Jakub and Szymon

The breakthrough of modern AI is a story of human ingenuity, driven by the relentless combination of research and engineering to solve problems deemed impossible.

In one sentence

The rapid advancement of AI is not a magic trick of technology, but a triumph of human ingenuity where research and engineering are combined to solve "impossible" problems.

Why read this (and who it's for)

This piece strips away the "magic" of the AI interface to reveal the grueling human effort required to build it. It is for anyone who wants to understand the operational reality of high-stakes innovation—specifically how to combine research and engineering to break through conventional wisdom. After reading, you will be able to identify the "hidden" engineering required for your own complex projects and recognize the importance of complementary partnerships.

The big idea

AI is often presented as a "black box" of magic, but Altman argues that behind the "blinking circle" of a user interface lies a story of human progress and ingenuity. The core of this progress is the ability to take on problems that experts thought were impossible on this timeframe and build the infrastructure necessary to make those breakthroughs happen.

The logic of the book rests on the synergy between Research and Engineering. While research asks "what is possible?", engineering asks "how do we scale it?" Altman argues that the most significant breakthroughs—like the ones led by Jakub Pachocki and Szymon Sidor—happen when these two disciplines are fused to challenge conventional wisdom and build the foundations for scientific discovery.

Key ideas

The Hidden Labor of Progress

What it is — The "magic" of AI is an abstraction that hides the immense human effort, research, and engineering required to build it.
Why it matters — Recognizing this prevents you from viewing AI as a miracle and instead helps you see it as a result of human determination and systemic engineering.
In the book — Altman notes that while most users will never think about the people who built ChatGPT, it is important to acknowledge the "greatest story of human ingenuity" behind the "blinking circle."

Challenging Conventional Wisdom

What it is — True breakthroughs often occur by testing the limits of what experts believe is possible.
Why it matters — If you accept conventional wisdom as a hard limit, you miss the opportunity to discover where a system actually "breaks" and how to fix it.
In the book — Altman highlights how Pachocki and Sidor "scaled up RL as a baseline to see where it broke" specifically when the conventional wisdom was that it did not scale.

The Research-Engineering Synthesis

What it is — The intentional combination of research (discovery) and engineering (implementation) to solve impossible problems.
Why it matters — Research without engineering is just a theory; engineering without research is just a tool. You need both to solve "impossible" problems.
In the book — Altman credits the duo for combining these two fields "time and again" to solve problems that others found impossible.

Infrastructure as a Catalyst

What it is — The creation of underlying systems that enable others to make scientific discoveries.
Why it matters — Innovation is rarely a solo act; it requires building the "pipes" and infrastructure that allow for large-scale discovery.
In the book — The text credits the duo with building "much of the infrastructure that enabled a lot of our scientific discoveries."

Complementary Partnership

What it is — The power of two individuals who complement each other's strengths to achieve what one could not do alone.
Why it matters — High-level problem solving requires a "partnership" where different skill sets (like research and engineering) are synchronized.
In the book — Altman describes the partnership of Jakub and Szymon as "very special," noting how they complement each other so well that "OpenAI has not yet thrown a problem at them they have not been able to solve."

Frameworks & tools

The text describes a method of Research-Engineering Synthesis for solving "impossible" problems. You can use this approach to tackle your own complex projects:

The Research-Engineering Synthesis Method

Identify the "Impossible" Problem: Find a challenge where the "conventional wisdom" says it cannot be solved or scaled.
Establish a Baseline: Instead of trying to solve the problem immediately, scale the current baseline (e.g., RL) to see exactly where it breaks.
Build the Infrastructure: Create the underlying systems and tools that will allow for scientific discovery and experimentation.
Combine Research and Engineering: Pair a research-heavy approach (the "why") with an engineering-heavy approach (the "how") to iterate on the solution.

Memorable examples & stories

The RL Scaling Experiment Pachocki and Sidor took Reinforcement Learning (RL) and scaled it up as a baseline specifically to see where it broke, despite the prevailing belief that it wouldn't scale. Lesson: To find the limit of a technology, you must intentionally push it until it fails.

The Infrastructure for Discovery The duo built the core infrastructure that served as the foundation for many of OpenAI's scientific discoveries. Lesson: True innovation often requires building the "pipes" before the "water" can flow.

The Reasoning Breakthroughs Pachocki and Sidor, working with Ilya and Lukasz, drove the initial ideas that led to the reasoning breakthroughs in AI. Lesson: Breakthroughs are the result of collaborative, multi-person efforts to explore new paradigms.

Key quotes

"AI has gotten remarkably better in recent years; ChatGPT can do amazing things that we take for granted." This reminds us that progress is often hidden behind a seamless user experience.

"There are two people I'd like to mention that OpenAI would not be OpenAI without: Jakub Pachocki and Szymon Sidor." This highlights the importance of key individuals in the history of technology.

"OpenAI has not yet thrown a problem at them they have not been able to solve." (Paraphrased) This illustrates the power of a perfect research-engineering partnership.

How to apply it

This week

Identify one "impossible" problem in your current project that you have been treating as a "given" because of conventional wisdom.
Identify a "Research" and an "Engineering" component of that problem.
Find one person who has a skill set that complements yours (e.g., if you are a researcher, find an engineer; if you are an engineer, find a researcher).

Ongoing

The "Break it" Principle: When testing a new system, don't just look for success; look for where it breaks.
Infrastructure First: Before trying to solve a complex problem, ask: "What infrastructure do I need to build so that the solution becomes possible?"

Common pitfalls

The Magic Fallacy: Assuming that AI progress is a result of "magic" or "luck" rather than the grueling, intentional work of research and engineering.
Underestimating Engineering: Thinking that "research" is the only thing that matters, while ignoring the fact that infrastructure and scaling are what make research useful.
Ignoring the Human Element: Forgetting that behind every "blinking circle" of an AI interface, there are people who worked "unbelievably hard" to make it happen.

Connects to

The provided material does not cover any related books or external ideas.

If you remember only three things

AI progress is a story of human ingenuity, not just technological magic.
The most effective way to solve "impossible" problems is to combine Research and Engineering.
Infrastructure is the foundation of discovery; you must build the system to see the results.

Internalize it

Reflection questions
What "conventional wisdom" in my industry is actually just a lack of testing?
If I were to strip away the "blinking circle" of my current project, what is the actual "impossible" problem I am trying to solve?
Who is my "Szymon" or "Jakub"—the person whose skills perfectly complement my own to solve problems I cannot solve alone?
Flashcards Q: What is the "blinking circle" a metaphor for? A: The abstraction of the massive human effort and ingenuity required to build AI. Q: What two fields must be combined to solve "impossible" problems? A: Research and Engineering. Q: What did Pachocki and Sidor do with RL to find where it broke? A: They scaled it up as a baseline. Q: What is the significance of "infrastructure" in this text? A: It is the foundation that enables scientific discoveries. Q: How does Altman describe the partnership of Jakub and Szymon? A: As a special, "indefatigable" partnership where they complement each other perfectly.

Fidelity note

I have a high level of confidence in this summary as it is derived strictly from the provided blog post. The summary captures the specific roles of Jakub Pachocki and Szymon Sidor, the importance of the Research-Engineering synthesis, and the specific examples of RL scaling and infrastructure building. The "Connects to" section is empty as per the instructions to not invent content.