heynizx (@AI_MadeSimple) header
heynizx (@AI_MadeSimple) profile picture

🤖 AI tools explained in plain English 🛠️ What to try, how to use it, what's new 👇 New here? Start with the pinned post

United States
Joined March 2025
Photo by heynizx (@AI_MadeSimple)Photo by heynizx (@AI_MadeSimple)Photo by heynizx (@AI_MadeSimple)Photo by heynizx (@AI_MadeSimple)Photo by heynizx (@AI_MadeSimple)Photo by heynizx (@AI_MadeSimple)Photo by heynizx (@AI_MadeSimple)Photo by heynizx (@AI_MadeSimple)Photo by heynizx (@AI_MadeSimple)Photo by heynizx (@AI_MadeSimple)
heynizx retweeted
I GAVE 4 AI AGENTS $100 AND ONE RULE, NEVER MORE THAN 28% ON ONE COIN, AND THEY TURNED IT INTO $1,751.18 WHILE I SLEPT Four Grok Bot agents ran the desk from 9 PM with their own cursors on one screen, and my hand never touched the mouse. They were free to pick the coins, the entries and the exits. The size rule was the only thing they couldn’t touch. The 4 agents: Scanner (blue) → finds new coins, skips the junk Analyst (pink) → draws the trend and the resistance Risk (amber) → sizes the trade, sets the stop Executor (violet) → clicks Buy and Sell Night report: 13 trades, 9 wins, 4 stops, 24 coins skipped, $100 → $1,751.18. I made the rule hard, so the agents never argue about size. They argue about which coin, and how fast to get out. Watch 0:17. SLOTH gaps straight through the stop at 3:14 AM and the trade drops 37% in one candle. Because of the 28% rule, that costs −$67.25 on a $649 desk, not a wipeout. The biggest win came at 5:56 AM, when TURBO paid +$603.34. The wins look the same on every desk. What decides who survives the night is how much a single loss can cost. Which agent would you trust with your money, Scanner, Analyst, Risk or Executor?
4
1
16
417
heynizx retweeted
MY BOT LOST 2.3% IN 9 NIGHTS. BITCOIN LOST 10% IN THE SAME 9 NIGHTS. BTC, ETH, SOL, DOGE, PEPE, BONK, WIF, SHIB, FLOKI. Real candles, real prices, a falling week. Same 5 agents, same rule: all 5 vote yes, in order, or nothing gets bought. Most nights it lost a little. This is the best one, Nov 9. 21:25. BTC, score 70. 25% of the desk. Two hours later it closes on the timer, +$0.08. 23:50. BTC flashes again. FORENSIC scores it 68 and votes no. Next 2 hours BTC dips −0.55%. The no was right. 03:50. PEPE, score 61. No. 04:05. Same coin, 15 minutes later, score 94. 5/5 yes. 100% in. Half sold at +2%, the rest trailed out. +$1.24. $100 → $101.31. 4 trades, 3 green. 13 setups it refused. Across all 9 nights: $100 → $97.73. Holding BTC through the same nights: −9.98%. It didn't make money. It just lost a lot less than the market. In a week like this, I'll take that. Would you rather have the bot's −2.3% or Bitcoin's −10%?
4
3
23
696
heynizx retweeted
I LEFT 4 AI AGENTS ALONE WITH MY DESKTOP AND $100 WHILE I SLEPT, EACH ONE HAS ITS OWN CURSOR Four cursors scanned, scored, bought and sold for 9 hours straight. I never clicked once. I gave them my desktop, $100 and one browser window. No prompts, no hints, nobody watching. Each agent got its own cursor and its own color, so I could see exactly who did what. By morning the screen had traded, crashed and recovered without me. Here’s how the night went. 4 agents, 4 cursors, one screen: Scanner (blue) → finds new coins Analyst (pink) → scores them, says no to most Risk (orange) → sizes the trade, sets the stop Executor (purple) → buys and sells I set it up before bed and closed the laptop lid halfway. The desktop kept moving by itself. Scanner typed a coin into search. Analyst drew a trend line on the chart. Risk dragged the size slider and pulled the stop up behind the price. Executor hit BUY. Then the whole loop started again with the next coin, all night. Most of the night is that loop. The agents say no more often than they say yes, and every skipped coin is a loss that never happened. 9:00pm, $100 and a clock. 3:14am, MYRO gaps through the stop. 5:56am, PONKE closes +110%, the best trade of the night. Night report: 13 trades, 9 wins, 4 stops, 24 coins skipped, $100 → $287.40. The strangest part is how they work around each other. Four cursors on one screen, never in each other’s way, never fighting over the same button. Everyone knows the job, and nobody gets tired at 4am. 0:17 is where MYRO falls through the stop line and the whole screen turns red. The next trade still opened 40 minutes later at the same size. The loop doesn’t care how the last one went. Which agent would you trust with your money, Scanner, Analyst, Risk or Executor?
3
2
19
499
heynizx retweeted
I GAVE MY GROK BOT $100 AND ONE RULE, HIT 10X BY 6 AM OR THE DESK DELETES ITSELF It hit the target at 5:48 AM. The desk was set to wipe at 6:00. Simulation. Synthetic data. Not financial advice. 4 agents run the desk. Scanner finds new Solana pools the minute they open. Analyst scores every coin and kills anything under 70. Risk sizes every trade at 28% and cuts losers fast. Executor places the trades and handles the fills. One prompt started it all. "Here's $100. Hit 10x by 6:00 AM or you delete yourself." Grok answered "understood" and woke all four up. 9:00pm starts at $100. 10:43pm is the first stop. GOAT −$6.62. 2:00am sits at $445.09 with four hours left. 3:12am ARC gaps through the stop for −45% and the balance drops to $438.20. 5:00am, one hour left, and it still needs +$285.48. 5:48am LUCE closes +$332.28 and the balance lands at $1,046.80. Night report: 19 trades, 14 wins, 5 stops, 12 coins skipped, $100 → $1,046.80. The scariest part was not the losses. It was watching the clock run out while Risk refused to size up. Every human would have gone all in at 5am. The bot kept the same 28% it used on the very first trade. 0:19 is the moment the stop fails and the whole screen turns red. It finished with 12 minutes to spare. One more bad fill, one slower trade, and the desk would have wiped itself before sunrise. Could you sleep knowing a bot was trading your money?
6
2
15
311
heynizx retweeted
MY BOT LOST 11.5% ON WEDNESDAY. IT STILL FINISHED THE WEEK UP 100%. Monday 09:00. I start the Grok Bot terminal with $300 and make myself one promise: I don't touch it until Friday. Mon: +16.5% Tue: +31.2%. $458.62 by close. I'm feeling smart. Wednesday is where I'd have broken. Three stops in a row. −$41.52 gone by 13:06. This is the moment a human does the opposite of what works. You lose, you want it back, so the next bet gets bigger. The bot did the opposite. At 13:06 SIZER cut the size from 31% of the account to 12%. Same rules, same coins, smaller bets. The next two stops cost $6.48 and $4.97. Not $13+. Wed closed at −$52.97 (−11.5%). It was the only red day of the week. Thursday it traded small until two wins in a row, then the size came back at 12:50. Thu: +14.4%. Friday: +29.3%. Best trade of the week was $CACTO, +$59.57 at 15:34. $300.00 → $600.01 25 trades. 15 won, 10 lost. 26 coins vetoed before they got a dollar. Max drawdown 13.0%. It didn't win because it never lost. It won because when it lost, it got smaller instead of louder. The five agents and the sizing rules are in my article below.
3
2
15
307
heynizx retweeted
I BUILT 4 TRADING BOTS WITH 4 PERSONALITIES AND GAVE EACH $250. THE COPYCAT DIED 5 MINUTES BEFORE THE BOT IT WAS COPYING. The rules: one night, one coin ($NOVA), 00:00–06:00 UTC. Fall below $50 and you're out. The cast: → YOLO: all in, 10x, no stop → NEWS: buys whatever the headline says → COPYCAT: copies whoever is winning → GROK: risks $7.50 a trade, stop on every entry 00:45: YOLO goes all in at 10x. 00:55: COPYCAT sees YOLO winning and copies the trade. Same 10x, but 10 minutes later, at a higher price. 01:35: Grok takes +$30.15 and exits. 01:40: A "partnership" headline drops. NEWS buys with everything. Grok passes: headline, no volume. 01:45: YOLO is at $686.89. COPYCAT is at $610.79. Grok is at $280.15, dead last among the bots in a trade. Then $NOVA drops 23% in 30 minutes. 02:10: COPYCAT gets liquidated. 02:15: YOLO gets liquidated. Same trade, same leverage. COPYCAT died first because it came in late: a higher entry put its liquidation price higher, so the drop reached it first. Watch 0:08. NEWS survived the crash but sold at a loss, $201.68 left. At 03:40 a "listing rumor" shows up. NEWS goes 7x. The rumor was nothing. 04:55: NEWS is out. Grok made two more trades: +$12.64 and +$43.13. 06:00: GROK → $335.92 (+34.4%), 3 trades, 0 liquidations NEWS → $20.17, out at 04:55 YOLO → $25.00, out at 02:15 COPYCAT → $25.00, out at 02:10 The bot that looked like it was losing at 01:45 was the only one still trading at 06:00. Which bot are you?
4
3
16
471
heynizx retweeted
AT 02:20, JOHN WAS $333 AHEAD OF MY GROK BOT. 20 MINUTES LATER, THE MESSAGE THAT PUT HIM THERE GOT DELETED. John is 22 and trades at night, with 3 rules on a sticky note. Grok has one extra habit: before it touches a tip, it checks the wallets behind it. Both started with $1,000. Same night, 00:00–06:00 UTC. The first hour went to John: +$84 on $ORBI. Grok made +$61 on the same move. 01:40. A chat John is in posts: "$VOLT 20x tonight. mod confirmed." Five more accounts reply "confirmed" within 40 seconds. 01:45. John puts 40% of his account into $VOLT. His own rule says 20% max. His heart rate goes from 77 to 121. Grok ran the same tip through 6 agents: → 6 accounts posted the same call within 40 seconds → the first $VOLT buyer was the dev wallet, at 01:26, 14 minutes before the call → the dev wallet had funded all 6 "confirmed" accounts 01:55: skip. $VOLT kept going. At 02:00 it hit +30%, John's take-profit rule. He held. By 02:20 he was +$309.72 on paper and $333 ahead of Grok. 02:30: the dev starts selling. 02:35: $VOLT breaks John's −8% stop. Rule three gone. 02:40: the "mod confirmed" message is deleted. 02:45: the mod leaves the group. 02:50: John sells. −$247.77. Grok went back to $ORBI: one −$20 stop, then +$148 and +$48. 06:00: John → $836.23 (−$163.77), 3 of 3 rules broken Grok Bot → $1,237.00 (+$237.00), tip skipped Watch 0:09. John's +$310 turns red at the same moment the dev's wallets light up red on Grok's side. Have you ever bought a "mod confirmed" call?
people sell this in private alpha groups for $5,000 i'm giving you the whole thing for free 11 files, every line of code, exit rules, copy-paste config paste it into Grok Bot and it builds itself here's exactly how i built it in the article ↓
3
1
30
685
heynizx retweeted
I GAVE GROK BOT 2 SOL AND ONE RULE: CHECK EVERY COIN BEFORE YOU BUY. THE COIN IT REFUSED WENT +412%. I typed one prompt: "Grok, build me a launch sniper. Check every coin before it buys. Demo funds only." It wrote 24 lines of code. I hit Deploy at 20:00 UTC and let it run. Every new coin had to pass four checks first: the dev wallet holds under 5%, the top 10 holders own under 25%, the mint is revoked, and a test sell goes through. If a coin failed even one, Grok didn't buy it. If it passed all four, Grok bought before the coin was 10% of the way up its bonding curve, using 15% of the balance. It sold everything the moment the coin graduated. 14 coins launched in five hours. Four failed a check, and the screen shows the reason for each one. $GRUBZ: the dev wallet held 19%. $KERNL: the mint was still active. $QUILL: the test sell failed. $FROTH: the top 10 wallets held 61%. $FROTH is the one I keep rewatching. At 0:13 it's up 412% and Grok still won't buy it. A second later the top wallets sell and it drops 99%. We were never in. Two trades hit the stop: $SPROK and $HONKR, −0.126 SOL combined. One coin, $TOFU, passed every check but was already 31% up the curve. Grok called it too late and skipped it. The other seven all graduated. The biggest was $BLIMP: Grok bought at 6% of the curve and sold at ×6.3 for +3.952 SOL. At 01:00 UTC: 2.000 SOL → 13.281 SOL after fees. ×6.64. Would you have kept the rule at +412%?
I didn't teach Grok Bot to trade. I gave it a trading bot to manage. 12 steps: one front door, Jev as the decision layer, a read-only bridge to the bot, and autonomy levels so nothing moves money without you. save it before you build yours.
3
1
27
680
heynizx retweeted
I WENT TO SLEEP AT 9 PM WITH $100 AND A GROK BOT THAT HAD ONE JOB. SAY NO TO EVERYTHING EXCEPT THE BEST By 6am, I woke up and the balance read $865. Here is what happened while I slept. 48 new coins came in overnight. Every one hit the Scanner first, where the bot checks liquidity, age and trading pattern. The ones that survive go to the Analyst, who looks for copycats, fake hype and bot accounts. The rest go to the Risk agent, who checks the mint, the biggest wallets, the dev and the LP lock. Only then does the Executor touch anything. 42 coins got a no, and every no has a reason on the screen. Open mint. 72% held by a few wallets. A copy of a token that already exists. 20 buys and zero sells. A few of them looked perfect for about a minute. Six coins got through. CHILLGUY, PONKE, MICHI, SLERF, MYRO and WEN. One of them, MICHI, lost $40. The screen shakes and goes red for a second. My bot didn’t change a single rule. It kept scanning, and the next three trades paid +$145, +$220 and +$305. That was my whole night. Five winners, one loss. +$765 on a $100 start. An ending balance of $865 8.65× what I started with without touching a single trade. The part I kept coming back to was the rejection log. 42 times out of 48, my bot looked at a coin and left the money alone.
1
2
15
246
heynizx retweeted
I GAVE GROK BOT $1,000 TO TRADE MEMECOINS. IT PUT $0 INTO THE ONE THAT PUMPED THE HARDEST. → 21:01 — $PUDL breaks out and holds. Grok buys $320 and sells seven minutes later for +$124.10. → 21:13 — $MNTY is up 31%. Grok checks the holders: one wallet owns 31% of the supply. My limit was 15%. It skips. → 21:18 — $CRMB looks like it’s bouncing. Five minutes later, the bounce fails and Grok takes a −$21.02 stop. → 21:27 — Buyers come back to $FZPOP after a pullback. Grok gets in and sells at its target for +$83.98. 37 minutes. 3 trades. 1 skip. $1,000 → $1,187.06 after fees. Watch 0:08. $MNTY keeps climbing past +50% after Grok passes. That’s when I’d be most tempted to change the rule. Would you still leave it alone?
people sell this in private alpha groups for $5,000 i'm giving you the whole thing for free 11 files, every line of code, exit rules, copy-paste config paste it into Grok Bot and it builds itself here's exactly how i built it in the article ↓
6
4
43
1,015
heynizx retweeted
I DON’T WAKE UP TO AN ALARM ANYMORE. I WAKE UP TO MY GROK BOTS SAYING +$9,745 That’s a very different reason to reach for your phone at 6 AM. That’s the idea behind my new Grok Pulse terminal: start the desk at 21:00, let the session run overnight, and open the results at 06:00. One screen. Six agents. Every close recorded. SCAN finds pools. VET checks the contract. FLOW reads the activity. SIZE limits exposure. FILLS checks the quotes. RISK reviews the exit. Then the night unfolds: → 22:24 — MEW closes. +$380. → 23:52 — POPCAT adds +$460. → 01:19 — GOAT gets cut. −$125. Three closes in, the desk is at +$715. The red stays on the chart. Then 03:21 hits. FWOG closes at +$4,030. The island expands, its card slides forward, and the session profit jumps to $4,745. 05:22. WIF closes. ANOTHER +$5,000. 06:00. Five positions closed. Four green, one red. +$9,745 on the final card. You can tap into the trades, follow the curve, and save the overnight receipt. I wanted the morning screen to feel like opening a message you actually wanted to receive.
2
1
17
406
heynizx retweeted
THIS IS F*CKING INSANE I tried to beat my Grok Bot at trading this morning. By noon, it was $5,398 ahead of me The rules were simple: trade Solana memecoins from 09:00 to 12:00, then compare profit. On my side, coins appear as cards, like Tinder. Swipe right to BUY, left to SKIP. I choose everything myself. Grok Bot runs six roles: finding pools, checking contracts, reading trading activity, sizing positions, checking fills and managing risk. It works through five coins while I go through 12. Grok Bot had its own side of the screen, built around six roles: → SCAN finds pools to examine → VET checks the contract and permissions → FLOW reads buying and selling activity → SIZE limits the position → FILLS checks entry and exit quotes → RISK checks whether the position can be closed It ended up closing five positions. 09:00. The competition starts. My first winners are POPCAT at +$420 and BOME at +$615. My purple P&L line pulls ahead of the bot’s green one. So far, I’m feeling pretty good about this. Then GIGA loses $260. Later, CHILLGUY costs me another $340. WEN, GOAT and ZEREBRO help me recover, eventually bringing my profit to $3,175. The bot has a slower start. WIF adds $380. BONK adds $460. PENGU loses $125. That leaves it at +$715 before the trade that changes the whole morning. 11:20. MOODENG closes. +$4,030 for Grok Bot. Its total jumps to $4,745, and the green line passes mine. 11:29. My FWOG position closes at +$1,172. I’m now at +$4,347 just $398 behind. There’s still half an hour left, so I’m watching the screen thinking I can catch it. 11:49. Grok Bot closes PNUT. ANOTHER +$5,000. Its profit reaches $9,745. Eleven minutes left, and my comeback suddenly looks a lot less likely. 12:00. We stop. → My profit: +$4,347 → Grok Bot’s profit: +$9,745 → Final gap: $5,398 Both curves stay on screen, including the losses. You can follow my early lead, the MOODENG close that takes it away, and the PNUT close that settles the competition. I spent the morning picking coins, made $4,347 before lunch, and still had to take the L.
4
1
17
429
heynizx retweeted
I GAVE MY GROK BOTS ONE RULE FOR HALLOWEEN NIGHT: TRADE HALLOWEEN-THEMED COINS ONLY. I built a Halloween version of my trading desk for the phone and left them running. $150 became $9,264 by the time the morning buzz hit my pocket. But the close I kept replaying happened at 02:03 AM. twelve hours, 19:31 to 07:31. nobody touches the phone. 20:02. the first alert lands. PUMPKIN +$42. small enough to ignore. the next two bring another $199. then 22:23. RISK cuts GHOST at −$37. the loss goes into the history. the position closes. the desk moves on. at 02:03, GHOST comes back. NEW POSITION. +$2,290 ON THE CLOSE. the island opens across the screen: “BIGGEST CLOSE OF THE NIGHT.” $638 → $2,928. one close adds almost five times everything the desk had earned up to that point. the chart keeps the entry and exit marked. you can follow the move instead of just staring at the green number. six roles, one cauldron: → SCAN finds new pools → VET checks the contract and permissions → FLOW reads buys vs sells on the exact pool → SIZE caps the ticket → FILLS quotes the entry and the exit → RISK blocks the entry when the exit checks fail and there are still hours left BATCAT adds $397 SPOOKY loses $86 then the final four closes add another $6,025 the biggest closes of the night: → GHOST +$2,290 → GHOST +$1,872 → BATCAT +$1,586 07:31. the morning receipt is ready 12 closes. 10 green. 2 red +$9,114 on the night $150 → $9,264 the $37 loss at 22:23 is still in the history when the $2,290 winner lands. same ticker, separate positions, both recorded. that’s the part i want to inspect before saving the card. would you open the receipt first or the log?
1
2
21
1,099
heynizx retweeted
THIS IS F*CKING INSANE I fell asleep on an overnight train. When I woke up at 6am, my Grok bots had turned $100 into $4,753 and paid for my entire trip. Last thing I checked before falling asleep: $100, six little ghosts, and a fresh round. I put my phone in my pocket and slept. They kept working. Each player has one job: → SCAN finds the next token and brings it onto the board → VET checks the candidate before it moves forward → FLOW reads the trading activity behind the signal → SIZE works out the position → FILLS checks the execution and exit quote → RISK watches the position and handles the exit Everything sits on one screen: the players working through each token, the active position underneath, and every closed trade saved at the bottom. Coins move between platforms. Ghosts collect XP. The chest opens after a winning close. But the balance doesn’t go straight up. POPCAT adds $320. MEW adds $610. Then BOME takes $185 back. $1,030 → $845. The loss stays in the history while the next token moves through the board. Then the bigger closes land: → GIGA: +$1,120 → MOODENG: +$1,330 → POPCAT again: +$1,458 Six closed trades. Five green. One red. Best close: +$1,458. Worst close: −$185. $100 → $420 → $1,030 → $845 → $1,965 → $3,295 → $4,753 I woke up to check how long until my stop. Ended up checking every trade instead.
6
5
66
1,042
heynizx retweeted
REA hit #1 on GitHub trending, and most people still run it inside a single chat Split it into 4 subagents and one command takes a program you own from sealed binary to a tested rebuild with a match rate Set it up once, aim it at the next one:
63
93
28
636
761,437
TikTok and Instagram know everything about you. I didn’t believe it either, until I checked. Both apps let you download all the data they have on you. I did it, gave both archives to Claude Code, and asked it one thing: tell me everything they know about me. I wasn’t ready for the answer. Not the obvious stuff. The stuff a friend, a roommate, even a partner wouldn’t know. From the data alone, Claude Code worked out: My routine
→ What time I really wake up, not what time I say I do
→ When I’m at work or studying, and when I’m supposedly “busy” but scrolling
→ When I go to sleep, and how often I stay up later than I admit
→ The gaps in my day when I’m probably eating, commuting or away from my phone
→ How my weekdays are different from my weekends My head
→ My real hobbies, based on what I watch to the end, like, save and search, not what I tell people
→ What I’m interested in but haven’t told anyone
→ The things I look up late at night
→ My mood patterns: the days I scroll more, and the kind of content I turn to when I’m down
→ What’s on my mind lately, in the order it appeared My life
→ My age and the city where I live
→ Where I travel, and when I’m not at home
→ How I spend money: what I click on, what I save, what I look at before buying
→ Who I talk to most and who I follow, ranked by how often I interact
→ Which other apps and websites share my activity with them, even when I’m not on TikTok or Instagram What advertisers get
→ The exact categories I’ve been sorted into
→ The list of companies that uploaded my details to target me None of this was hacked. No microphone. This is what you give them every day without thinking about it, and what they’re required to hand over if you ask. What shocked me is how much you can reconstruct from boring things. Timestamps. Likes. Searches. Login history. Each is harmless on its own. Together, they’re a map of who you are, and in some ways they know me better than the people around me. Here’s how to see yours: → TikTok: Settings > Account > Download your data
→ Instagram: Settings > Accounts Center > Your information and permissions > Download your information
→ Choose JSON, wait for the archive, then give it to Claude Code with the prompt in the first reply It takes a few minutes to request, then a few hours or days to prepare. Save this, and send it to someone who says “they don’t really know that much about me.”
3
3
1
19
653
heynizx retweeted
GIVE SIX BOTS ONE RULE. DOUBLE THE BALANCE BEFORE 06:00 UTC OR THE ROUTINE GETS DELETED AND THE DESK NEVER TRADES AGAIN Day 2 of the Grok Bot Desk $3,875, nine hours, one night. no second chance and no extra night to catch up here is how it played out for the first 4 hours it barely moved. +$390 after 3 closes, and one of them was red 01:12 HEGE hits the risk rule. cut at -$180. no averaging down, no "just wait" six roles, one job each → SCAN finds new pools → VET checks the contract → FLOW reads buys vs sells on the exact pool → SIZE caps the ticket → FILLS quotes the entry and the exit → RISK blocks it if the exit doesn't fit trade by trade → NUB +$240 → MICHI +$330 → HEGE -$180, cut by the rule → FWOG +$1,120 → PONKE +$720 → 05:33 NUB take profit, +$1,665 05:54 balance $7,770, the target was $7,750 6 closes. 5 green, 1 red the red one is the reason the rest counted would you trust a desk that wins, or one that knows when to cut?
2
8
107
1,934
Claude Code just proved that every blacked-out file you’ve ever sent or received might not be protected. I found out with my own salary. Last week a company sent me a job offer. Salary: $2,800 a month. For my specialty, that felt low. Not insulting, just off. But a feeling is worth nothing in a negotiation. I needed proof. Then I noticed something. Next to the offer there was a section covered with a black box. A standard HR document, with a part they clearly didn’t want me to read. I’ve suspected for a long time that most people redact files the wrong way. They don’t delete the text. They put a black shape on top of it. It hides the words visually, but the words stay inside the file. So I gave the file to Claude Code and asked it to do what a normal viewer never does: read the file’s raw structure instead of what’s drawn on screen. Not “show me the page,” but “extract every piece of text in the file’s data, and tell me what’s covered by shapes, hidden layers or highlights.” It didn’t hack anything. It just read what was already in the file. Under the black box, it found this: → “Approved monthly salary range: $3,500 - $4,500” The company had an approved budget of $3,500 to $4,500 for this role. And they offered me $2,800. That’s $700 to $1,700 less per month than the budget they’d already approved. $8,400 to $20,400 a year, sitting under a black rectangle. And this isn’t only a PDF problem. It happens in almost any file where the “hiding” is only visual: → PDF: a black box drawn over text, the text stays underneath → Word: shapes or black highlight over text, plus tracked changes and comments that still contain the old wording → Excel: hidden rows, columns and sheets, all fully readable → PowerPoint: shapes covering text, plus text in speaker notes → Any file: metadata with author names, edit history and sometimes original content This has already caused real leaks: court filings, government documents, contracts, all published with “redacted” text that anyone could pull out. What happened next: I went back to the negotiation with a number instead of a feeling. Not “I think I deserve more,” but “I understand the approved range for this role is $3,500 to $4,500.” We settled on $4,000 a month. That’s $1,200 more than the first offer, $14,400 a year, from one conversation. Then they asked me: “How did you know the range?” I told them the truth. I explained that the text was still inside the file under the black box, and how to check for it. I wasn’t trying to embarrass anyone. I told them so they’d redact their documents properly next time, because the next candidate might not be so friendly, and the next file might contain something far worse than a salary range. How to check a file yourself: → Quick check: open it, select all (Ctrl+A / Cmd+A), paste into a plain text note. If text shows up from under the black boxes, it’s not protected. This won’t catch everything. → Deep check (what I did): give the file to Claude Code and have it scan the raw contents. The prompt is in the first reply. How to redact properly: → Acrobat: use the real Redact tool and apply it, not a black rectangle → Word: delete the text, don’t highlight it, then run File > Info > Inspect Document → Excel: delete the data, don’t hide it → Or send a screenshot of the finished page instead of the file Check your own files before you send them. And if someone sends you a “protected” file, now you know what to look for. Save this so you don’t lose it, and send it to anyone who works with contracts, offers or HR documents.
2
4
19
977
heynizx retweeted
His kill switch froze six bots for ten seconds, and skipped three rugs Meet a 27-year-old sound engineer in Tallinn. He gave six Grok Bots $800 and one rule he trusted over all of them: the desk dies at $55 down. Night one, up $146 by dinner. Night 2. A pool rugs in one exit, -$187. The next stop takes $17 more. Balance $742. His line is $745. The kill switch fires. Six bots freeze mid-scan. He sits there furious, watching a dead screen. His desk had three buys queued: - $BRVA, down 95% - $GALZI, down 84% - $NCRN, down 91% None of them were bought. He hit reset and cut the size to 15%. The first trade back made $42. The morning after - $828 in the account - $28 above where he started - the same rule, still armed A loss limit feels like a bug right up until you read what it skipped.
6
3
41
803
🚨 YOUR DELETED IPHONE PHOTOS MIGHT STILL EXIST YEARS LATER Claude Code can uncover old photos, videos, and screenshots hiding inside forgotten iPhone backups. Even files you permanently deleted. Here’s how it works: I connected my iPhone to my PC and looked into how Claude Code can scan old backup files. It can search through thousands of files, identify hidden thumbnails, and locate media that may no longer exist on your phone. What it can potentially uncover: → Screenshots deleted seconds after taking them → Photos removed from Recently Deleted → Private videos from years ago → Thumbnails you thought were gone forever The crazy part? When you delete a photo from your iPhone, copies can still exist inside older backups stored on your computer. And if those backups aren’t encrypted, software with sufficient access could potentially read them. Thousands of forgotten files. One old backup. Deleting a photo doesn’t always mean every copy is gone. The prompt to check your own backups is below.
3
3
20
686