
Jakub Pachocki is OpenAI's chief scientist; MIT Technology Review reports that he "played key roles in the development of both GPT-4" and reasoning models, and he described his job as "setting the research road map for the company." — MIT Technology Review: OpenAI is throwing everything into building a fully automated researcher (Will Douglas Heaven, March 2026).
Part 1: Competitive Programming & Algorithmic Rigor
- On Competitive Programming: Pachocki said "I became very passionate about the idea that there is some problem and it's not clear at first that it is solvable, but you can write a program that makes the computer do that in a reasonable time." — Before AGI podcast with Jakub Pachocki and Szymon Sidor (July 2025), unofficial bilingual transcript page.
- On Transitioning to Deep Learning: Pachocki said "the thing that forced me to re-evaluate was AlphaGo." — Before AGI podcast with Jakub Pachocki and Szymon Sidor (July 2025), unofficial bilingual transcript page.
Part 2: OpenAI Five & Reinforcement Learning
- On Dota 2: Pachocki said "The particular project we were trying to solve was the game of Dota 2." — Before AGI podcast with Jakub Pachocki and Szymon Sidor (July 2025), unofficial bilingual transcript page.
- On Scaling Instead of a Barrier: Pachocki said "it didn't fail, it just kind of worked. We ended up just scaling things up." — Before AGI podcast with Jakub Pachocki and Szymon Sidor (July 2025), unofficial bilingual transcript page.
- On expecting a barrier: Pachocki said "What we thought would be hitting a barrier and then going back to the drawing board ended up being a two-year process of scaling things up." — Before AGI podcast with Jakub Pachocki and Szymon Sidor (July 2025), unofficial bilingual transcript page
- On the Legacy of Five: Pachocki wrote that the scaling projects of 2017 "included scaling self-play, robotics, and, in retrospect most notably, scaling recurrent networks to model language, which was a precursor to the GPT line of research." — OpenAI: An Alien Mind, by Jakub Pachocki (September 6, 2026).
Part 3: The Development of GPT-4
- On Scaling: Pachocki wrote "At a high level, progress in machine intelligence is driven by increasing computational power." — OpenAI: An Alien Mind, by Jakub Pachocki (September 6, 2026).
- On committing to compute: Pachocki wrote "We at OpenAI deeply internalized this around 2017, after seeing consistent returns to scaling across multiple research projects." — OpenAI: An Alien Mind, by Jakub Pachocki (September 6, 2026)
- On algorithmic progress: Pachocki wrote "I see them largely as discoveries along the path of scaling; the science of deep learning is still nascent, and meaningful algorithmic progress tends to correlate with access to compute." — OpenAI: An Alien Mind, by Jakub Pachocki (September 6, 2026)
- On a research roadmap: Pachocki said "the primary thing I'm responsible for is setting the research road map for the company." — The OpenAI Podcast Ep. 5: AGI progress, surprising breakthroughs, and the road ahead (August 2025, video)
- On generality: Pachocki said "we seek to create intelligence that is very general." — The OpenAI Podcast Ep. 5: AGI progress, surprising breakthroughs, and the road ahead (August 2025, video)
Part 4: The Shift to Reasoning Models
- On Reasoning Models: Pachocki wrote that in mid-2023 "we saw the first results that gave us confidence that we will be able to scale the training of reasoning models, unlocking the capability of pretrained models to form their own chains of thought." — OpenAI: An Alien Mind, by Jakub Pachocki (September 6, 2026).
- On Continuous Growth: Pachocki said "I definitely see this as a signal that like you know something here is on track." — Unsupervised Learning, Ep 84: OpenAI's Chief Scientist on Continual Learning Hype, RL Beyond Code, & Future Alignment Directions (April 2026, video).
Part 5: Automating Scientific Research & The AI Intern
- On the AI Research Intern: Pachocki said "What we're really looking at for an automated research intern is a system that you can delegate tasks to that would take a person a few days." — MIT Technology Review: OpenAI is throwing everything into building a fully automated researcher (Will Douglas Heaven, March 2026).
- On a research lab in a data center: Pachocki said "I think we will get to a point where you kind of have a whole research lab in a data center." — MIT Technology Review: OpenAI is throwing everything into building a fully automated researcher (Will Douglas Heaven, March 2026)
- On proving the technology: Pachocki said "That lets you prove that the technology works before you connect it to the real world." — MIT Technology Review: OpenAI is throwing everything into building a fully automated researcher (Will Douglas Heaven, March 2026)
- On not prioritizing a mathematician: Pachocki said "If we really wanted to, we could build an amazing automated mathematician." — MIT Technology Review: OpenAI is throwing everything into building a fully automated researcher (Will Douglas Heaven, March 2026)
- On real-world research: Pachocki said "We are much more focused now on research that's relevant in the real world." — MIT Technology Review: OpenAI is throwing everything into building a fully automated researcher (Will Douglas Heaven, March 2026)
- On Automating Discovery: Pachocki said "I think it is just hard to internalize like this is possible to automate most of this process." — The OpenAI Podcast Ep. 5: AGI progress, surprising breakthroughs, and the road ahead (August 2025, video).
- On Productivity Tooling: Pachocki said "I can have it run experiments in a weekend that previously would have taken me like a week to code." — MIT Technology Review: OpenAI is throwing everything into building a fully automated researcher (Will Douglas Heaven, March 2026).
- On being pedantic about code: Pachocki said "I'm very pedantic about my code" and "I like to type it all manually in vim if I can help it." — MIT Technology Review: OpenAI is throwing everything into building a fully automated researcher (Will Douglas Heaven, March 2026)
- On not letting it take the reins: Pachocki said "I don't think it is at the level where I would just let it take the reins and design the whole thing." — MIT Technology Review: OpenAI is throwing everything into building a fully automated researcher (Will Douglas Heaven, March 2026)
- On doubters: Pachocki said "I can believe some people don't find it very useful yet." — MIT Technology Review: OpenAI is throwing everything into building a fully automated researcher (Will Douglas Heaven, March 2026)
- On Shifting Human Focus: Pachocki said "Nobody really edits code all the time anymore. Instead, you manage a group of Codex agents." — MIT Technology Review: OpenAI is throwing everything into building a fully automated researcher (Will Douglas Heaven, March 2026).
- On the 2028 Horizon: Pachocki said "Even by 2028, I don't expect that we'll get systems as smart as people in all ways." — MIT Technology Review: OpenAI is throwing everything into building a fully automated researcher (Will Douglas Heaven, March 2026).
- On Near-Term Limits: Pachocki said he did not expect systems you could simply tell to go solve alignment, "not this year." — Unsupervised Learning, Ep 84: OpenAI's Chief Scientist on Continual Learning Hype, RL Beyond Code, & Future Alignment Directions (April 2026, video).
Part 6: Alignment and Chain-of-Thought Monitoring
- On Alignment as a Core Problem: Pachocki wrote "The core problem in AI research is that of alignment - getting the AI to “try to do the right thing” by human standards." — OpenAI: An Alien Mind, by Jakub Pachocki (September 6, 2026).
- On Chain-of-Thought Monitoring: Pachocki wrote that chain-of-thought monitoring "is based on an appealingly scalable idea: a lot of the model's capability comes from a verbalized reasoning process (chain-of-thought)." — OpenAI: An Alien Mind, by Jakub Pachocki (September 6, 2026).
- On not supervising the reasoning: Pachocki wrote "If we scale optimization on the outcomes of that process, but do not supervise the process itself, that chain-of-thought has no direct incentive in training to hide any misaligned ideas or objectives." — OpenAI: An Alien Mind, by Jakub Pachocki (September 6, 2026)
- On diminishing monitorability: Pachocki wrote "our evaluations indicate our ability to rely on CoT monitoring is progressively diminishing." — OpenAI: An Alien Mind, by Jakub Pachocki (September 6, 2026)
- On confidence in monitoring: Pachocki wrote "I expect general AI progress to increasingly be bottlenecked by confidence in monitoring." — OpenAI: An Alien Mind, by Jakub Pachocki (September 6, 2026)
- On depending on monitoring: Pachocki said "Once we get to systems working mostly autonomously for a long time in a big data center, I think this will be something that we're really going to depend on." — MIT Technology Review: OpenAI is throwing everything into building a fully automated researcher (Will Douglas Heaven, March 2026)
- On the Challenge of Generalization: Pachocki wrote "The fundamental challenge of AI alignment is generalization." — OpenAI: An Alien Mind, by Jakub Pachocki (September 6, 2026).
- On Automated Alignment Research: Pachocki wrote that a north star is "building an automated AI researcher, iterating with it on the alignment problem and finding ways for people to remain part of the self-improvement loop." — OpenAI: An Alien Mind, by Jakub Pachocki (September 6, 2026).
Part 7: AI's Societal Impact and Power Concentration
- On Racing Forward: Pachocki wrote "The idea of racing forward at all costs seems absurd once one internalizes the seriousness of the stakes." — OpenAI: An Alien Mind, by Jakub Pachocki (September 6, 2026).
- On Human Judgment: Pachocki said "Of course, you still want people in charge and setting the goals." — MIT Technology Review: OpenAI is throwing everything into building a fully automated researcher (Will Douglas Heaven, March 2026).
- On Power Concentration: Pachocki said "It's going to be a very weird thing. It's extremely concentrated power that's in some ways unprecedented." — MIT Technology Review: OpenAI is throwing everything into building a fully automated researcher (Will Douglas Heaven, March 2026).
- On a data center doing the work: Pachocki said "Imagine you get to a world where you have a data center that can do all the work that OpenAI or Google can do. Things that in the past required large human organizations would now be done by a couple of people." — MIT Technology Review: OpenAI is throwing everything into building a fully automated researcher (Will Douglas Heaven, March 2026)
- On personal responsibility: Pachocki said "I do feel personal responsibility" and "I don't think this can be resolved by OpenAI alone." — MIT Technology Review: OpenAI is throwing everything into building a fully automated researcher (Will Douglas Heaven, March 2026)
- On voluntary slowdowns: Pachocki wrote "I expect and hope for voluntary slowdowns to become commonplace until shared safety bars are established." — OpenAI: An Alien Mind, by Jakub Pachocki (September 6, 2026)
- On keeping people in the loop: Pachocki wrote that the core challenge of automating AI research is getting there "in a way that keeps people a part of the continued improvement process, and leaves the future in humanity's hands." — OpenAI: An Alien Mind, by Jakub Pachocki (September 6, 2026)
- On Speed vs. Safety: Pachocki wrote "Scaling AI systems has to be constrained by our confidence in safety." — OpenAI: An Alien Mind, by Jakub Pachocki (September 6, 2026).
- On Societal Governance: Pachocki wrote "I believe that international coordination on future AI development needs to become a top priority for governments around the world." — OpenAI: An Alien Mind, by Jakub Pachocki (September 6, 2026).
- On Human Agency: Pachocki wrote "We need to find ways to preserve human agency and enshrine an intrinsic value to being human, in a world where most tasks could be performed by AI." — OpenAI: An Alien Mind, by Jakub Pachocki (September 6, 2026).
- On Systemic Risk: Pachocki said "Until you can really trust the systems, you definitely want to have restrictions in place." — MIT Technology Review: OpenAI is throwing everything into building a fully automated researcher (Will Douglas Heaven, March 2026).
Part 8: Philosophy on Deep Learning as a Natural Science
- On Deep Learning as a Science: Pachocki wrote "The study of deep learning-based AI is largely an experimental science." — OpenAI: An Alien Mind, by Jakub Pachocki (September 6, 2026).
- On Empirical Observation: Pachocki wrote "our large-scale training runs are experiments, and we are sometimes surprised by their results." — OpenAI: An Alien Mind, by Jakub Pachocki (September 6, 2026).
- On interpreting results: Pachocki wrote "as the systems become more capable, the results become harder to interpret." — OpenAI: An Alien Mind, by Jakub Pachocki (September 6, 2026)
- On grown systems: Pachocki wrote "AI is grown more than designed." — OpenAI: An Alien Mind, by Jakub Pachocki (September 6, 2026)
- On surpassing enough: Pachocki wrote "the AI does not need to match or exceed all human capabilities; it just needs to surpass enough of them." — OpenAI: An Alien Mind, by Jakub Pachocki (September 6, 2026)
- On being transformative: Pachocki said "you don't need to be as smart as people in all their ways in order to be very transformative." — MIT Technology Review: OpenAI is throwing everything into building a fully automated researcher (Will Douglas Heaven, March 2026)
- On the Limits of Pure Theory: Pachocki wrote that, as in neuroscience, the system's "overall action evades a description we can fully understand." — OpenAI: An Alien Mind, by Jakub Pachocki (September 6, 2026).