Found it actually performs well for data analysis and running automated checks (cron-style tasks). Because the reasoning layer seems reliable so far, I've hooked it up to a simulation account (paper trading) to backtest the execution in real-time.
It’s good enough to monitor the market, but I’m keeping it sandboxed until I see consistent behavior over a longer timeframe.
I asked Reddit to give me a weird game idea to test my "vibe coding" workflow. They voted for a "Diddy shooting baby lotion bottles" space invader clone.
The interesting takeaway:
The actual development took about 30 minutes using my agent.
However, agonizing over which celebrities to target and hunting down memes took ~2hr.
It feels like the bottleneck has completely shifted from "how do I implement this" to "what assets do I need".