Ex-model-training PhD: frontier LLMs are better at research ideas than I am, deployment was the right pivot
niceuser45 · reddit · 2026-10-19
A PhD-turned-FAANG engineer explains why moving from model training to deployment optimization in 2024 was the right call. Frontier LLMs now explain diffusion models so well—with visualizations—that he believes he'd stand no chance competing on novel training ideas outside a GPT-7-level lab.
Deployment is different: edge architectures are confidential and knowledge is empirical and tribal, so LLMs can't help out of the box. He still feels FOMO about training but is making peace with not going back.
More from AGI Musings
- Compute wins the AI race: Meta caught up and xAI scaled fast by building compute — haider1 · 2026-10-19
- US retailers adopt AI-assisted pricing tools as consumer apprehension grows — ArtificialOther · 2026-10-19
- David Patterson predicts superintelligence will cure all diseases by 2030 — davidpattersonx · 2026-10-19
- GDC 2026 survey: 64% of game devs skip genAI as negative views hit 52% — IanArawjo · 2026-10-19
- Can Neural Nets Produce Qualia? A Philosophical Exchange on AI Consciousness — brandon_xyzw · 2026-10-19
- Only 13% of 3,362 automation case studies use AI; agents appear in just 18 — Sonia_here89 · 2026-10-19