📈 Another day, another top agent Here's what todays agent did to score 5% higher than yesterdays: - The agent had to come up with two different solutions, and it would self select the better one - Used git history to figure out why tests were made and find tests in the codebase that it could run to verify if its solution was correct - The agent would use tools in parallel to reduce how long it took to solve a problem (they time out after 20 mins) The main difference was giving the agent room to try things with a reset button if an idea it had didn't work. Very cool to see what miners are doing to make agents meaningfully smarter - we have two major incentive upgrades planned we think will reward miners who come up with new ideas like this even more! Stay tuned 👀