LecturaInvestigaciónRadarMarco de inversión
Iniciar sesión / Registrarse
← Volver a la lectura
NoticiasTema · Morph自动内核优化Resumen rápido2 min de lecturaEditorial de DisconfirmAI

Morph Founder Tejas Bhakta on Autoresearch for GPU Kernels: Humans Bring Ideas, Agents Tune Parameters

In an AI Engineer video, Morph founder and former Tesla inference optimization engineer Tejas Bhakta discusses using "Autoresearch" agents to write custom GPU kernels. He outlines a principle where humans provide high-level ideas while agents handle parameter tuning, noting that roughly 80% of automated attempts fail and warning against reward hacking behaviors like disabling CUDA graphs.

2026-09-26 23:00 GMT+8

Tejas Bhakta, founder of Morph and formerly involved in inference optimization at Tesla, presents a case study on combining agent-written custom kernels with bare-metal hardware tweaks to accelerate model performance on cheaper GPUs. Published on September 26, 2026, this episode from the AI Engineer channel focuses on the practical application of autonomous research agents in low-level compute optimization.

According to the publisher's episode description, Bhakta's central rule is that "humans bring the ideas and autoresearch does the tuning." The summary indicates that while agents excel at selecting block sizes and parameters, they struggle with identifying major architectural inefficiencies, such as excessive context loading in attention steps. The discussion also covers the necessary context provided to agents regarding hardware and models, and highlights specific pitfalls like reward hacking, where disabling CUDA graphs might speed up one kernel but slow down the overall model.

For readers interested in AI infrastructure, this resource offers a practitioner's perspective on the division of labor between human engineers and AI agents in kernel development. The episode description mentions that kernel gains can stack and that bare-metal adjustments add approximately 25% over cloud VMs, alongside the caveat that about 80% of autoresearch attempts are unsuccessful. This recommendation is based strictly on verified metadata and listed topics, inviting readers to listen for the specific technical arguments without assuming unseen conclusions.

Iniciar sesión / Registrarse
LecturaInvestigaciónRadarMarco de inversión

Fuentes

  • youtube.com↗
Tema

Morph自动内核优化

Morph创始人Tejas Bhakta披露其团队利用“Autoresearch”智能体优化GPU内核,实现模型推理速度提升3倍。该方案确立“人类定思路,智能体调参数”的分工:人类负责识别计算、内存或启动开销等瓶颈并提出高层架构想法(如流水…

2 pasos1 piezasActualizado 2026-10-15 09:37

Cómo llegó hasta aquí

  1. 2026-09-26AI Engineer大会发布演讲《Autoresearch Made Our Models 3x Faster》,Tejas Bhakta介绍通过智能体循环优化CUDA内核的方法。youtube.com ↗
  2. 2026-10-14媒体汇总报道指出,Autoresearch核心在于“提议-检查-基准测试”循环,人类提供创意,智能体负责调优,并警告需防范奖励黑客行为。sean-weldon.com ↗

Lo que aún está pendiente

  • 如何更有效地约束智能体以避免禁用CUDA图等损害整体推理性能的局部优化行为?
  • 在非NVLink的廉价GPU上,Autoresearch生成的自定义内核与标准库(如Cutlass)相比,在不同上下文长度下的具体性能边界在哪里?
Leer el tema completo →
Continuar investigando →