The Shortcut Is... Using Your Brain
Why the real shortcut to success in ML engineering isn’t what you think.
There will not be any office hours for the next two weeks (March 31 - April 14th)
Many of the folks I work with come to me looking for hacks:
What’s the hack to passing the interview?
What’s the hack to onboarding faster?
What’s the hack to getting that sweet, sweet promo?
What’s the hack to getting into ML in the first place?
Honestly, I get it. The brain evolved to conserve energy. Thinking is metabolically expensive. And in tech, we’re trained to optimize, automate, and abstract. So of course we go looking for the shortcut.
But ML engineers take this to another level. I don’t know if it’s the abundance of frameworks, the obsession with leetcode, or the meme culture of “vibe-coding,” but a lot of smart people are trying to coast through hard problems with the least amount of cognitive friction.
Use your brain
Early in my career, I was one of them.
I had gotten pretty good at applying ML patterns to problems. Linear regression here, a little XGBoost there, throw in some scikit-learn preprocessing, ship it. But then I ran into a problem that didn’t fit the pattern.
I asked my teammate Kevin what the trick was.
He looked at me and said:
“Use your brain.”
It was an epiphany.
I had outsourced the hard part: thinking. And Kevin’s blunt advice forced me to re-engage. Not with some magical insight, but with deliberate thought.
That’s the thing no one wants to hear…
Engaging your brain is the shortcut.
It looks slower at first. It feels scarier.
But it pays off faster than any hack.
You retain more. You build reusable knowledge.
And when the next problem (or interview) rolls around, you’re not googling for your duct-taped solution—you understand what’s going on.
This is especially obvious with tools like ChatGPT.
Yes, it’s great at offloading mental effort.
Yes, it’s fun to vibe-code for small tasks.
But if you’re trying to level up your career, and you let it do all the thinking, you’re giving up the best muscle you have.
Use it tactically, not habitually.
Add friction, not convenience, to force engagement.
At Meta, I saw this play out in onboarding too.
Some folks coasted through onboarding and were shocked when the job got hard. Others struggled in onboarding, but ramped faster when they hit real projects.
Onboarding should be hard—so the job gets easier.
It’s better to front-load the mental pain. Learn the weird parts of the codebase. Ask the “dumb” questions. Dig into the tooling. Accumulate real knowledge. The thing is, you can be the “dumb person on the team” when you start, but it is much harder to pull off a year later.
It takes a bit longer per problem.
But you’re building a foundation that lasts.
One more tip: write things down.
Not just notes—compressed notes.
Writing forces clarity. Reviewing forces repetition.
That combo = sticky memory.
The Neuroscience of “Think First”
Cognitive effort feels hard because it is hard. The prefrontal cortex, responsible for reasoning and decision-making, burns more energy per gram than nearly any other tissue in the body. That’s why our brains evolved habits, patterns, and automation: to conserve fuel.
But deliberate thinking activates deeper learning.
Research shows that effortful retrieval strengthens memory (Roediger & Karpicke, 2006).
And when we struggle slightly before solving a problem, we encode it more deeply (Bjork’s “desirable difficulties”).
So yes, it’s harder. That’s the point.
How To Engage Your Brain (and Still Use ChatGPT)
1. Force yourself to think first.
Before Googling or prompting ChatGPT, write down what you think the answer might be. Even if you’re wrong. This primes your brain.
2. Use ChatGPT for reflection, not replacement.
After solving something, ask ChatGPT to critique or optimize your solution. You’ll get feedback without outsourcing the core effort.
3. Introduce friction by design.
Block certain websites during deep work. Use pen and paper for design docs. Turn off autocomplete temporarily. Anything to make you do the work.
4. Compress your knowledge.
Write atomic notes on problems you’ve solved. One note = one concept. Add tags, links, and revisit often.
5. Practice desirable difficulty.
Mix up topics. Quiz yourself. Try coding problems without looking at docs. If it feels uncomfortable, you’re probably doing it right.