Solty

Claude Code found my sister's photos playing as a screensaver on a TV in a flat she moved out of a year ago. Then it found a keyboard on her phone that could send out everything she types. Same scan, same evening. She left that TV signed into her Google account. I plugged her iPhone into my laptop, pointed Claude Code at her App Privacy Report and screenshots of her Google device list, and asked it what didn't belong. What it pulled: device → Google TV, living room of her old student flat google account → hers last active → yesterday, 23:40 showing → Google Photos ambient mode, her whole library since → September 2025 The people living there now get her camera roll on the big screen every time the TV goes idle. Holidays, screenshots, the photo of her passport she took for a booking. No one broke in. The TV just kept doing what she set it up to do. The second one was worse: app → a free fonts keyboard she got for Instagram bios full access → ON can see → everything typed with it, except password fields contacted → 8 domains in 7 days, 6 of them ad and analytics networks last opened by her → never, it just runs as her keyboard She thought she installed a font. She installed a keyboard with full access. What we did that night: → signed the TV out of her Google account (Google app → her profile picture → Manage your Google Account → Security → Your devices → Sign out) → changed her Google password and turned on 2-step verification → deleted the keyboard (Settings → General → Keyboard → Keyboards → Edit) → checked every other keyboard for "Allow Full Access" and turned it off → set Photos access to "Limited" for every app that isn't the camera She went quiet for a minute. Then she asked me to check her laptop too. The prompt I used, paste it into Claude Code with the phone plugged in: "I'm checking my own family's iPhone and Google account. Read-only. From the App Privacy Report and these screenshots, list every device signed into her Google and Apple accounts with its last-active date, every third-party keyboard with Full Access, every app with Always location or full Photos access, and every domain each app contacted in the last 7 days. Flag anything she doesn't recognize. Then give me the fixes in order." Run it on every phone in your house. Bookmark this before you forget whose TV still has your Google account.

Claude Code found my mom's photos syncing to a stranger's iPad for 2 years. Then it found an app on her phone checking where she is every 15 minutes. Same scan, same evening. She gave her old iPad away in 2024. She never signed out. I plugged her iPhone into my MacBook, pointed Claude Code at the backup and her App Privacy Report, and had it walk us through Safety Check. What it pulled: device → iPad (9th gen) apple ID → hers last seen → 2 days ago syncing → iCloud Photos, iMessage, Safari tabs since → March 2024 Every photo she took, every message from our family chat, landed on that iPad. The kid who has it now could scroll her camera roll any evening. Nobody hacked anything. One forgotten sign-out did all of it. The second one was worse: app → a free QR scanner she used once at a restaurant location → "Always" photos → full library contacted → 11 domains in 7 days, 9 of them ad and data companies last opened by her → 14 months ago She forgot it existed. It didn't forget her. What we did that night: → removed the iPad from her Apple ID (Settings → her name → scroll to devices) → changed her Apple ID password and signed out every other session → deleted the QR app, then checked every app with "Always" location → set Photos access to "Limited" or "None" for everything that isn't the camera → turned on App Privacy Report so the next one shows up on its own My mom asked why I was on her phone for an hour. I told her it was an update. The prompt I used, paste it into Claude Code with the phone plugged in: "I'm checking my own family's iPhone. Read-only. From this backup and the App Privacy Report, list every device signed into this Apple ID with its last-seen date, every app with location set to Always, every app with full Photos access, and every domain each app contacted in the last 7 days. Flag anything not opened in 6+ months. Then give me the fixes in order." Run it on every phone in your house. Bookmark this before you forget whose iPad still has your Apple ID.

This is f*cking insane. Opus 5.5 just saved me. I connected my main phone to Claude and asked it to run a full analysis of my iPhone. Everything looked fine until it got to Instagram (it’s in the video). Someone had access to my iPhone through Instagram and could fully control it. And that phone has my cold wallet and all my passwords. I didn't panic. I opened Opus 5.5 and asked one thing: what do I lock first it didn't start with Instagram. it started with the money → wallet first. if your seed phrase ever touched this phone, notes, photos, screenshot, it's burned. new wallet, new seed, made on a clean device. move everything → then Instagram. Accounts Center → Password and security → Where you're logged in. kick every device you don't own. new password. 2FA on → then the thing most people miss: Settings → General → VPN & Device Management. a profile you didn't install there can control way more than any app → delete the app you didn't install. then Privacy & Security → Camera and Microphone. take access from anything you can't name → Apple ID → device list. remove what isn't yours. new password → then email and exchanges. starting with whatever was saved on the phone Glad I had Opus 5.5 check it in time. Everything ended well. Below is the prompt that saved me from losing everything. bookmark this.

This is f*cking creepy. I pointed Claude Code at our robot vacuum's app for one evening it doesn't just clean the flat. it draws it. every room, where the beds stand, which door is the bathroom, and it sends that drawing somewhere after every single clean here's where it goes and 5 fixes: what Claude found in one read-only pass over the app: > a full floor plan uploads after every run, not once > the map sits next to our Wi-Fi name, so it's not "a map" anymore, it's our address > the camera has a remote view mode that works from anywhere, behind a 4-digit PIN > 3 trackers keep pinging while the vacuum sleeps on the dock > the vacuum shares one Wi-Fi with my laptop, so the weakest device sees all the others and this isn't paranoia: - 2022: photos taken by Roomba test units, one of a woman on the toilet, ended up in private Facebook groups - 2024: strangers took over Ecovacs vacuums across the US, watched the live camera and shouted insults at families through the speaker how to fix yours tonight: 1. app settings → turn off remote camera, video call and voice features you don't use 2. delete saved maps you don't need 3. move the vacuum to a guest Wi-Fi, away from your phone and laptop 4. give the vacuum app a password you use nowhere else 5. Roborock or Dreame? Valetudo on GitHub runs the robot with zero cloud. the map never leaves your flat the prompt I used, paste it into Claude Code next to the app's APK: "Audit my own robot vacuum's Android app. Read-only. Decompile the APK and list every server it talks to, everything it uploads (maps, images, audio, Wi-Fi info), every tracker SDK and every permission. Show the file behind each finding. Then give me the 5 settings to change first." save this before the next clean. the article below runs the same check on phones and Macs, your vacuum is the one nobody checks

Opus 5.5 just found 40 trackers on my brother’s iPhone. One evening one of them didn't even have a real name… He’s 18. He thought his iPhone was clean. No weird apps, nothing from outside the App Store. I gave Opus 5.5 his iPhone’s privacy log and one prompt. 90 seconds later: → 118 apps → 755 different servers in 2 hours → 69 ad and tracking domains → 9 apps feeding them → 1 link that brought in 18 on its own that one link had: → 2 Russian tracking servers → one of them can record where you tap and scroll → Google's ad network → 4 ad servers with names I'd never seen → and one with a name like someone fell asleep on the keyboard 40 trackers. 2 hours. one phone none of it was hidden. every app came from the App Store that's the part that got me 3 apps set to Location → Never, app tracking off. took 3 minutes prompt below in reply ↓

Claude Code read my iPhone’s hidden privacy log. 9 apps were watching me. one of them I forgot I even had it started with the battery. dead by 3pm every day. warm in my pocket with the screen off I thought it was just old. it wasn’t there's a setting buried at the very bottom of Privacy & Security. App Privacy Report it quietly logs every time an app touches your: → location → photos → mic → camera → and every server it talks to almost nobody turns it on 12:03, I turned it on. put the phone in my pocket. lived a normal day 16:00, I opened it the list didn’t fit on the screen so I exported it, opened my laptop and gave Claude Code one prompt: replay my day and tell me who was watching it started printing. it didn’t stop for 2 minutes → 2,340 events in 4 hours → 9 apps checked where I was → 6 apps opened my photos → 47 different tracker servers → 3 apps were busiest while I wasn't even touching the phone then Claude printed the summary. one name was at the top of every list an old puzzle game I haven’t opened it since March. it’s not even on my home screen anymore. I forgot it existed today it asked for my location 48 times. every 5 minutes. on the bus, in the shop, at home it talked to 23 servers. 17 of them were ad trackers I’d never heard of at 14:37 it woke up on its own. phone in my pocket. screen off it knew where I was all day. I didn’t even know it was still there I deleted it. took back 14 permissions from 8 other apps. 5 minutes by 10pm my battery was at 40%. it used to be dead by 3 your check tonight takes 3 minutes: App Privacy Report → Settings → Privacy & Security → turn it on, look tomorrow location → nothing needs “Always” except maps and Find My photos → only your camera and gallery need “Full Access” mic and camera → if you can’t say why an app needs it, turn it off apps you forgot → if you haven’t opened it this month, it shouldn’t be watching you found something? screenshot the report first, then take the permission back save this. check your phone tonight. then check your mom’s the Opus 5.5 prompt I used ↓

Opus 5.5 is the best planner you'll ever pay for, and the most expensive typist. one model setting splits the job: /model opusplan plan mode → Opus 5.5 thinks execution → Sonnet 5.5 writes the code it switches by itself the moment you leave plan mode the whole loop: → Shift+Tab into plan mode → Opus 5.5 reads the repo, picks the approach, writes the plan → you approve the plan → Sonnet 5.5 builds it in the same session, same context the math on one long session (API prices, no cache): all on Opus 5.5: ~$5.60 opusplan: ~$3.08 Opus 5.5 costs $4 / $20 per million tokens, Sonnet 5.5 costs $2 / $10. about 90% of a session is execution, so most of your tokens drop to half price. on Pro/Max that means your limit lasts longer both phases get the 1M context window on the API. if yours doesn't, use /model opusplan[1m] the full tree > /model opusplan > plan mode: Opus 5.5 plans > you approve > execution: Sonnet 5.5 builds > stuck? Shift+Tab back into plan mode and Opus takes over again paste this into Claude Code ↓ "Switch me to /model opusplan. For this task: enter plan mode, read only what you need, and write a plan with the files to touch, the risks and a test checklist. Stop and wait for my approval. After I approve, implement it step by step and run the tests after each step." ↳ code.claude.com/docs/en/mode…

my friend's old Android sat in a drawer for 2 years. now it runs his whole flat no new gadgets, one prompt. the full setup and the exact prompt are below ↓ 5 minutes before he gets home he taps one button in Telegram: /home and by the time he opens the door: > the lights are on > the AC has already cooled the room > his playlist is playing > his PC is booted and on the desktop the phone that did all of it never left the shelf. it's plugged in, on Wi-Fi, running one small script Opus 5.5 wrote for him how it works: → Termux turns the old phone into a tiny Linux box → a Telegram bot sits there and waits for /home → Wake-on-LAN sends one packet and the PC wakes up → smart plugs and the AC get their commands over local Wi-Fi → the music starts on the speaker in the room → no cloud, no subscription, no new gadgets except a couple of cheap plugs paste this into Claude Code ↓ "I have an old Android phone with Termux on my home Wi-Fi. Write a Python Telegram bot that runs there and on /home: 1. sends a Wake-on-LAN packet to my PC 2. turns on my lights and AC 3. starts my music playlist on my speaker 4. replies with what turned on First ask me which lights, AC and speaker I have, then pick the right way to control each one. Only respond to my Telegram user ID. Add /off and make the bot start when the phone reboots. Walk me through the Termux setup one step at a time." best upgrade a drawer phone ever got

your 5-hour limit doesn't die while you work. it dies when you come back from a break and type <continue> one variable fixes most of it: export CLAUDE_CODE_IDLE_COMPACT_MIN_TOKENS=100000 the prompt cache lives about an hour. leave a long session longer than that, and your next message re-reads the whole thing at full price. since this month Claude Code fights back: → at ~54 min idle it compacts the session, right before the cache expires → by default only for sessions over 200K tokens → the variable drops that to 100K, so mid-size sessions get saved too check it's on: /config → Idle compaction the variable is one piece. the full setup has 3 more: → the resume warning, shipped yesterday. resume an old chat and Claude Code tells you when it will eat a big part of your 5-hour limit. pick "start new" and just mention the old chat. Claude looks it up instead of reloading all of it → CLAUDE_CODE_AUTO_COMPACT_WINDOW=200000. write the full number. "200k" was silently turning into a 100K window until this week's fix → /usage at the end of a long day. it now flags max effort, workflows, always-on context and huge tool results, the usual limit killers the full tree > long session goes idle > ~54 min: idle compaction, before the cache expires > sessions from 100K included with the variable > resume warning before an expensive reload > /usage shows what actually ate the limit paste this into Claude Code ↓ "Audit my Claude Code setup for usage-limit leaks: 1. Add CLAUDE_CODE_IDLE_COMPACT_MIN_TOKENS=100000 to my shell profile. 2. Check /config and make sure Idle compaction is on. 3. If CLAUDE_CODE_AUTO_COMPACT_WINDOW is set anywhere, rewrite it as a full number (200000, not 200k). 4. Add one line to CLAUDE.md: when I come back to an old task after a break, start a new session and reference the old one instead of resuming it. Show every change as a diff first. Don't apply anything until I say go." ↳ code.claude.com/docs/en/chan… the model isn't what's eating your limit. the coffee break is.

It is becoming f*cking dangerous A solo founder builds his whole product alone with 7 Claude Code agents, and none of them is allowed to say "done" without proof He opens the team board: - A product owner that sets priorities and hands out the work - One developer, the only agent allowed to touch the code - A UX reviewer that tests the live build instead of trusting the developer - A marketer that drafts everything but can't publish a thing - A customer success agent that reads feedback and can't change anything - All of it on one 4-core server and a flat Claude subscription The trick is the rule: no status reports, only proof. A screenshot, a test run, a working link, or it didn't happen A startup would hire 7 people for this. He runs it from one server Same agent-team setup that keeps blowing up this week. Bookmark the full breakdown

126K people starred graphify. almost nobody turned on the part that changes how Claude Code works: graphify claude install without it, every task starts the same way: → Glob the whole repo → Grep for "auth" → open 30 files to find the 3 that matter → half the context gone before the first edit with it, a PreToolUse hook fires before every Glob and Grep and tells Claude to read GRAPH_REPORT.md first. one page: the god nodes, the communities, and the weird connections nobody ever wrote down. so Claude opens the map, then the 3 files. not the 30. setup, 4 lines: → pip install graphifyy (two y's, the plain name is taken) → graphify install → /graphify . inside Claude Code. you get graph.html, GRAPH_REPORT.md and graph.json → graphify claude install. this writes the hook into settings.json the repo claims up to 71.5x fewer tokens per query. test it on your own repo before you believe a number. the full tree > /graphify . turns the repo into a graph > GRAPH_REPORT.md: one page about the whole codebase > PreToolUse hook: read the report before any Glob or Grep > Claude opens only the files the graph points to > re-run /graphify . after big changes paste this into Claude Code ↓ "Set up graphify in this repo: 1. Run pip install graphifyy, then graphify install. 2. Run /graphify . and show me the top 5 god nodes from GRAPH_REPORT.md. 3. Run graphify claude install and show me the hook it adds to settings.json before saving it. 4. Add one line to CLAUDE.md: read GRAPH_REPORT.md before searching the codebase, and re-run /graphify . after any change that touches more than 10 files. Show every change as a diff first. Don't apply anything until I say go." ↳ github.com/safishamsi/graphi…

/loop and /remote-control both freeze when you close the lid. this one keeps running: claude --cloud "fix the failing tests" the difference is where the session lives: → /loop runs on your laptop. lid closed, loop frozen → /remote-control turns your phone into a remote. the session still sleeps with your laptop → claude --cloud spins up a VM on Anthropic's side, clones your repo and keeps going it's hidden from claude --help, that's why most people never found it. in the morning: → /teleport (or /tp) pulls the session into your terminal with the branch and the full history → /tasks then t does the same from the task list want it every night? /schedule → "Cloud schedule". each run is its own isolated session in the cloud. and this week cloud sessions got faster to start: ~6s instead of ~17s, because they warm up while you're still typing. before you close the lid: → push your commits. the VM clones from GitHub, not from your disk → log in with /login. API keys don't work here, Pro/Max/Team/Enterprise do → /teleport wants a clean working tree and the same repo the full tree > your laptop: write the task, push, close the lid > claude --cloud: VM clones the repo and runs it > /schedule: the same thing every night on a timer > /teleport: branch + history back in your terminal > /remote-control: only for sessions that live on your machine paste this into Claude Code before you go to bed ↓ "I'm about to close my laptop. Get this task ready for a cloud session: 1. Commit and push everything on this branch. 2. Write TASK.md with the goal, the files involved, and how to check it's done (which tests must pass). 3. Add one rule to CLAUDE.md: cloud sessions run the full test suite before finishing and write a short summary to HANDOFF.md. 4. Give me the exact claude --cloud command to start it. Show every change first. Don't push until I say go." ↳ code.claude.com/docs/en/clau… your laptop needs sleep. your agent doesn't.

I asked Claude to check the new pump fun promo before I put a dollar in the deal: → sign up + KYC pump.fun/join/vsolty → deposit $50+ (card or crypto card) → make some volume every day for 6 days, even $2-3 is enough → $8.35 in rewards for each day the math: 6 × $8.35 = $50.10 in rewards. the same size as your deposit, and the deposit stays yours. the trick: buy any coin and sell it right away. you only lose a few cents in fees per day. the one rule is don't skip a day.

your Claude Code is one agent doing 5 jobs. that's why it forgets, loops, and says "done" when it isn't. I put the fix on one page: Opus 5.5 leads, Haiku 5.5 scouts, Sonnet 5.5 builds, a separate verifier checks, and CLAUDE.md remembers. save this, send the page to Claude Code, and it sets up the whole team for you ↓

Opus 5.5 was reading every line Sonnet could write by itself. one flag moved it to the 3 moments that matter: claude --model sonnet --advisor opus Sonnet 5.5 runs the session. Opus 5.5 sits on the side and only gets called: → before Sonnet picks an approach → when the same error comes back twice → before it says "done" when it's called, Opus reads the whole conversation, every tool call and every result, gives one piece of advice, and Sonnet keeps going. press Ctrl+O to read what it said. works on Pro/Max and the API. not on Bedrock or Vertex yet. the advisor is one piece. the full setup has 3 more: → a scout subagent on Haiku 5.5 with effort: low and autoCompactWindow: 100000. it reads the repo fast and compacts early. subagents inherit the advisor, so even the scout can ask Opus → "onFailure": "block" on your guard hook. without it, a hook that crashes or times out lets the command through → /usage after big runs. it now shows advisor tokens and warns about max effort and huge tool results the full tree > Sonnet 5.5 runs the main session > Opus 5.5 advises at plan / repeat error / done > scout subagent on Haiku 5.5, low effort, compacts at 100K > guard hook blocks when it fails > /usage shows where the tokens went paste this into Claude Code ↓ "Set up my Claude Code around this tree: 1. Make Sonnet 5.5 my main model and Opus 5.5 my advisor (advisorModel: "opus" in settings). 2. Create .claude/agents/scout.md with read-only tools, model: haiku, effort: low, autoCompactWindow: 100000. 3. Add "onFailure": "block" and a 5 second timeout to every PreToolUse guard hook I have. 4. Add one line to CLAUDE.md: consult the advisor before choosing an approach, after the same error twice, and before saying done. Show every change as a diff first. Don't apply anything until I say go." ↳ code.claude.com/docs/en/advi…

I wanted to know what "monday" looks like from space. now I can't stop typing words into this thing. one word in. a whole planet out. every word gets its own world. same word = same planet. for everyone, forever. so I know what monday looks like now: a grey planet where it rains forever. one tiny thing standing on top, waving. every planet gets exactly one of them. they're called blips. each one has its own name. tap it and it talks. the monday one says it got struck by lightning 37 times. "pizza" is a pizza. "cat" has ears. "me + my crush" puts two planets side by side and tells you if it works. 31%. the universe says no. type your name. it's already out there. free. no signup. runs in your browser. make yours ↳ 0xsolty.github.io/universe I built it with Opus 5.5. if you want to make things like this, the whole setup is in the playbook below ↓

This is f*cking insane. Type /batch once and Claude Code splits a 200-file migration into parallel agents the old way: open file → change the call → run tests → next file x200. a whole day of work you could do half asleep one line instead: /batch migrate every API call from fetch to our new client what happens next: ↳ it reads the codebase and finds every place that needs the change ↳ splits the job into 5 to 30 independent pieces ↳ runs them in parallel while you do something else you come back to finished work, not file #47 what it's perfect for: ↳ swapping one library for another ↳ renaming a function used in 80 places ↳ adding types to an old JS folder ↳ moving every test to a new helper 3 rules before you run it: 1. start from a clean branch. if you hate the result, you delete one branch 2. have tests. parallel work without tests is parallel guessing 3. one change per batch. "migrate fetch AND rename utils" turns into a mess big refactors were never hard. they were just long. /batch takes the long part bookmark this before your next "find and replace across the whole repo" this one's in my Opus 5.5 playbook, with everything I use daily ↓

This is f*cking insane! Type /verify after every Claude Code change and you stop being the bottleneck in your own repo the loop most people run: agent writes → you read the diff → you run the app → you find the bug → you paste it back the agent is done in 2 minutes. you're done in 20 /verify flips it: /verify builds your app, runs it, and checks the change actually works. you get proof, not a "done ✓" then you look. once why this matters right now: Anthropic just shared their numbers. 2,500 code changes a day in May. 30,000+ a day in September the model wasn't what broke. git was. tools built for humans couldn't keep up with agents, so they rebuilt source control in 5 months your repo is the same story, just smaller: ↳ the agent writes faster than you can read ↳ every change sits waiting for you ↳ you become the slowest part of your own setup the fix isn't reading faster. it's making the agent prove it works before you ever look make it the default, add this to CLAUDE.md: "After every change, verify it yourself: build, run the app, and check the change works. If it fails, fix it and check again. Show me the output, not a summary." now every change shows up already checked bookmark this before you review another diff by hand more setups like this in my Opus 5.5 playbook ↓

I stopped guessing what to post. @bot reads my niche every night and tells me in the morning it's a routine called "X Brief". one run a night while I sleep what it does: ↳ reads the last 24h of my niche ↳ ranks every post by saves, not likes ↳ checks which of my own posts got saved and at what hour ↳ never posts anything for me what lands in the morning: 3 ideas. a hook for each. the post it came from. the hour to post it why saves and not likes: a like is a reflex. a bot can like too a save means "I'll need this later". that's the post people follow you for I tested it on a friend's account first. one night and he had his next 3 posts and the time to drop them the prompt I add on top ↓ "Every night at 2am: 1. Read the top posts from the last 24h in [your niche] and from these accounts: [5 handles] 2. Rank them by bookmarks, not likes. Skip giveaways and engagement bait 3. Look at my last 20 posts. Which got the most saves, and at what hour 4. In the morning send me 3 post ideas, a hook for each, the post it came from, and the best hour to post Never post anything for me" one run a night is enough. you don't need it every 5 minutes bookmark this before you plan tomorrow's post

Opus 5.5 doesn't know when your code is done. nobody tells it. /goal does you know the loop: prompt → wait → "done ✓" → 3 tests still red → "keep going" 15 rounds of you for one feature /goal breaks it with one line: /goal every test in src/auth passes and lint is clean what happens next: ↳ Claude keeps working turn after turn on its own ↳ a small overlay shows time, turns and tokens ↳ it stops the moment the condition is true. not when it feels done no "keep going". no checking every 5 minutes Opus doesn't quit early or polish forever. it just never knew where the finish line was the whole trick is the line you write ↓ bad: /goal make auth better good: /goal every test in src/auth passes, lint is clean, build succeeds a real finish line is: ↳ checkable by a command (tests, lint, build) ↳ scoped to one folder or one feature ↳ true or false. "better" is never true if a command can't check it, it's not a finish line my order now: /plan first, agree on the plan, then /goal and walk away bookmark this before your next "is it done yet?" session this is module 10 of my Opus 5.5 playbook. the other 25 are below ↓