Pinned post
Latest posts 5
Holy sh*t... Fable 5.2 just turned Claude Code into a 24/7 hedge fund
send this prompt before everyone else does
thank me later:
send this prompt before everyone else does
thank me later:
Fable 5.2 is UNREAL as a brain for teams of agents...View quoted post on X
1 orchestrator + 3 cheap workers = a team that runs for days
In this article, I show you how to set it up tonight: x.com/i/article/2120…
Damn... AQR - Cliff Asness’s quant fund - put 137 years of trend-following into one 16-page PDF
→ tested on 67 markets back to 1880 (they hand-copied Chicago futures data from 1877)
→ positive in all 14 decades
→ made money in 8 of the 10 worst 60/40 crashes: Great Depression, Stagflation, 2008
now send it to Fable 5.2 / Claude Code:
1 → upload the PDF
2 → say: “rebuild the 1m/3m/12m time-series momentum strategy from this paper on my tickers, vol-target it”
3 → say: “backtest 2000–today vs buy & hold, show every drawdown >10%”
4 → the step everyone skips: “now show me 1987 and 2022, where it struggled”
bookmark it. thank me later:
→ tested on 67 markets back to 1880 (they hand-copied Chicago futures data from 1877)
→ positive in all 14 decades
→ made money in 8 of the 10 worst 60/40 crashes: Great Depression, Stagflation, 2008
now send it to Fable 5.2 / Claude Code:
1 → upload the PDF
2 → say: “rebuild the 1m/3m/12m time-series momentum strategy from this paper on my tickers, vol-target it”
3 → say: “backtest 2000–today vs buy & hold, show every drawdown >10%”
4 → the step everyone skips: “now show me 1987 and 2022, where it struggled”
bookmark it. thank me later:
View quoted post on X
this is f*cking INSANE... the stack two sigma quants worked on is sitting on github for free
pandas was led for years by a Two Sigma managing director. Lightweight Charts is TradingView's own chart engine. TradingAgents is an AI trading desk that hit #1 on GitHub trending
together they have 189,000+ GitHub stars. and now the whole thing runs inside your Claude Code for $0
first, get how it works:
1. TradingAgents (121K stars) - the brain: AI analysts, a bull vs bear debate, a trader and a risk team. it says BUY, HOLD or SELL and gives the reason
2. pandas (50.5K stars) - the memory: every price, every call and every dollar of P&L lives in its DataFrames. "fast, flexible, and expressive data structures"
3. Lightweight Charts (17.8K stars) - the eyes: "one of the smallest and fastest financial HTML5 charts", so every BUY and SELL lands on your screen
now the setup:
1 → install Claude Code: npm install -g @anthropic-ai/claude-code, open a new folder and run claude
2 → paste the 3 links (below) and say: "install TradingAgents and pandas in a Python env, and lightweight-charts with npm"
3 → the brain: "connect TradingAgents to Claude and run it on NVDA for every Friday of the last 12 months. save each BUY, HOLD or SELL with its reason to a CSV"
4 → the memory: "load that CSV with pandas and replay the calls: enter on BUY, exit on SELL, 0.1% fee plus slippage. compare it with just holding NVDA"
if it loses to just holding, that's your answer
5 → the trap: "check the model's training cutoff and rerun only on dates after it. if the result collapses, the AI was remembering the news, not reading it"
this is the step everyone skips
6 → the eyes: "build a page with lightweight-charts: NVDA candles, a green marker on every BUY, a red one on every SELL, equity curve underneath"
7 → only if it survived steps 4 and 5: "run it every Friday on paper and send me the call with its reason"
paste links on the repos below:
pandas was led for years by a Two Sigma managing director. Lightweight Charts is TradingView's own chart engine. TradingAgents is an AI trading desk that hit #1 on GitHub trending
together they have 189,000+ GitHub stars. and now the whole thing runs inside your Claude Code for $0
first, get how it works:
1. TradingAgents (121K stars) - the brain: AI analysts, a bull vs bear debate, a trader and a risk team. it says BUY, HOLD or SELL and gives the reason
2. pandas (50.5K stars) - the memory: every price, every call and every dollar of P&L lives in its DataFrames. "fast, flexible, and expressive data structures"
3. Lightweight Charts (17.8K stars) - the eyes: "one of the smallest and fastest financial HTML5 charts", so every BUY and SELL lands on your screen
now the setup:
1 → install Claude Code: npm install -g @anthropic-ai/claude-code, open a new folder and run claude
2 → paste the 3 links (below) and say: "install TradingAgents and pandas in a Python env, and lightweight-charts with npm"
3 → the brain: "connect TradingAgents to Claude and run it on NVDA for every Friday of the last 12 months. save each BUY, HOLD or SELL with its reason to a CSV"
4 → the memory: "load that CSV with pandas and replay the calls: enter on BUY, exit on SELL, 0.1% fee plus slippage. compare it with just holding NVDA"
if it loses to just holding, that's your answer
5 → the trap: "check the model's training cutoff and rerun only on dates after it. if the result collapses, the AI was remembering the news, not reading it"
this is the step everyone skips
6 → the eyes: "build a page with lightweight-charts: NVDA candles, a green marker on every BUY, a red one on every SELL, equity curve underneath"
7 → only if it survived steps 4 and 5: "run it every Friday on paper and send me the call with its reason"
paste links on the repos below:
0:26
View quoted post on X
paste this into Grok Bot tonight and go to sleep. you'll wake up to a trading plan...
while you sleep, it reads all of X, news and filings, catches the narratives before they hit price, backtests every idea and paper-trades only what survives.
steal it before your timeline does:
while you sleep, it reads all of X, news and filings, catches the narratives before they hit price, backtests every idea and paper-trades only what survives.
steal it before your timeline does:
View quoted post on X