Irony: The DeepSeek Team is "mainly in their mid-20s" While Many Aged Computer Science Professors Are Now Rushing into AI/ML Research

Age is not the deciding factor; mindset and purpose matter.

Feb 07, 2025

The biggest tech news of 2025 so far is the emergence of DeepSeek, a breakthrough AI product developed by a small AI startup in China.

A few days ago, I published an article, “Don’t you think it’s absurd that most university IT researchers are focusing on AI and machine learning?”, reflecting my observations. It was selected by Medium Curator as “Boosted”!

Credit: ‘Silicon Valley” TV Show, SE0203.

Today, I came across a post stating that the DeepSeek team is predominantly young—very young. Curious, I looked it up and found it to be true.

Table of Contents:
· Many DeepSeek Team Members are early 20s
· A Notable Example of a Computer Science Professor
· Chasing hypes is quite common in commercial software development, too
· Software Startup Founders: Leverage the Potential of Young Talents
· Age is NOT the issue; the Mind Age is
· Many software companies now lack engineering discipline and have become overly bureaucratic

Many DeepSeek Team Members are early 20s

“The (DeepSeek) team seems to be composed of young graduates, many of whom have recently completed their degrees at elite Chinese institutions. A lot of them graduated in the past couple of years, whether it’s with a masters or a PhD. Some were still listed as being in graduate school. It is a very young team, mainly in their mid-20s.” - - University World News (2025-02—05)
My note: mid-20s after dedcuting 2 years research, started at early-20s.

This was not an accident but rather intentional.

“Liang Wenfeng (39), the company (DeepSeek')’s founder and CEO, told Chinese media in a 2023 interview that technical roles were taken up mostly by fresh graduates or ‘those with one or two years of working experience’ ".
- University World News (2025-02—05)

Essentially, a founder in his 30s led a team of some early 20s to achieve a groundbreaking AI research breakthrough that amazed the world.

Here’s another fact: much of the research and investment in AI took off after the release of ChatGPT in late 2022. In other words, the starting timeline is roughly the same.

Now, consider this question: “Do university Computer Science professors—typically in their 50s—have a higher success rate in AI research compared to younger individuals?”

The answer, based on DeepSeek's story (and history, as noted below), has to be a resounding NO. To my knowledge, universities haven't produced groundbreaking AI research since the late 1990s.

IBM’s DeepBlue beating the world chess champion in 1997
DeepMind’s AlphaGo (later a part of Google) defeated the Go world champion in 2017.
ChapGPT by OpenAI (late 2022)

Why did such a high percentage of Computer Science professors abandon their previous research areas and jump on the latest bandwagon? The only logical explanation is hype-chasing. With hype comes an immediate benefit for them—easier to publish papers, attract post-graduate students and get research funding.

A Notable Example of a Computer Science Professor

Professor Andrew S. Tanenbaum created Minix for teaching Operating System courses.

Many university lecturers would have dismissed his effort (huge): coding a teaching-purpose-yet-functional Unix OS, lacking value for academic publishing (nor commercial value). However, he fulfilled his role as an educator by seeking better ways to teach students, didn’t he?

Minix “directly influenced Linus Torvalds to develop Linux a few years later”. As we know now, Linux transformed the software industry, for a lot better.

Chasing hypes is quite common in commercial software development, too

When Kent Beck introduced eXtreme Programming (the first Agile methodology), many software companies dismissed it as “cowboy coding.” In 2000, when I presented XP internally (in a team brown-bag session) at Nortel Networks (my first commercial job after leaving research), my manager, Gary, really hated the concept, labelling it “cow-boy act”.

Later, as Agile gained popularity, a flood of Agile Coaches and Scrum Masters emerged—most of them, of course, were fake. I’m not here to judge; this just happens in certain fields.

What I can do, however, is ensure that fake Agile Coaches don’t infiltrate my company. My test is simple: demonstrate hands-on UI test automation on the spot. Without solid automated end-to-end regression testing, there is no Agile. Even a primary school kid can run stand-ups, drag stories in Jira, and coordinate retrospectives.

Software Startup Founders: Leverage the Potential of Young Talents

Elon Musk understood this and skillfully tapped into the potential of young talent. After laying off about 80% of Twitter's staff, the company continued to thrive. To my knowledge, SpaceX also employs many young professionals. This time, with DOGE, Musk has taken it even further.

Age is NOT the issue; the Mind Age is

It might sound strange for me, a person in the 50s, to write about this. However, I want to clarify that this is not age discrimination (clearly), but rather mind-age. Many software engineers stop learning once they graduate from university.

Here's a question for software professionals: When was the last time you read a technical book?

I’m in my 50s now, and I still (purchase and) read more technical books than my early-20 daughter (working at FAANG as a software development engineer). In my late 40s, I started re-learning C++ and re-developing the TestWise IDE (a desktop app) in C++, even though I had spent nearly my entire programming career as a Java web developer.

Currently, I am learning the Japanese language.

Many software companies now lack engineering discipline and have become overly bureaucratic

The software industry is no longer new; it has grown immensely and become a key part of the workforce in many countries.

It’s not surprising that closed mindsets and bureaucracy have developed in software companies—something I've witnessed frequently, especially as a test automation engineer/consultant. Senior managers, executives, and principal engineers often make apparently absurd decisions, such as

Hiring an agile coach to conduct various 'ceremonies' and measure fake velocity using made-up user story points.
How could this actually improve software development productivity? One defect can set back development by days, weeks, or even months. Often, these projects are riddled with defects, yet they’re labelled 'Agile' simply because they have an Agile coach or 'Scrum master' on the team."
“Your Selenium Ruby test automation indeed works well, but we insist on using Java as that’s our coding language”
E2E testing is black-box testing, with nothing to do with code.
“The Cypress demo looks great; it seems easy to use. Yes, you mentioned that Cypress doesn't support frames and multi-windows, features necessary for testing our app. I'm confident there will be a solution for this in the future.”
Cypress.io (the company behind Cypress) is dying. Every Cypress test automation I witnessed was a complete failure. For some requesting my help, I rescued with Selenium WebDriver + RSpec, quickly.
“We've decided to adopt Cucumber for E2E test automation. Despite the additional maintenance overhead you mentioned, the principal engineer believes the benefit of business analysts being able to read and update tests outweighs it.”
It failed, of course. In reality, business analysts never actually read or updated the Cucumber tests, and many of the tests failed because they weren't kept up-to-date.

Why do I use test automation to highlight this? Automation and quality control (reliable and repeatable) are the essence of engineering. Furthermore, it helps identify issues quickly because it involves continuous testing.

The Agile Way

Discussion about this post