Register for "A Guide to Data Products: Everything You Need to Understand, Plan, and Implement" - Friday, May 31,1:00 p.m. ET

Lightning Learning and Deep Work

In his book “Deep Work”, Cal Newport argues that long sustained periods of uninterrupted and focused work are becoming increasingly important today. Indeed, deep work is now required in order for you to keep ahead of some, soon to be created, AI that wants to take your job.  

But deep work is hard and requires you to concentrate, study and learn about difficult topics. There is so much to learn that the learning process needs to be as efficient and motivating as possible. The new paradigm of lightning learning is a great way to accomplish this.

Lightning learning is a learning approach that is rapid and narrowly focused (a brilliant flash or ‘lightning strike’ of learning). It might seem then, that it might not be a good match for the long hours of dedicated learning required for performing ‘deep work’.

This is, however, not true. The rules of lightning learning still apply even when you are going deep. The trick is that the learning path must be well defined. Like the stepping stones across a stream. With lightning learning you can learn one skill at a time and then, step by step, eventually you look up and realize you have made it across to the other side. The key to success is the placement of the stepping stones or learning path.

Deep work is the Zen antidote to our ADD culture

Deep work is not a new idea. It is just common sense. In a nutshell, shallow work is easy, requires limited knowledge and is more immediately gratifying. Deep work is hard, requires difficult and dedicated study and produces greater and more long-lasting results.

What make ‘deep work’ an interesting concept today is that we do so little of it when so much of it is required. Just when the best jobs require learning about very challenging topics we find our learning time overwhelmed with shallow tweets, emails and Facebook articles. Our day-to-day working lives have become fragmented and interrupt-driven. Our current jobs almost seem to reward and encourage a form of attention deficit disorder (ADD).

In contrast to this ‘business-driven ADD’, deep work is very much like the practice of Zen concentration. It is focused on one topic, with complete immersion and without distraction. It allows great and difficult things to be accomplished.

Not that shallow is all bad

That doesn’t mean, however, that shallow work (or shallow learning) is not valuable. Knowing ‘shallow skills’ is still important (like knowing how to add a new customer to HubSpot or how to add a filter to an Excel column), but they are generally skills that most eighth-graders could acquire if they cared to. Unfortunately, they are also skills that could be automated and taken over by an algorithm or an AI.

The problem is that many of us are using our learning superpowers (and limited time) to only acquire shallow knowledge to do shallow work. We’re missing out on all the benefits of performing deep work and the enjoyment that comes from deep understanding. We need to reallocate time to do more deep work and find a way to learn deeply as well.

Shallow skills can be easily replaced 

Newport proposes the following definition for deep work:

 “Acquiring and using skills that are difficult and are not easily automated or learned by others.” 

He warns us that to be successful in the future, “you must master the art of quickly learning complicated things”. He even cites learning about data science as an example of something valuable and complicated that requires deep work.

In the future, AIs will replace many routine jobs and shallow work. The creative jobs will be the ones that still require human intelligence. If you aren’t learning how to build an AI, your job might be performed by one.

Learning is different for shallow and deep work

To understand the difference between the learning required for shallow versus deep work let’s revisit an old joke. It contrasts the broad liberal arts education from Harvard with the focused engineering education from MIT:

“The difference between Harvard and MIT is that at Harvard, you learn less and less about more and more until you know nothing about everything. At MIT you learn more and more about less and less until you know everything about nothing.”

According to this joke, Harvard is shallow and MIT is deep and in the extreme neither is useful. But, of course, in practice they both are. (Ideally, you might want to get a degree from both Harvard and MIT to cover your bases.) This is true in learning for both shallow and deep work.

Here is a more formal look at some of the differences between the types of learning required for shallow versus deep work:

Learning required for Shallow Work
Learning required for Deep Work
Example: learning how to use the latest photo filter on your iPhone
Example: learning how to write the AI algorithms for the next photo filter released on the iPhone
Could be done by an 8th grader
Requires background knowledge, and mature levels of concentration and discipline
Immediate gratification
Long term gratification and not guaranteed payoff
Acquiring knowledge that could be replicated with an AI
Acquiring knowledge that can’t be replaced with an AI
Learning no more than what you need to know
Learning deeply without being sure of the immediate utility
Small time commitment
Large time commitment
Sometimes the right thing to do (e.g. Harvard)
Sometimes the right thing to do (e.g. MIT)

Deep work is really important for AI

Deep work is usually a requirement in fields of study that are rapidly changing, such as machine learning and AI. This is because these fields are generally complex and have regular breakthroughs that must be understood and assimilated in order to be used. Let’s look at how much AI has changed within its short history.

Back in the early 1990s AI was the hot new technology. Japan was running the 5th Generation Computer project to leapfrog the US in AI. And thanks to a book of the same name by Edward Feigenbaum and Pamela McCorduck, Japan’s AI initiative was considered to be an existential threat to the United States’ technological and economic superiority. All hope, at that time, rested on the programming language LISP for creating AI programs and the LISP machine computers to run those programs. You had to know LISP very well if you wanted to do AI.

But just twenty years later things have changed dramatically. The American economy is strong, Japan is a powerful economic ally and the knowledge of LISP, as a programming language for AI, has become nearly inconsequential. reports that LISP is now used just 0.4% of the time for AI programming and Python (a language that had almost no use for AI just ten years ago) accounts for 65.8% of all AI projects today. Time to apply some lightning learning to acquire Python skills.

Lightning learning is good for shallow and deep work

Lightning learning is based on the idea that knowledge needs to be acquired quickly with a ‘just in time’ approach that is focused and directed. This is an excellent approach for shallow work where you might just need to watch a 5-minute video to gather the knowledge that you need.

But lightning learning can be used for deep work as well. The trick is to have a good plan that ties together a series of rapid and small learning events into more complete knowledge. This is accomplished via a well-defined BOK (Body of Knowledge), curriculum or learning path that leads the learner from one small accomplishment to the next until a sizeable and otherwise intractable subject area is mastered.

Such an approach provides all the benefits of lightning learning’s speed while expanding its use from simple shallow work projects to deep work. In both cases, the learner will be acquiring small flashes of knowledge that are very focused. For shallow work, a few flashes may be all that is required. To accomplish deep work, the learner may need many such flashes, all organized into a prolonged ‘thunderstorm of learning’. At the end of the thunderstorm deep work can be accomplished.


Cal Newport’s book “Deep Work” is a call to action for everyone working in the hi-tech industry today. It reminds us of what we all used to know quite well: sometimes you need to really concentrate to get something important done. That thing is usually something considered to be intellectually difficult. Lightning learning provides important principles to accomplish deep work as rapidly and effectively as possible. The key to success is to have a well-defined learning path.

Some predictions: four years from now

In four years, we will look back and see that we have made very rapid progress in improving and deploying lightning learning for accomplishing deep work:

AI will help. AI, machine learning and data science will help to solve some of the shortcomings currently inherent in eLearning and to apply the principles of lightning learning to accomplishing deep work. AI will be used to provide ‘just in time’ knowledge. Recommendation engines will be applied to learning to recommend the ‘next best skill’ for someone to learn based on their personal preferences and the success of other students who came before them – effectively building the learning path on the fly. These algorithms will make achieving a learning goal enjoyable and rapid. It should be a virtuous cycle of teaching students how to build better AIs and then having those AIs help make learning better for the next group of students.

Deep work will catch on… sort of. It would be lovely if Cal Newport’s ideas catch on and we can balance our lives and our learning between required near term work and much more difficult, but eventually more beneficial, deep work projects. However, the reality is that deep work is hard and not everyone will choose to do it. But for those who do, the rewards will be significant. Newport’s insight wasn’t that deep work exists and is important (we all knew that already), but rather how our current ADD business and social culture makes it so hard to perform.

Shallow learning will still be important. There is nothing wrong with shallow work or shallow learning. In fact, look for colleges and business education to embrace and teach shallow knowledge to their students by breaking much bigger topics down into skills that can be learned in a few hours rather than a few years. Lightning learning techniques will make this happen.


“Deep Work”, Cal Newport. 2016.

Python leads the 11 top Data Science, Machine Learning Platforms: Trends and Analysis. Piatetsky”. KDnuggets. 2019.

“The Fifth Generation: Artificial Intelligence and Japan's Computer Challenge to the World”, Edward Feigenbaum, Pamela McCorduck

Stephen J. Smith

Stephen Smith is a well-respected expert in the fields of data science, predictive analytics and their application in the education, pharmaceutical, healthcare, telecom and finance...

More About Stephen J. Smith

Books by Our Experts