Ryan Dietz

I built an AI tool for doing Amazon Product Research, instead of having to click through hundreds of Keepa filters, you can simply chat with AI. Still really early but looking for sellers that want to collaborate. We're also building Ad automation, keyword harvesting and bid…

What really runs LLM models.. IYKYK @weaviate_io

Update on the LLM trading bot side quest. GPT-o3 is doing decent, 5 wins / 4 losses for +5.2% in 24hrs. Grok-4 is terrible. Even after the adjustments its still not able to understand market conditions. 1 win and 6 losses for -5.5%. Im going to let it run a little longer to…

Update on my little LLM trading bot experiment. I reset the bots Monday with some tweaks to the indicator data. I added candle OHLCV data for the 5m, 15m, 1h and 4h to try and catch wick outs and give HTF trend data, but it did not work. Grok-4 has gone haywire revenge trading…

I finally made my own blog site! No paywalls. No distractions. Just pure dev content. Check it out 👉 devlogs.xyz And feel free to subscribe to the newsletter if you want updates!

Can anyone guess the most used first prompt.. 💀

I don't think the timing of GPT-4.5 being removed, and the potential release of GPT-5 is a coincidence. People love 4.5 for its creative writing, and the way it understands human logic, so I think openAI is consolidating and we'll get that back here soon with GPT-5.

I finally get to build out a true SaaS with the sticky gamified UX I've always wanted.. Usually I'm stuck doing more enterprise type solutions, but this one is going to be fun. I like trying to reverse engineer user psychology to get those quick dopamine hits.

Could it be possible they are shifting focus from the open source weights, to focus on the GPT-5 release. With the Grok-4 drop it kind of puts the pressure on imo.

My old ass refused to swap out CSS for @tailwindcss for the longest time.. But I've officially made the switch, and kick myself for not doing it sooner..

Whether you think AI will replace you, or make you better, you're right.

Are there any senior devs that actually get value out of cursor, or LLM driven IDEs when working in prod? I use AI all the time for dev support, but having it directly in vsCode is more annoying than anything. Maybe it's just my dumbass 🤷

PhD level models are cool and all, but the majority of us simpletons don't need all that. Better accuracy with basic tasks is more paramount for the majority at this point.

Zuck is spending hundreds of millions on talent, @elonmusk is spending hundreds of millions on GPUs.. Zuck has zero frontier models, Elon has arguably the best frontier model. Do with this what you will..

So far I'm not impressed with Grok4, just being honest.. I've tried through the API and App, maybe just my dumbass, but it's not living up to the hype yet. I'm not testing it on PhD level questions so maybe it's like driving a Ferrari on a go-kart track..

All bots have been hot swapped to Grok4. Pretty pumped to see the delta between the 2 models.

Switching from dev mode to bizdev mode is one of the hardest parts as a founder, and completely under rated. Those who can do it well, will kill it.