Person/Sean Grove/YouTube/25/07/The New Code

• The New Code — Sean Grove, OpenAI - YouTube

  • Summaries

    • My Notes
      • In particular, I'm interested in reading the Deliberative Alignment paper he mentioned at 13:48, where an AI/Model/Spec ( #Example: OpenAI/GitHub/model_spec (what I have been calling an AI/Behavior/Catalog)) would be used by a more powerful LLM-as-Judge to score the input output pair, which then was fed back into fine-tuning.

    • official, via youtube

      • In an era where AI transforms software development, the most valuable skill isn't writing code - it's communicating intent with precision. This talk reveals how specifications, not prompts or code, are becoming the fundamental unit of programming, and why spec-writing is the new superpower.
      • Drawing from production experience, we demonstrate how rigorous, versioned specifications serve as the source of truth that compiles to documentation, evaluations, model behaviors, and maybe even code.
      • Just as the US Constitution acts as a versioned spec with judicial review as its grader, AI systems need executable specifications that align both human teams and machine intelligence. We'll look at OpenAI's Model Spec as a real-world example.
      • Finally, we'll end on some open questions about what the future of developer tooling looks like in a world where communication once again becomes the most important artifact in engineering.
      • About Person/Sean Grove
        • Sean Grove works on alignment reasoning at OpenAI, helping translate high‑level intent into enforceable specs and evaluations. Before OpenAI he founded OneGraph, a GraphQL developer‑tools startup later acquired by Netlify. He has delivered dozens of technical talks worldwide on developer tooling, APIs, AI UX and design, and now alignment.
        • Recorded at the AI Engineer World's Fair in San Francisco.

    • via AI/Engineer/Twitx TwitX here

      • 🆕 Structured communication is the bottleneck. Whoever communicates best becomes the most valuable programmer.
      • As LLMs increasingly do more and more code generation, the need for clear communication with Agents and with Humans goes up.
      • The clearest artifact of communication is a specification, whether it is a PRD between PMs and Engineers, the US constitution between government and its people, OpenAI's model spec for LLMs.
      • Sean applies the Model Spec to the GPT 4o glazegate issue - which actually preserved trust in the community that sycophancy was a bug! And since everything is a spec... we should think hard about how good specs are written, as we write them for our own AIs.

  • #Video