@FinanceYF5: 2/ His name is Lenny Bogdonoff. He joined OpenAI when it only had 250 people, while GPT-4 was still being trained and ChatGPT hadn't launched yet. His first task: rebuilding the Jupyter code execution environment, which later became the prototype for the 'AI computer' concept. He didn't realize how important this was, and most people didn't either.

A Decade of AI: One Person’s Story

1/ Expelled from high school, dropped out twice, worked over 30 jobs before age 22.

Later, he was part of ChatGPT’s journey from 0 to 1, witnessing the months when hundreds of millions of users flooded in.

Now he’s thinking about what to do for the next decade.

A self-narrative worth reading.

2/ His name is Lenny Bogdonoff.

He joined OpenAI when it had only 250 people — GPT-4 was still being trained, and ChatGPT hadn’t launched yet.

His first task: rebuild the Jupyter code execution environment, a prototype for what later became the “AI computer” concept.

He didn’t realize how important this was. Most people didn’t either.

3/ ChatGPT launched in November 2022.

No one predicted it would reach this scale.

Traffic grew week by week. All GPUs were reallocated. The entire company revolved around a single variable: compute.

Database IDs started overflowing. Every early architecture decision broke down one after another. A team of fewer than 10 people was supporting hundreds of millions of users.

4/ The most important thing he did at OpenAI wasn’t code.

It was realizing: The user scale of ChatGPT far exceeds any outsourced annotation team.

If you could get users involved in the data flywheel, model quality would reach a different level.

Whether a model is good or not is largely an operations and quality management problem — not just an algorithm problem.

5/ In early 2024, he left OpenAI to become a VC.

He spent a year and a half diving deep into new fields every day: AI’s impact on the physical economy, energy supply chains, raw materials…

He says it was the fastest period of non-technical learning in his career.

Then the fund was acquired, the founder left, and he left too.

6/ His core judgment now:

It’s not “how powerful AI is,” but the speed of intelligence — the distance from knowing something to taking action is being compressed by AI.

He’s seen acceleration happen from the inside, and he’s also seen where it gets stuck.

Where it gets stuck is often not a technical problem, but an organizational one.

7/ He wants to spend the next decade on that stuck place.

“Where AI is cheap and fast isn’t where the biggest returns are.”

Few people see this clearly — those who’ve both been inside a lab and sat in an investor’s seat.

What industry do you think has the slowest AI penetration but the highest potential?

That’s all.

If you like this topic:

1. Follow me (@FinanceYF5)
2. Like + repost the first post below