Bercus (@BercusNA) header
Bercus (@BercusNA) profile picture

Shipping AI agents in public. Real stacks, real prompts, real invoices.

Joined October 2026
Photo by Bercus (@BercusNA)Photo by Bercus (@BercusNA)Photo by Bercus (@BercusNA)Photo by Bercus (@BercusNA)Photo by Bercus (@BercusNA)Photo by Bercus (@BercusNA)Photo by Bercus (@BercusNA)Photo by Bercus (@BercusNA)Photo by Bercus (@BercusNA)Photo by Bercus (@BercusNA)
Bercus retweeted
19 year old Japanese student built a trading bot with Claude Code in 2 days. Used his iPad as a second monitor. First night: $6,732 profit. Starting capital: $68. Total profit so far: $750,000. [ ๐๐จ๐ญ๐ž: ๐…๐จ๐ฅ๐ฅ๐จ๐ฐ ๐Œ๐ž @monicaa_AI ๐…๐จ๐ซ ๐ข๐ง๐ฌ๐ญ๐š๐ง๐ญ๐ฅ๐ฒ ๐š๐ฎ๐ญ๐จ ๐ƒ๐Œ] Here's how it works๐Ÿ‘‡ The bot scans over 50 markets simultaneously. Syncs live BTC data from Binance every second. Spots price errors before humans even notice. The edge is pure speed + pattern recognition. While traders stare at charts trying to predict the next move, his bot is already executing on mispricing across dozens of markets. No guessing. No emotions. No hesitation. Just Claude Code logic finding gaps that close in seconds. He built the entire system in 48 hours: โ†’ Claude Code handles the trading logic โ†’ Binance API feeds real-time BTC data โ†’ iPad displays multi-market monitoring โ†’ Executes trades when arbitrage windows open The system runs 24/7. Every price dislocation = profit opportunity. Most people are still trading manually, refreshing charts, second-guessing entries. Meanwhile this 19 year old engineering student turned $68 into $750K by letting Claude Code do what humans can't: process 50 markets instantly and execute without fear. Why are people still trading manually? I'm giving away the exact Claude Code setup for free. 24 hours only. To get it ๐Ÿ‘‡ 1๏ธโƒฃ Comment " Claude " 2๏ธโƒฃ Like and Repost 3๏ธโƒฃ Follow @monicaa_AI (so I can send it via DM)
246
199
2
782
76,055
Bercus retweeted
THIS IS F*CKING INSANE, I CASHED OUT $60,000 AND SLEPT, MY GROK BOTS SOLD A TRUMP COIN 90 MINUTES BEFORE IT CRASHED 78%, $1,873 IS NOW $8,977 in 3 nights Jev + my Grok bots turned $100 into $61,873 the replies said nobody ever cashes out, it's just numbers on a screen so yesterday i moved $60,000 to my bank and left them the change, $1,873 if they're actually good, the change is enough the coin: VeilTrump, a Trump meme coin, 1 day old, $866K at 1pm the night: โ†’ 1:35pm first trade made $45, i almost turned them off โ†’ 3:20pm the dip to $714K, bought, sold $1.71M an hour later, +$2,400 โ†’ 7:05pm $1.32M to $2.10M, +$2,615 โ†’ 10:30pm skipped a coin up 218% in 40 minutes, its pool got emptied at 11:30pm โ†’ 12:25am $1.27M to $1.42M, +$777 โ†’ 5:00am sold at $1.10M, then all cash the morning: โ†’ 6:30am the coin fell from $1.03M to $230K in 25 minutes โ†’ my bots had been out for 90 minutes $1,000 held in that coin is $265 my $1,873 is $8,977, 4.8x on a coin that lost 73% honestly i laughed at the 5am exit until i saw 6:30am at 0:09 my spider locks the $60,000 in a silk safe at 0:24 the coin falls off the screen, my bots already in cash $100 โ†’ $61,873 โ†’ $60,000 to my bank โ†’ $8,977 still working every prompt and the full bot setup are in the article i quoted below โ†“ free they trade on one exchange, it's my first reply, sign up there and the bonus is yours bookmark this, one morning your coin drops 78% before you're up send it to the friend who never sells would you have cashed out the $60,000 or let it ride?
48
47
3
383
92,138
Bercus retweeted
It is becoming f*cking dangerous: Anthropic documented every piece of this separately, and nobody noticed what they add up to 1. Create a folder on your PC - this is the only world your agent will ever see 2. Launch Claude Code inside it and drop a CLAUDE.md in the root with your rules - it loads into every single session, you never paste context again 3. Create a MEMORY folder and add one rule to CLAUDE.md: write every durable fact there, read it before planning anything 4. Create .claude/agents/ and define three read-only Haiku 5.5 subagents - they explore the files, your main model never burns context on the search 5. Give it a real browser: claude mcp add playwright -- npx -y @playwright/mcp@latest 6. Connect Gmail, Calendar and Drive at claude.ai/customize/connectoโ€ฆ - they load into the CLI automatically on the same account 7. Wrap the whole chain in .claude/skills/<name>/SKILL.md so the entire thing runs as one command: /<name> the whole setup takes around 20 minutes now the part nobody wires up. every one of those steps has the same slow spot: a tiny fork that doesn't need a thinker. which file. which tool. retry or stop. keep this chunk or drop it. that's where Jev fits. it can't write a word. it returns the choice with a probability in 70 to 500 milliseconds, at $0.042 per million input tokens, output free. that's thousands of decisions for the cost of one LLM call. > let Claude think. > let Jev decide what deserves Claude. > let code do the actual work. what you end up with: the full Anthropic stack running on your own machine - persistent memory, a swarm that explores in parallel, a real browser, your mail and drive, and Jev making the thousands of tiny calls in between the step-by-step setup is in the article below. save this before your next session.
REA is the most dangerous repo on GitHub right now It gives your AI agent a real disassembler, so it reads any app down to the binary Every claim comes back with the address behind it I broke down the exact 8 steps:
35
85
1
773
113,681
Bercus retweeted
this is f*cking wild: you still give 1 agent 1 big task. A dev gave 12 Anthropic agents 64 papers and 4 jobs. The result: a lab you run in a browser. Change the setup. Watch the result move. They split the work: Readers pull the equations and settings, so builders start from clean inputs. Builders turn each method into code, so every method runs on its own. Reviewers check each result against the paper, so a sim that only looks right gets caught first. Interface agents build the controls, so you change one thing and see what happens. Per the post: 11 methods, 28 reference scenes. 7 matched the reference results. 4 stayed flagged for review. Try it today: start with 1 paper. Get 1 result right. Then add agents. In my video, the spider does the same with 2 models. Astra searches wide. Opus checks deep. The spider collects it all. The point is not more agents. It is clear jobs and a reviewer. You hire more agents. Then you skip the reviewer.
REA is most dangerous thing on the internet right now... It lets your AI agent reverse engineer any app, down to the binary If you use it correctly, you can rebuild any feature you pay for, for FREE In this article, I show you how:
9
26
1
219
15,771