Everybody must use AI at the workplace. If you’re not using AI, you’re not making effective use of your time. Period.

The romantic engineer

Some romantics are still attached to their older ways of working. Getting a problem, designing the algorithm with pen and paper, then coding it out, testing it, debugging it, optimizing it. Figuring out why it isn’t working, why it suddenly stopped working, how the same piece of code was working earlier. The Zen state of coding uninterrupted for hours together.

Believe me, I get it.

I’m a hardcore romantic as well. I love the idea of sitting in the corner of the office with black coffee and audiophile headphones, as I bang a mechanical keyboard loud enough for every star in the galaxy to hear. I love obsessing over the nits - obsessing over how declarative code is more readable than imperative code, debating how a React component shouldn’t be thought of as a class but as a pure function of state and props, insisting that no piece of logic must ever be repeated. Choosing to build simple libraries myself just so I can optimize bundle size and avoid unnecessary third-party code creeping into my code paths. I absolutely devour these things.

But it’s history now.

We still review AI code, but it’s time we accept that the rigor we once had around low-level design is gone. I still try to obsess over it in code reviews, but I can clearly see code quality degrade over the past couple of years. And I don’t see us ever getting back to that world. Things are only going to get worse. AI abstraction is going to get even more abstract and even more easy to use.

But it’s okay.

It took me time to accept this, but it’s fine. Things work well. There are no obvious performance issues, and our latencies are far better than what SLAs demand. People launch things fast, get feedback fast, and products get better fast. The internet keeps improving faster.

Orthodox engineers like me would have preferred to control design more tightly and optimize things better, but there’s no real choice here. AI has “happened” to us. The world has become extremely fast and extremely competitive. It’s capitalism. Only performance matters. Everything else is a facade, no matter how much companies pretend otherwise. It’s like performance-enhancing drugs in bodybuilding. If you want to compete, you use them. If you don’t, you lose.

But the romance must not end.

Redefining romance

The romance must shift.

The romance now is being able to control AI in a way that it builds the best possible software, saves massive amounts of time, and empowers you to execute at a scale you never imagined. We can still build beautiful things. In fact, we can build more, better, and faster. And all of that is defined by the how. How you use AI.

I truly believe AI augments us rather than replaces us, if we learn to use it well.

Our comprehensive study of AI-assisted development tools across 300 engineers provides statistically robust evidence for both the significant potential and practical limitations of integrating AI into software development workflows. [0]

The noise

And that’s where the noise comes in.

The number of AI tools thrown at us lately is mind-boggling. I don’t want to blame the builders. Builders will build, and they should. But we are spoilt for choice today. There’s AI in our cars, AI in our code editors, AI in design tools, AI in our documents, AI in our wallets. Everyone is competing to build autonomous AI agents that automate your day-to-day work. There are assistants, agents commanding sub-agents. There’s AI with IoT, AI for therapy, AI for dating, and what the fuck not.

The big dogs are building models and integrating with everything - OpenAI, Claude, Perplexity. Smaller players promise better platforms and better experiences. Everyone promises intelligence.

Limited time and bandwidth

But we still have just one brain. Limited attention. Limited bandwidth.

So what do we make of all this intelligence being thrown at us?

I felt genuinely overwhelmed by this question for the past few years. I was wasting time onboarding tools I didn’t really need, depending on systems that looked impressive but didn’t solve the problems I actually had. I was choosing tools because they were offered to me at peanuts price, not because I had clearly defined what I needed.

Eventually, out of that overwhelm, I adopted a way of thinking that helped me use AI in a way where its utility was clearly scoped. This came from wasting time, making bad choices, and realizing that the real problem wasn’t AI capability, but the space in our work lifecycle.

When we choose an AI tool, do we actually know why we’re choosing it? Are we choosing a solution to a real problem, or are we impressed by a powerful demo that looks like it could revolutionize how we work? What about it should we obsess over? Latency? Accuracy? User experience? Beauty? Time saved? Our business metrics?

Most of us don’t stop to ask this. And when we don’t, we delegate by reflex. With over 20 AI tools in the pocket, every task starts looking like an AI task. This reflexive delegation is now being called cognitive offloading [1] - a loop where thinking slowly disappears and laziness gets normalized as productivity. The result isn’t leverage. We lose our human edge - and that’s how the noise wins.

This is where the TIE framework begins.

I believe that if we can lay down a principled way of building a personal AI operating system - one that is principled, logical and grounded in real problems at work, we can reason clearly about where an AI solution fits, what value it adds, and whether it deserves a place in our system at all.

I’ll be writing about the framework I use in the upcoming parts of this series. I call it the Time Intelligence Economy framework. It’s what helped me silence the noise and focus only on signal. Over the next couple of weeks, I’ll be covering:

1. The AI Noise (current): Why AI is unavoidable at work, how noise creeps in, and why choosing how you use AI now matters more than the tools themselves.
2. What must stay human?: AI claims it can do all things: coding, planning, analytics. What’s actually true? Where do humans still matter?
3. Divide and Conquer: Active vs Passive: Not all AI tools are equal. Where should you obsess over latency? Where should you obsess over accuracy? Where should you not care about anything and just jump on it?
4. Active AI tools: Is AI helping you use your time better, or are you sitting idle while an LLM “reasons”?
5. Passive tools and digital employees: Are digital employees real? The truth about automations, workflows, and agents that work without you watching them.
6. Building your personal AI operating system: Beyond philosophy, what are the must-haves? How do these tools come together into a system that compounds?
7. My AI stack and workflows, explained: A case study, not advice. How I use AI to multiply the value of every second of my work life.

If this resonates, drop your email below. I’ll send each piece to your inbox as it’s published.

---

References

• [0] - Measuring AI’s true impact on developer productivity - https://arxiv.org/abs/2509.19708
• [1] - AI Tools in Society - Impacts on Cognitive Offloading - https://www.mdpi.com/2075-4698/15/1/6