Working at a fast-growing AI startup has taught me an infinite number of lessons that have made me a better engineer. The key aspects of this are summarised below.

**Decision Making**

Decisions in software engineering should be made on the following criteria:

- How critical is this decision?

- Is this decision reversible?

- Are we worsening the problem by adopting a workaround solution?

- Does making this decision now increase or reduce the cost of making the decision in the future?

This framework has allowed me to build a mental picture of complex decisions and make them more digestible when explaining my thought process and rationale during my time at eesel. I’ve come to realise that while the decision itself is important, the reasoning behind it matters even more. A well‑reasoned decision can’t be regretted.

---

**Problems and Simplicity**

Most things are simply solvable. Services tend to operate intuitively: if a downstream layer isn’t functioning as expected, it’s almost always an issue with your own code. There isn’t much that is truly blocked, and blockers are rarely real.

---

**Incidents and Ownership**

At eesel we’ve had a number of incidents simply because code breaks. It’s not possible to catch every issue before it occurs. What matters is the post‑incident process. Acting to resolve incidents before they ever happen again is deeply satisfying.

I’ve been responsible for many parts of incident management at eesel, including PIRs and managing incidents when services go down. As much as this can suck when you’re off work, it genuinely feels like the core of software engineering.

One of my favourite problems I’ve solved was debugging a Python memory leak. I spent almost a full day working with the Python garbage collector, tracking down a finicky leak that was taking our entire server down during periods of high load.

That experience was some of the most fun problem solving I’ve ever done. There’s something magical about working through a truly difficult technical issue all the way to completion.

---

**Communication**

Communication during incidents is one of the most important factors. Rapid updates keep people aligned and ideas flowing. It’s essentially a long sparring session.

---

**Code, Intuition, and AI**

Good code design matters even more for AI. At eesel, we’ve built systems that were thrown together quickly, and others that had real time and effort invested in them. Even with my bias toward the latter, it’s clear that good code is more important than ever.

Intuition doesn’t get lost with AI. Building intuitive services is key to unlocking AI‑driven 10× productivity.

Background agents are powerful. As much as I dislike admitting it, software engineers are moving into more senior roles by default. Understanding low‑level systems is becoming increasingly important.

---

Software engineering, at its core, is about decision making, problem solving, and owning systems when they fail. All of these skills feel even more important in an AI‑driven wo