Blog post

Why Juniors Are Important

Fabian Walther

There is no substitute for making a bad decision and slowly realising that. Often at a point where that decision led to a constant uphill battle dealing with the fallout. But more often bad decisions really do their impact when no one can even connect the dots anymore and what started all the mess in the first place. Realising why something didn’t work out as planned is – from my perspective – the most effective way of learning and becoming a person with professional experience. Hence the name: You have to experience something in all its glory to really remember to make it better the next time. This takes time. You can’t fast-track that fundamentally with AI. The human mind needs its time to let knowledge and a sense of good taste and judgement sink in.

Currently there is a tendency in the industry to bet everything on persons with experience getting amplified by AI. Getting into the industry is becoming more and more difficult. This is fundamentally wrong for two reasons.

Sustainability

The obvious one first: this is not sustainable for every company that isn’t a startup only trying to survive until Series B funding. If you only rely on seniors (and the demand for seniors is higher, not lower with AI) there will be no one taking over (with slowly built up domain knowledge and experience) when your seniors move on to the next opportunity. And they will, when realizing they are working in an unsustainable environment just draining knowledge out of people which they built up somewhere else. Company culture is ubiquitous. Not investing in the personal growth of new employees often also means: I don’t care about people at all.

Learning is not a one-way street

The second reason is even more important: Gaining experience by learning together with other humans is not a one-way street. I’m deeply convinced that I must be able to explain a possible solution to someone not deep into a topic to truly understand it myself. My decisions get better by discussing them back and forth with people with an unbiased, fresh point of view. Also, teams only consisting of seniors “knowing exactly how everything should be done” tend to block each other. Often there is more than one good solution and you just have to decide. No decision at all is a very bad decision because you won’t even know what does not work. Let alone what would have a good outcome.

Review decisions, not code

In teams we see seniors getting overwhelmed by huge amounts of code generated by juniors. We should reject that tendency immediately. Politely nonetheless. Huge pull requests after weeks of work from the basement thrown over the wall to a colleague always were bad behaviour. Only accepted because of the (maliciously weaponized) sunk cost fallacy. This bad behaviour is only amplified by AI usage. Encourage reviewing ideas and challenge decisions instead. Don’t review code, review decisions leading to code. Use AI to generate less code and invest time in discovery and discussion to generate code solving the right problem in the first place.

You can’t learn from mistakes in generated code. Because there was no effort involved in creating it. No one will remember, just change and move on. You can only learn if you have to invest a significant amount of time and energy into the rework. Correcting randomness generated by an LLM is not sustainable. Correcting your beliefs and mental model after a healthy discussion is.

The Contribution of the Agentic Trio

And this is exactly what one could try to facilitate with the Agentic Trio: Encourage learning from meaningful interaction with people from different crafts to gain a holistic view of the software development lifecycle instead of trying to battle with huge amounts of generated code creating team friction and grinding everything to a halt through Slop and mental team overload. Division of labor hasn’t happened just because of massively scaling the output of program code. Specialization and the synergetic cooperation of experts and newcomers alike lead to software and systems solving the right problems the right way.

Call to Action

Don’t stop hiring juniors because AI makes seniors more productive. Rotate newcomers through the Trio roles, let them challenge decisions, and review the thinking that leads to code rather than the code itself.