Jon Saad-Falcon
Showing 1–2 of 2 transcripts.
- Y Combinator1h 0m
Self-Improving Harnesses, Local Personal AI And YC's Agent For Work | YC Paper Club
Seth Karten, Jon Saad-Falcon, Josh France, Regan Bell
YC Harness Club researchers demonstrate that "self-improving harnesses" capable of modifying their own code and memory have driven Arc AGI success rates from 30% to 100%, surpassing standard prompt engineering by leveraging recursive learning architectures. Presentations by Seth and John Sad Falcone detail specific implementations like Prime Agent's multi-tier context management and OpenJarvis's local on-device stack, which collectively achieve frontier-level performance while reducing costs by 800x and enabling autonomous long-horizon research. Furthermore, the QM project establishes a centralized cloud-hosted framework with dynamic sandboxing and fine-grained permissions, addressing critical challenges in system-wide coordination and data security for scalable multi-agent deployment.
- Y Combinator1h 16m
Kernels and Chips: Cursor, NVIDIA, and Meta Researchers on GPU Performance | YC Paper Club
Stuart Sul, John, Francois Chaubard, Jon Saad-Falcon, Mark Saroufim, Misha Smelyanskiy, Brennan Shacklett
The event synthesizes critical industry shifts toward specialized ASICs and disaggregated architectures that separate training and inference workloads to maximize intelligence per watt. Technical deep dives highlight breakthroughs in multi-GPU kernel optimization and local inference, demonstrating that consumer accelerators can now deliver nearly 90% of frontier model utility while reducing energy costs by 70%. These innovations collectively address systemic bottlenecks in network communication and latency, establishing a roadmap for distributed, cost-efficient AI ecosystems that minimize reliance on centralized cloud resources.