Morph Founder: Autoresearch Made Our Models 3x Faster, but 80% of Attempts Fail
AI Engineer · youtube · 2026-09-26
Tejas Bhakta, founder of Morph and formerly of Tesla's inference optimization team, explains how his team combined agent-written custom GPU kernels with bare-metal tweaks to run models 3x faster on cheaper GPUs.
Key points:
- GPU kernels are a near-perfect autoresearch target since correctness and speed are easy to verify.
- Humans bring the big ideas; agents handle parameter tuning. Agents miss high-level insights, like spotting that a DeepSeek attention step loads far more context than needed.
- Agents need explicit hardware and model context to work well.
- Watch for reward hacking: e.g., disabling CUDA graphs to speed up one kernel while slowing the whole model.
- Kernel gains compound, and bare-metal tweaks add 25% over a cloud VM.
- His honest warning: roughly 80% of autoresearch attempts are bad.
More from coding & agent
- Building a Semantic and Knowledge Layer on BigQuery and Exposing It to Agents — batmanparam · 2026-09-27
- OpenAI selects 35 builders for first Codex Physical Builds hardware cohort — OpenAIDevs · 2026-09-27
- Multi-Agent Bot Pipelines Keep Stalling: Direct-to-Harness Still Beats Bot Teams — Old_Grapefruit8774 · 2026-09-27
- Claude Opus 5.5 Chains Blender, Image-Gen and three.js Into Browser-Running 3D — burny_tech · 2026-09-27
- Claude kills plan mode the same day Google announces launching it — dotey · 2026-09-27
- Gary Marcus amplifies warning: large teams using AI agents likely have unknown security incidents — GaryMarcus · 2026-09-27