S2 = The curiosity year

2024 feels like it reshaped so much about how we, the cogs and gears of the creative economy, spend our days — making me wonder how our work might evolve next.

A more clickbaity title for this piece could be “How not to lose a job in 2025”, but that feels a bit off. The future will definitely bring huge changes to how we work, but the reality is quite nuanced. AI won’t replace most of us anytime soon — if anything, it’s becoming clear that human-AI collaboration (“human in the loop”) works far better than either human or AI alone. I don’t see much sense in panic, and I don’t believe in any version of a “Terminator” future (nor a “Dune” one).

Yet, what I truly believe is that people in this future will more and more become generalists.

“Product developer”, Leonardo da Vinci, 1490

---

For me, it has always been really hard to specialize — choosing between math and social sciences in school, engineering and design at university, hands-on work and leadership in my career.

At first, it felt like a weakness — the “Jack of all trades, master of none” vibe. Later, I saw it as just one of my traits — “sometimes, it’s useful to have a bunch of different things done by one person”. Today, though, it feels like a key ingredient for doing great work.

A long time ago, I read Range by David Epstein, and it had a big impact on how I think about the generalist approach. David’s main point is that specialization works best in “kind” environments — those with clear, stable rules and repetition (e.g., sports, legislation). However, it’s less effective in “wicked” environments — ones that are constantly changing and full of uncertainty (e.g., medicine, technology). Generalists thrive in these settings because they can adapt to change, bring a diverse perspective, and transfer knowledge across different domains.

Now, in the intelligence age, it seems like the environment is becoming more and more “wicked”, while “kind” areas are disappearing every day. The best thing we can do in such a world is to follow our curiosity, trust our instincts, and explore as many interests and identities as possible rather than committing to a single, highly specialized skill or area of knowledge.

Here are three observations on this shift from areas I know a bit about: design, engineering, and management.

Observation 1 — The rise of design engineers

I’ve been doing design for a living for several years, and I’m fascinated by how this huge, complex field is transforming so quickly.

To me, the most important things a designer should have today are curiosity and taste. These qualities are so valuable and so rare that the best hiring filter I found this year was just asking what product or technology candidates recently found truly excellent.

You can quickly learn a lot of specific skills, but these two are very hard to study. There are no courses — you need to invest in them every day by exposing yourself to what happens outside of your silos.

The rest of a designer’s skillset is quickly changing, and the biggest trend I see is the growing demand for designers who ship — not just influencing the product through mocks and specs, which are far removed from users, but directly shaping the real product that users interact with.

It has always been important for product designers to be technically grounded — to deeply understand constraints and be able to talk with devs in the same language. But now, designers can add to their systems thinking tools that democratize building technology, like Claude, v0, or Cursor. As a result, you don’t even need to know code to build something real — you just need to precisely understand what you need, clearly explain it to the AI, and curate the final result.

Building a web application in Cursor: an 8-year-old's perspective

That said, knowing at least some code dramatically increases your ceiling. Designers who can code and design engineers — a new and increasingly popular role at the intersection of design and engineering — now have an extensive toolset to quickly prototype their solutions. We’ve already been transitioning from Figma mocks as the main focus of design work to production code, and 2024 has greatly accelerated this shift.

Not long ago, we realized that designers can’t just focus on pixels — instead, they need to deeply understand their product and users. Similarly, designers are now becoming more like generalists, with an increasing overlap with engineering skillsets.

This trend isn’t just about designers. The intersection of different roles is expanding from all sides. Today, the best companies building products want their engineers to care about craft as well. A great example of this — though a bit extreme for me — is how engineers design at PostHog.

Having such a team of generalists prototyping the path is a really nice way of building things. When our work is live and in our and our users' hands, we can postulate, debate, prototype, test, and move with greater speed and impact. We can look at what works, look at what doesn't, learn from it all, and keep moving.

We’re quickly transitioning from double diamonds, waterfalls, and sticker-heavy board processes to a much more collaborative, real, and, frankly, fun job.

Observation 2 — AI starts writing code

At the time I was at university, some of my peers learned Assembler — a low-level machine language.

A simple program in Assembler looks like this:

section .bss
    result resb 1

section .text
    global _start

_start:
    mov al, 5
    add al, 3
    add al, '0'
    mov [result], al
    mov eax, 4
    mov ebx, 1
    lea ecx, [result]
    mov edx, 1
    int 0x80
    mov eax, 1
    xor ebx, ebx
    int 0x80

I have no idea what this means except that, according to ChatGPT, this program basically adds two numbers.

In Python — one of today’s most popular programming languages — this would look like:

print(5 + 3)

The difference is huge. Even if you don’t know anything about Python, you can easily understand what these lines of code do.

I believe this is not the end of the evolution. The core value of programming isn’t about knowing all the details of formal syntax. At its essence, it’s about structuring an abstract, vague task into a clear set of requirements. These requirements could range from low-level instructions like “put this number into this memory cell” to more human-readable commands like “print 5 + 3”. Over time, programming has become increasingly declarative and readable, and with the rise of new language models, it’s now possible to finally translate it into actual human language — and back.

AI agents and assistants are a relatively new area, but we already have tools like:

• Claude — a simple assistant that can build apps and games in seconds without requiring any knowledge of code.

• v0 and Replit Agent — more code-oriented prototyping tools that let you build and iterate just by prompting, but also allow you to switch seamlessly to writing and editing code when needed.

• Cursor and GitHub Copilot — AI code editors and chatbots that 10x speed up and simplify the daily work of traditional software engineers.

That said, we’re still far from fully automated software engineering agents. As I mentioned, the collaboration between humans and AI is much more powerful than either one alone. If you want to build, you still need to learn code. In fact, the return on learning to code is steadily increasing. You just have to do much less of it — freeing up time to focus on what really matters.

As a short glimpse of the future, I highly recommend checking out the demo at the start of the next video to see how working with an agent looks today:

Further on, the great engineers will be defined by systems thinking, communication, explanation, and orchestration skills — not just by memorized knowledge of specific tools or systems. Once, we had a market full of frontend developers, backend developers, Java developers, and any other overly specialized roles. In the future, engineers will increasingly become generalists — just “builders”.

One of the curious consequences of this is the lowering of barriers and changing job requirements. For example, an experienced school teacher, skilled at understanding and explaining complex theoretical concepts, could, with just a few months of reskilling, become a far better software engineer than a fresh graduate who spent a year learning the intricacies of specific programming frameworks. This also highlights why it has always been cool for engineers to learn at least the basics of math.

Observation 3 — Switching to “founder mode”

If you’re on Twitter, you probably haven’t escaped “founder mode” this year. The term originated from a talk Brian Chesky gave at YC, which later turned into a highly influential essay written by Paul Graham.

Founder mode is a leadership style of founders who stay deeply involved in all parts of their company, working directly with teams and being in the details, even as the company grows. Brian and Paul argue that while traditional advice suggests founders should switch to a hands-off management approach as their company scales, this shift can distance them from the company’s core and hurt innovation and growth. Instead, they recommend to remain involved in the details.

I see this is another example of the growing popularity of the generalist approach. To effectively collaborate with different teams and functions, you need to have a good understanding of what they do — you need broad expertise.

This year, I saw firsthand how much more powerful this alternative approach can be. At TripleTen, our product brings together a lot of moving parts — product development, engineering, curriculum design, the learning community, the career team, and many more. Aligning and orchestrating all these people has always been a big challenge, and it has only been resolved by our new CPO, who brought a radically different mindset and a strong willingness to roll up her sleeves.

We’re wrapping up this year in really great shape — more aligned than ever with a clear direction ahead. This is largely the result of having a single person capable of working closely with with all the key areas of the product, draw boundaries between them, and give each one a clear focus. I don’t think anyone could have been nearly as effective without being a generalist.

According to OpenAI, the final stage of the path to artificial general intelligence is organizational AI — systems capable of running entire companies by orchestrating workflows and managing autonomous agents. The fact that this is considered the last step highlights that orchestration remains the most complex and challenging task, as well as the hardest to replace.

Ironically (or not), the takeaways for managers are quite similar to those for designers and engineers:

• The future of the role is about curating and orchestrating (human and artificial) agents.

• To do it well, you need to know a little about a lot of different areas.

Instead of a conclusion

While I’m writing these lines, there’s a curious debate about American culture unfolding on Twitter.

To me, it feels in a way related to the thoughts above. I believe that if you want your kids to succeed, you don’t need to force them to choose between Whiplash and Friends, between science competitions and cartoons, between books and TV.

All in all, this diverse perspective, this cross-disciplinary knowledge, this range of experiences, and the things you’re genuinely passionate about — make you truly stand out and become your unique imprint that no other human, skilled migrant (hi there!), or AI can replace.

Definitely not in 2025. Happy new year!

---