If you work in tech, and especially if you work anywhere near AI, it’s easy to feel like you’re constantly behind: new frameworks, new models, new tools, new papers each week.
The natural response is Just in Case (JIC) learning: reading the book, watching the videos, or following the tutorial so that, just in case, you’re ready when you need it. I’ve done plenty of this myself. And while JIC learning can feel productive, a lot of it turns out to be wasted effort.
The problem with Just in Case learning
Technology moves fast. Faster than most learning plans. When we invest heavily in learning specific tools or workflows without a concrete need, we’re betting that those details will still matter later. Often, they don’t. There’s also a deeper issue: JIC learning tends to be passive and shallow.
Without a real problem to solve, there’s little context. And without context, retention is poor. If the knowledge isn’t used soon, it fades.
The result is a false sense of progress: we felt busy learning, but little of it shows up in our work.
Just in Time learning as a core skill
By contrast, Just in Time (JIT) learning happens when there’s an immediate need: You’re building something, you’re solving a real problem. So you learn exactly what’s required, apply it immediately, and move forward.
After more than 25 years in software, I can’t think of a single meaningful project I’ve worked on that didn’t require JIT learning. Every project had something new:
- A language I hadn’t used before
- A framework I’d never touched
- A domain I knew nothing about
- A constraint that forced a new approach
What mattered wasn’t having pre-existing knowledge of everything. It was the ability to learn quickly, with intent, and in context. That ability compounds.
Strong JIT learners:
- Know how to find the right information
- Can filter signal from noise
- Learn with a clear goal in mind
- Retain what they learn because they use it immediately
In practice, this skill is far more valuable than trying to “stay ahead” of every trend.
When Just in Case learning does make sense
This doesn’t mean JIC learning is always bad. It makes sense when the learning is foundational and unlikely to change. Mathematics is a great example. Linear algebra and calculus have been around for centuries and will be relevant long after today’s ML frameworks are obsolete.
When I first learned machine learning, I took Andrew Ng’s course. At the time it was taught using Octave, an open alternative to Matlab. I remember almost nothing about Octave itself.
But the foundational concepts stuck:
- Overfitting vs underfitting
- Gradient descent
- Impact of regularization
Those ideas have served me across many different tools, languages, and systems. The surface-level details changed. The fundamentals didn’t.
That’s the sweet spot for JIC learning: durable concepts that transfer across contexts.
Being selective is the real advantage
If you’re feeling overwhelmed by the pace of change in tech, the problem may not be your ability to learn. It may be what you’re choosing to learn. Useful questions to ask are:
- Is this foundational, or is it tool-specific?
- Will this still matter in a few years?
- Do I have a concrete use for this right now?
Being deliberate about JIC learning and excellent at JIT learning is one of the highest-leverage strategies for a long, sustainable career in tech.
You don’t need to know everything. You need to know how to learn when it actually matters.