AI

OpenAI's Yann Dubois: Why AI Progress Suddenly Feels Real

Data Driven NYC
Data Driven NYC•May 21, 2026

Why It Matters

GPT‑5.5’s reliability and efficiency breakthrough makes AI a dependable productivity partner for businesses, accelerating deployment in coding, security, and knowledge work while reshaping competitive dynamics.

Key Takeaways

  • •Reliability threshold crossed in December, making AI tools truly usable.
  • •New models accelerate development by improving coding and tooling efficiency.
  • •Reinforcement learning shifted from verifiable tasks to real‑world applications.
  • •Horizontal improvements ensure consistent performance across diverse verticals.
  • •Model efficiency doubled, cutting latency and token usage significantly.

Summary

In a candid conversation on the Mad Podcast, OpenAI’s post‑training frontiers lead Yann Dubois explains why the release of GPT‑5.5 feels like a sudden step‑function in AI progress. He argues that a reliability milestone was reached around December 2023, after which the models became trustworthy enough for real‑world workloads, turning continuous capability gains into a perceptible leap.

Dubois outlines three drivers behind the acceleration: the reliability breakthrough, the self‑reinforcing loop where better models speed up both research and tooling, and the migration of reinforcement‑learning techniques from math‑oriented, verifiable rewards to messy, production‑grade coding and cybersecurity tasks. He also describes OpenAI’s organizational split between vertical specialist teams and a horizontal frontiers team that smooths performance across use‑cases, ensuring the model behaves consistently.

Memorable remarks include, “We just crossed that threshold, now we can trust these models to do a lot of the work we’re doing,” and “once you start having models that are really good you accelerate yourself.” Dubois notes the internal excitement waves surrounding GPT‑5.5 and highlights a two‑fold pride: a 2× efficiency gain and a company‑wide alignment on a single north‑star model.

The implications are clear for enterprises: with higher reliability and doubled efficiency, AI can now be deployed for critical coding, security, and knowledge‑work tasks at scale, reducing latency and token costs. However, Dubois cautions that the “last mile” of domain‑specific reliability remains an open challenge, urging continued investment in both vertical expertise and horizontal robustness.

Original Description

AI suddenly feels like it has crossed a threshold, and Yann Dubois, co-lead of the Post-training Frontiers team at OpenAI, joins Matt Turck to explain why. Yann’s team has led the post-training behind the company's reasoning models, including the recent GPT-5.5 release. In this conversation, we go inside the shift from raw model capability to useful, reliable systems: what changed with GPT-5.5, why reinforcement learning is moving beyond math and coding competitions into messy real-world work, how reasoning models like GPT-5.5 actually work, the difference between GPT-5.5 Thinking and GPT-5.5 Pro, why post-training has become one of the most important frontiers in AI, and why evals, model-as-judge, hallucinations, agentic workflows, GDPval, and continual learning are now central to the next phase of frontier models. Yann also shares why continual learning remains one of AI's biggest unsolved problems three years after ChatGPT, and where startups still have massive room to build as frontier models race ahead.
Yann Dubois
OpenAI
Matt Turck (Managing Director)
FirstMark
Listen on:
00:00 - Cold open
00:34 - Intro
01:30 - Why recent AI progress feels like a step function
04:13 - Model reliability & the rollercoaster of shipping 5.5
07:33 - How OpenAI structures vertical and horizontal teams
09:49 - Improving model efficiency and test-time compute
12:32 - Yann Dubois' journey from Switzerland to OpenAI
15:37 - Reasoning in 2026: Real-world utility vs verifiable rewards
18:34 - GPT-5.5 Thinking vs Pro: Scaling test-time compute
20:09 - How reasoning models become more efficient
23:23 - Pre-training scaling and overcoming the data wall
27:03 - Multimodal data, synthetic data, and embodied AI
31:05 - Demystifying mid-training and post-training
37:21 - Does RL create new capabilities in AI?
38:53 - The challenges and frontier of scaling RL
43:09 - Is building AI models a craft or a strict science?
48:21 - How AI models generalize across different domains
54:18 - How reinforcement learning cures AI hallucinations
56:04 - Negative generalization and conflicting instructions
58:05 - Can RL scale to law, medicine, and the broader economy?
1:00:19 - The evaluation bottleneck and Model as a Judge
1:04:21 - Continuous AI progress & continual learning
1:08:49 - Will foundation models eat the agent harness?
1:11:23 - Why startups should focus on the last mile of AI

Comments

Want to join the conversation?

Loading comments...