Real Alpha

Unsupervised Learning · 2 min instead of 57 min · Oct 6, 2026

Post-training wins inference because the biggest workloads train first

YP

Yash Patil

CEO, Applied Compute

  • Jacob EffronHost · AI Investor, Redpoint Ventures

    Growth investor in Abridge, Ramp and Physical Intelligence; ex-Flatiron Health.

About this segment

Yash explains why the largest inference workloads are the first worth post-training. Training and serving can be co-optimized, so how a model is trained shapes how it is deployed.

From “Applied Compute CEO on the Limits of RL, the New AI Hyperscaler & Why Post-Training Wins Inference”, Oct 6, 2026. Watch the full episode

“the most at-scale workload, so the largest workloads where you're spending the most money, is actually where you're going to want to go and post-train first, right?”
YPYash Patilat 29:52 · Unsupervised Learning29:52

Takeaways

  1. Companies start on frontier models, then post-train once a workload matures
  2. The biggest workloads follow a power law and gain the most from training
  3. Training for tool-call parallelization changes how you set up inference
Watch 2:01 from 29:20

In this episode

Full episode

  1. 29:20Post-training wins inference because the biggest workloads train first2:01

  2. 24:24RL is a hill-climbing machine, so protect your evals2:03Up next
  3. 12:28Continual learning is blocked by data-efficient sparse-reward training1:27

Transcript

Transcribed by machine (Whisper) from the episode’s audio, so a word may be off here and there.

  1. 29:24

    And so what our view is is that the most post-training actually wins inference. So if you are – basically the trend we've been seeing is like companies will start off with frontier models. They'll do a lot of experimentation. Then their workload will become mature. They'll be scaling that workload. then they'll think about open models and how to sort of cost optimize, basically push the Pareto curve out.

  2. 29:54

    And the interesting thing is that the most at-scale workload, so the largest workloads where you're spending the most money, is actually where you're going to want to go and post-train first, right? Because that's actually where the most gains are to be had, and it's kind of power law between all the other use cases. And the interesting thing you can do is you can both co-optimize your inference and your training.

  3. 30:19

    So basically the way we train models actually influences how we go and set up inference deployments for customers. If it's like a heavy tool-using model and we train it to do a bunch of tool call parallelization and that kind of stuff. Can you say more about that? What does that actually mean in practice? If it was designed that way, how does that influence the inference choices you make? Yeah, yeah.

  4. 30:41

    So it's a great question. And so basically, if we are doing a long horizon tool using agent, the way we set up the inference deployment or the inference engine is different than if we were doing a decode-heavy workload. And you can set up these disaggregated setups where you're using multiple different types of chips for pre-fill and decode. And so having influence on how the models are trained actually can influence how we want to go and set up the inference deployments for very large-scale workloads.

  5. 31:14

    It's interesting. I'm really curious to dig into this because I feel like there's been this question for a while