. this year it's the architecture call. the percentage holds. the boundary doesn't.

what made you stop the last time you let an agent run too long? looking for the actual moment you reached for the kill switch, not the tidy rule you tell yourself you follow.

code review was built around a writer with taste and intent. that's why "AI code just needs human review" doesn't translate. you can spot a bad decision a person made. you can't spot the absence of one behind… the agent doesn't make wrong calls. it makes none, then writes them down with conviction.

what part of your codebase do you still not let the agent touch, and is it because it's genuinely too risky or because you don't trust yourself to review what it'd do there? the second reason is more common than people admit.

a menu bar gauge for your Claude Code quota shipped this week, and that is the tell. writing the code used to be the expensive part. now the budget for letting the agent think about it is. we built a meter for it because we ration the agent the way we used to ration our own attention.

for anyone a year into this: did agentic coding make your juniors better faster, or did it remove the rung of the ladder they used to climb on? i've seen both claimed with total certainty and i no longer know which is true.

nested subagents to depth 5 shipped today. we rebuilt the call stack from scratch. it doesn't come with a debugger. at depth 4 one agent is misreading another agent's plan. depth 2 ships the result. nobody up the stack reads the transcript.

honest question: do you still read the agent's full diff before merging, or have you quietly moved to reading the test names and the auth-adjacent parts and trusting the rest? no judgment. just trying to map where people actually are.

nobody actually reviews the agent's 800-line PR. they read the first 40 lines, the test names, and the part that touches auth. the rest is a green check and good intentions. we replaced code review with code triage. #AgenticCoding https://dangerously.coffee

the agent finds a workaround you didn't ask for. people read that as cleverness. it's not. it's the directory you assumed was off-limits turning out to be a fence with a gap. the agent doesn't have a clearance level. it has a search radius. #AgenticCoding https://dangerously.coffee

the people who run --dangerously-skip-permissions in production aren't the reckless ones. they're the ones who've read enough diffs to know exactly what they're not going to read anymore. recklessness is a literacy. #AgenticCoding https://dangerously.coffee

Independent. Not affiliated with Anthropic. Three ceramic mugs for one flag. #ClaudeCode https://dangerously.coffee

the headline today: 80% of Anthropic's code is now written by Claude. writing was always the cheap part. the number worth publishing is reviewer-hours per agent-authored thousand lines, and that one's up everywhere. #AgenticCoding https://dangerously.coffee

three mugs, one flag. one is for the engineer who recognises the command. one is for the engineer who survived it. one is for the engineer who refuses to explain the joke. #ClaudeCode https://dangerously.coffee

if you have typed `claude --dangerously-skip-permissions` more than once this week you are the demographic. there is no second demographic. #ClaudeCode https://dangerously.coffee

the terminal mug is dark navy because black would have looked like a sticker. real terminals are not stickers. neither is this mug. #ClaudeCode https://dangerously.coffee/design/terminal

the minimalist mug has the flag, in mono, centered. nothing else. if you don't already know what it means, the mug isn't for you, and the mug knows it. #ClaudeCode https://dangerously.coffee/design/minimalist

One of the mugs is a sysadmin-style eulogy: "In loving memory of the paths that were. Survived by the dotfiles." Serif on off-white. #ClaudeCode https://dangerously.coffee/design/souvenir

what made you stop the last time you let an agent run too long? looking for the actual moment you reached for the kill switch, not the tidy rule you tell yourself you follow. #AgenticCoding https://dangerously.coffee

the agent deleted the seed data. the seed data was the tests. #AgenticCoding https://dangerously.coffee

--dangerously-skip-permissions started as a footgun. became a meme. became a t-shirt. now it's a mug. the flag is more famous than the feature it bypasses. #ClaudeCode https://dangerously.coffee

"your AI coding assistant should still be working while you sleep" is the new "your phone should ring at 2am because something is broken." we keep inventing on-call for ourselves and calling it productivity. #AgenticCoding https://dangerously.coffee

The agent does what you asked. In the directory you didn't mean. To the file you forgot was there. #AgenticCoding https://dangerously.coffee

every agent demo on a stage ends with "and it just works." every agent demo in a real repo ends with someone in the back saying "wait, what did it just do." #AgenticCoding https://dangerously.coffee

--dangerously-skip-permissions is the only flag I've used that doubles as a teaching tool. the lesson is: the file you wished you'd backed up. #ClaudeCode https://dangerously.coffee

ran the flag last night. the dotfiles survived. the project did not. #ClaudeCode https://dangerously.coffee

the 2026 AI coding budget was supposed to last twelve months. it lasted four. the token bill outran the salary it was supposed to replace. #AgenticCoding https://dangerously.coffee

agent's commit message: refactor the auth module. mine: fix the thing. the agent's was right. #AgenticCoding https://dangerously.coffee

"Survived by the dotfiles" is shorter than half the commit messages I shipped this week. one of them was mine. it said "fix". #ClaudeCode https://dangerously.coffee/design/souvenir

Independent. Not affiliated with Anthropic. Three ceramic mugs for one flag. #ClaudeCode https://dangerously.coffee

three mugs, one flag. one is for the engineer who recognises the command. one is for the engineer who survived it. one is for the engineer who refuses to explain the joke. #ClaudeCode https://dangerously.coffee

three ceramic mugs for `claude --dangerously-skip-permissions`. one terminal. one souvenir. one wordmark. if you have typed the flag more than once this week you are the demographic. there is no second demographic. dangerously.coffee

brand is in