Tech hiring has become really hard…

Not from a lack of talent (Morocco has nothing to envy other markets), or a lack of time (we’ve got tools for that—shoutout to Invirtus).

Simply because we no longer know what to look for in a tech profile. In this era of heavily subsidized AI-coding assistants, tomorrow’s software engineer is a super-chameleon who adapts to a constantly shifting environment.

So here are my 4 core criteria (for early-stage startups doing intensive agentic coding)—and I’ll let the community debate the rest :)

1. The Systemic Decomposition Specialist => They listen to a startup pitch or a business problem and maintain a high-level view of what’s being built, complete with its strengths and flaws. They pinpoint ambiguous areas and isolate them within the agents’ workload. They anticipate test cases or the need for greater architectural generality. This is fundamentally someone who knows how to step back, see the big picture, and ideally carries the battle scars of years spent deploying into complex enterprise systems.

2. The Stochasticity Reducer => This one specializes in analyzing and predicting failure modes in AI processes (which are inherently unstable). Much like an engineer building in an earthquake zone, this technician sets up evaluation and testing harnesses and keeps a Plan B ready for every fragile section of the codebase. Hallucinations and model drift are their worst enemies. They bring peace of mind to business teams and make the application far more deterministic.

3. The Entropy Hunter => A real sniper who senses and eliminates unnecessary local complexity wherever agents produce it. A re-coded function instead of using an existing library, an overcomplicated method with zero business rationale. They spot an LLM API call looping five times when a single, broader call would have done the trick. Code review is a must for them, but they can smell a bad pattern without even digging into the fine print.

4. The Paranoid Tester => This person trusts nobody—least of all the 5 Opus 4.7 agents looping non-stop since last night at the mercy of rate limits. They test everything that moves: between every user story, every snippet of UI, and every API call, determined to keep an accurate mental map of what’s happening under the hood. Refusing to get overwhelmed by the volume of generated code, they take the time to digest the complexity of every feature (even though they aren’t the product owner) and re-adjust the agents before a major feature drift takes hold.

Does your tech team have these 4 skills? Do you have these 4 skills yourself? Do you see any other macro-skills critical for this new era?

A few clarifications

1. Analyzing these profiles highlights a clear reality: each of these four postures requires significant seniority. The real challenge for the tech world moving forward will be our capacity to train juniors so they can actually acquire these skills… (thanks Oualid for pointing this out)
2. Context Engineering has also emerged as an essential role in bridging the human and agentic worlds—specifically in breaking down the information asymmetry between these two perspectives… (thanks Yahya for pointing this out)

← All thoughts