Loading activity…
@18jeffreyma
CS PhD @Harvard + researcher @Meta Created http://swefficiency.com, http://programbench.com, http://swe-sweep.com
AI
A post says Carbon scored 88.2% on SEC-bench Pro and is rolling out first to cybersecurity specialists.
AI
A Google team member claims the agent ranked #1 on NVIDIA's SOL-ExecBench and reached 2× human experts on SWE-fficiency.
AI
A user comparing versions 1.1 and 1.3, both at xhigh, says 1.3 used 39% fewer tokens across the benchmark and brought the number of programs at least 95% resolved from eight to 33.
