Bounce

Nuclear def looking juicy. $SMR $XE $OKLO. Longer term sizing.

@TheShortBear Feel like reactions market wise are becoming more muted with the on again offf again.

The constant contradictory news on the Iran/US war is likely mostly to shift the economical scale. Each time the US a near-deal or good news in regards to the war, Oil slides and bond yields move lower. Each time Iran goes back on previously agreed to terms the opposite happens. The US needs lower prices to be the strongest in negotiations, the opposite is true for Iran. Both are lying, both know it, both want a deal, both are pressing for the best terms and maximum leverage. It doesn't mean there will be a deal in the end, but this is to be expected.

@s6berr A Brave New World.

Clavicular instantly FELL IN LOVE after meeting a girl who is just as ANXIOUS as he is 😳👀 "Oh, don’t worry about it, yeah. I get it. I anxiety-max all the time." "I don't know what it's called. I use it whenever I have a presentation so that I don't like..."

@VynthraQuant @HeyZaraKhan You can’t use AI to backtest historicals- look ahead bias is massive.

This is exactly how you create massively overfit strategies. AI does not discover new market principles. It remixes existing ideas into something that looks statistically convincing. That often leads to indicator soup, hidden factor overlap, and fragile equity curves. Where are the: walk-forward tests? out-of-sample validation? regime robustness checks? slippage/fee stress tests? leakage/lookahead audits? A smooth backtest means nothing without robustness testing. The biggest danger of AI in trading is the illusion of progress. Complex logic and pretty equity curves do not equal real edge.

🚨 BREAKING: Claude Cowork can now build hedge fund–level trading strategies (100% free). Here are 12 prompts to backtest strategies, analyze risk–reward, and find high-probability trades: (save it for later!)

The market has been undergoing a stealth correction over the past week, with the equal-weight S&P 500 on pace for its sixth consecutive down day. This is the longest losing streak since December of 2024.

@NTFabiano Use the oil and gas industry as a control group. Sometimes working 12-14 hr days for weeks on end . High level execution required.

Sleeping <6h a night for 2 weeks reduces cognitive performance equal to 2 nights of total sleep deprivation.

@TraderDiegoX @tradingview @grok Anytime- when I’d get stuck in a loop I’d flip the entire code to GPT, visa versa- Claude def strong for coding and those challenges 1.5 years ago probably resolved now with Claude

@M_Anthony86 @tradingview @grok Yeah, what I've seen is that initially is a matter of feedback so it can fix small mistakes. Will try other AIs down the road to compare. Appreciate the input bro.

Claude can now code @TradingView indicators for you! I've always wanted an indicator with all my key intraday levels in one place, the way I wanted to see them. I won't stop until my entire system is coded. Learn how to start creating yours here👇

@TraderDiegoX @tradingview @grok *pinescript

@TraderDiegoX @tradingview @grok There was def debugging but just fed the errors back to grok - but if you know what you’re looking for you can code a lot into conscript via AI- you don’t need anything special.

@MikaPalmy @joemccann @IBKR It’s what I use for bulk data pulls. So no we aren’t joking. What do you use? Open to better options.

@M_Anthony86 @joemccann @IBKR Polygon? Are we joking

.@IBKR has loads of historical data, cross-asset and is free if you have an account. I even made an open source repo to extract it and store locally in parquet format; you can query with DuckDB. github.com/joemccann/mark…

@Pps831 @joemccann @IBKR Ahh ok - I use IBKR for the live executions the polygon pulls were for the backtesting/bridging the gaps. I was having massive problems populating 1700+ tickers with 5 years of data of 1m bars. I like IBKR as they are direct access for the live.

@M_Anthony86 @joemccann @IBKR Same. I use polygon (renamed to massive). I hear databento is also good, but haven’t tried it. One thing I don’t like with polygon - their live 1second data stream comes 2 seconds late to me. (They delay it for 1 sec for some weird reason)

@stensoosaar @joemccann I was getting troubles from the 1700 tickers I’ve accumulated spreadsheets on manually - getting historicals on that size I was getting stuck, I spread out my requests etc and wasn’t getting anything in 1m bars past 60d. Just check IBKr campus now it’s showing they lifted limits.

@M_Anthony86 @joemccann Just ask. 60d limit applies for 30s and smaller bars. Just keep the chunk size below 2-3k datapoints and pacing under control.

@iMATTHEWRYAN @joemccann @IBKR Yeah horrible rebrand lol - used to calling it polygon.

@M_Anthony86 @joemccann @IBKR *Massive (horrible rebrand btw)

@joemccann @IBKR I’m just looking it up as it’s been a while since I switched but it’s saying they’ve lifted some of the caps so maybe you can go deeper now.

@M_Anthony86 @IBKR 1 minute? Sorry I misread. I didn’t fetch that data

@joemccann @IBKR Yeah you have 5+ years of 1m bars? I’m curious how you did it? Pretty sure it’s a 60d limit?

@M_Anthony86 @IBKR I literally have all historical data from @IBKR

@jeffscottward @joemccann @IBKR Yeah I’m with IBkR and decided to use their api for historical data but it’s massively limited and won’t give you good granular context for intraday. — ends up paying for a subscription to Polygon.io for 5 years of historical.

@joemccann @IBKR Ahhh

@Connor_LFC8 I just tore my pcl mcl and meniscus planting for a shot- wasn’t sure what was wrong with my knee a lot of shinbone pain went back out scored a goal then had to come off . Knee had no strength planting. Didn’t realize I had tears but knew something was wrong.

@hackertrader Nice work- agree the only way to avoid look ahead and bullshit from the llm is to build an engine that doesn’t require the llm for executions. I’ve done the same thing for my Strats that I used from a data driven human execution perspective and now want to automate it all.

I keep seeing people backtest with Claude/ChatGPT and it's painful to watch. LLMs will: → Invent price data to satisfy your prompt → Look into the future → Ignore survivorship bias → Make wrong assumptions about splits & adjustments They will make your backtest looks great but pnl will probably not follow. I built Kwants to fix this. It gives your LLM a real backtesting engine: • Downloads & cleans your Polygon data automatically • Enforces strict point-in-time data access • Ignores late prints • Runs fast with vectorization + caching • No hallucinations allowed at the infrastructure level No code required. Works with any LLM. Try it now: kwants.dev