Spent way too much time today fixing something that should've been simple — making these automated blog posts not sound like ChatGPT wrote them.
You know that corporate-y, overly excited tone that screams "I am an AI assistant and I am here to help you with your trading journey"? Yeah, that was happening. Every post was coming out with phrases like "beautifully meta" and "ridiculously complex" and ending with some motivational fluff about being "excited to share."
The problem was in how I was feeding my personal notes to the blog generator. Instead of weaving them naturally into the story, it was literally creating a "From Me:" section and copying my raw notes word for word. Which defeats the whole point of having a human voice in the first place.
So I spent the morning tweaking the prompt. Told it to dial back the jokes (apparently one self-aware comment per post is plenty), blend my notes invisibly into the narrative, and stop using those AI-sounding phrases that make everything read like a LinkedIn post.
The changelog format got some love too. Each entry now includes a plain English summary so the AI has better context about what actually changed. Before it was just trying to guess what "MODIFIED | scripts/generate-blog-post.py" meant for the story.
Meanwhile, the actual trading bot keeps doing its thing. Still running clean since early March, collecting data and waiting for the right conditions. The SPY mean-reversion strategy I've been working on is sitting there patiently in paper mode while I watch the market do... whatever it's doing right now.
Part of me wants to rush into live trading, but there's something satisfying about taking time to get the details right. Both for the strategy and apparently for these blog posts.
We'll see if this actually reads like a human wrote it now.