Wesley Chuen

[Wesley Chuen @wesleychuen]

A lot of people claim to have models that can predict stock prices accurately, but I haven’t seen any that hold up under real market conditions.

I recently came across a 2024 research paper reporting over 90% accuracy using an LSTM model on daily VN-30 index data.

The setup included:
• Indicators: SMA, MACD, RSI
• 60-day sliding window
• Four LSTM layers (30–60 units each)
• ReLU activation, 0.1 dropout
• 1,000 training epochs in TensorFlow and Keras

The paper doesn’t say if that 90%+ figure came from the training set, test set, or true forward predictions.

With only a few months of test data, that number might not hold up once market volatility kicks in.
I’m planning to replicate the setup using daily data to see how it performs in real conditions.
Will compare training and test accuracy against actual market prices.

Should be an interesting little side project.

Source: Nguyen et al. (2024), Humanities and Social Sciences Communications.

[Ken Kirtland IV @KenKirtland17]

Here's some more of his banger takes from the same panel

[AwakenedOutlaw⚒️ @AwakenedOutlaw]

Pfffft... A trillion dollars. Please.
That's like two daycares and a hospice.

[Wesley Chuen @wesleychuen]

@TeslaBoomerMama Schwab.
Requested 600 got 26
Requested 300 got 18
Requested 200 got 15

[Wesley Chuen @wesleychuen]

@vdruts @eCommKen @openclaw Same here. Spent lot of time over the weekend to revive mine by changing it back to .23. .25 did the same thing. Not going to upgrade until I hear this is fixed. Staying at .23 for now.

[Wesley Chuen @wesleychuen]

@XFreeze How it compare to kimi on pricing

[JoNation @JoNationLive]

Picked up a pizza order. As I walked out, an elderly couple was getting out of their new Tesla Model Y. I said, “Beautiful car.” He said, “Thank you. It’s the best car I’ve ever owned.”

I said, “Why?”

He said, “You see me—I’m 78 and can barely walk, let alone drive. I’m taking my wife out for date night again after 52 years of marriage. We couldn’t do this with our old car. I didn’t buy this to save the planet. I bought it to save me.” 😢

All choked up, I said, “Thank you for sharing. You two go enjoy your dinner!”

I see and hear this so often from elderly people buying Teslas, but it never gets old. Tesla is freedom🫶🏻

[Zara Zhang @zarazhangrui]

Almost every AI power user I know is MORE stressed and busier after using AI, not less

What people thought AI would do: 10x productivity so that we can finish work earlier & relax more

What it’s actually doing: 10x productivity so that we end up with 20x more things to do cos of the sheer possibilities

[ΛI DRIVR @AIDRIVR]

after this, I will shut up. just please:

put yourself in the shoes of a customer that:

- bought Model S/X in 2022/2023 (~$140k for a Plaid), or even 3/Y at ~$60-75k

- paid an additional $15k for FSD (12 ½ years worth of subscription at current pricing)

- doesn't have access to latest FSD versions because HW3 (missed cutoff point by a few months)

- can't justify upgrading to HW4 because they’re deeply underwater on their current car

- see $15k FSD purchase option drop by nearly half, even though Tesla said it was only going to get more expensive

- see a great deal on a Cybertruck with "Free FSD Transfer for loyal customers!" advertised right below the Order button. press "Order."

- aren’t chronically online, so miss Tesla quietly changing the terms retroactively

- show up (or are about to show up) for delivery in June only to be told “sorry, we’re not honoring that”

- then went online to figure out what happened… and got hit with “stop wanting things for free!”

these customers exist, and will never buy another Tesla again.

[Wesley Chuen @wesleychuen]

@Loopify @Eli5defi works fine here as well. If they ban me, I'll switch to other providers, probably will cost less.

[Wesley Chuen @wesleychuen]

@steipete @openclaw Ever since version 2026.2.12, all my full agents except the main one won’t response in discord channels anymore. I asked the main one to try to fix and nothing worked. Tried .2.13 and same thing. Had to downgrade it back to .2.9. I’m afraid to try the new versions now.

[Wesley Chuen @wesleychuen]

@unusual_whales Nice! Gotta try it out.

[Wesley Chuen @wesleychuen]

@jandatesla Yeah, it does that. We ended up in a ditch from our last ski trip. Never again.

[Wesley Chuen @wesleychuen]

@_MG_ @lauriewired Reminded me to bring a solar USB charger with me. 😁

[Wesley Chuen @wesleychuen]

@LeonXDreal @EliteDangerous Yes, I've stopped playing Elite dangerous for quite awhile now. What ship is this?

[Wesley Chuen @wesleychuen]

@AlejandroEV66 and taking up 2 spots of the charging space...

[Chris Bakke @ChrisJBakke]

When I reported to Elon, we would have X product meetings at 5 or 6pm (because he was with the Tesla team for 10 hours before that), then my 1:1 checkin would be at 10pm, but that would regularly get moved to 11pm or midnight or 1am.

Usually around 2am, he’d go take a nap for a couple hours in the office and then repeat the same schedule again the next day with a different set of companies, 7 days a week.

[Wesley Chuen @wesleychuen]

The first round of my AI stock prediction project wasn’t as accurate as I hoped, and it’ll take some time to refine and improve it.

In the meantime, people have asked what else AI can do with the market. So I built a Tool Augmented Generation setup using OpenAI GPT with custom APIs that let the model access live data.

It pulls technical indicators such as EMA, RSI, MACD, ATR, and volume, then combines them with company fundamentals like earnings growth, ROE, profit margins, and leverage.

The model interprets both together to give an AI-driven view of a stock. The example below is GOOGL.

*This is for information and experimentation only, not financial advice.

#AI #MachineLearning #Finance #StockMarket

[Wesley Chuen @wesleychuen]

Applying AI to the stock market – Part 2: Reality check

I spent the weekend replicating a 2024 deep learning model that claimed 90 %+ accuracy predicting stock prices using LSTM networks. The results were nowhere near that high.

Training performance
LSTM – Train 50.5 % / Test 50.4 % accuracy (essentially random guessing)
Random Forest – Train 90.4 % / Test 89.1 % accuracy (for comparison)

Confusion matrix results (Test set)
Model Accuracy TP_up TN_down FP_up FN_down
LSTM 50.36 % 48 21 33 35
Random Forest 89.05 % 71 51 3 12

I also built an automated hyperparameter tuner to optimize the learning rate, batch size, validation split, patience, recurrent dropout, and optimizer selection. Trying not to touch the other hyperparameters that the paper used yet. None of them moved the LSTM much above 50 %.

Just like I originally thought, using AI and deep learning models looks convincing in papers, but I really have my doubts when applying them to a U.S. index like SPX that’s influenced by so many economic and sentiment-driven factors.

I’ll keep refining the setup, including tuning the remaining hyperparameters and adding more engineered features, and see how much I can improve it.

[Wesley Chuen @wesleychuen]

A lot of people claim to have models that can predict stock prices accurately, but I haven’t seen any that hold up under real market conditions.

I recently came across a 2024 research paper reporting over 90% accuracy using an LSTM model on daily VN-30 index data.

The setup included:
• Indicators: SMA, MACD, RSI
• 60-day sliding window
• Four LSTM layers (30–60 units each)
• ReLU activation, 0.1 dropout
• 1,000 training epochs in TensorFlow and Keras

The paper doesn’t say if that 90%+ figure came from the training set, test set, or true forward predictions.

With only a few months of test data, that number might not hold up once market volatility kicks in.
I’m planning to replicate the setup using daily data to see how it performs in real conditions.
Will compare training and test accuracy against actual market prices.

Should be an interesting little side project.

Source: Nguyen et al. (2024), Humanities and Social Sciences Communications.