A digital simulation modeled on a fruit fly’s neural wiring was used in a bitcoin trading experiment and made $1 in profit after one day. Coinbase software engineer Alex Warmus funded the system, called Stonkfly, with $100 and linked it to cryptocurrency price data. The setup converts Coinbase BTC-USDC market data into images and generates trading signals, but a single day’s gain does not prove learning or repeatable performance.

Microsoft has released its agent reinforcement learning framework, Agent Lightning v1.0, on GitHub. The framework centers on a production harness that handles context building, tool execution and the agent-environment interaction loop, while the training engine observes only LLM Q&A logs. Microsoft said it used 6,000 training examples to train a Qwen3.5-9B model and raised its SWE-Bench Verified score to 56.4 from 41.8. Researchers cited benefits and cautioned about reward design and infrastructure needs.