The problem with PhDs is not the workload, but the lack of skill at structuring information. Unis/PIs don’t teach that. It’s like running a bank, but having no clue about accounting. I’ve finished my PhD in ~2.5 years and built a business on the side. It’s possible.

Read Papers as Diagrams, Not Text. This AI Skill is a game changer for academics.

AI summaries of papers are useless (just read the abstract), but a diagram of its findings genuinely boosts understanding. Effectively, it’s an AI-generated graphical abstract of everything the paper presents. This post teaches you how to build the skill for your domain.

Here’s an example for a biology paper:

This is a whole paper, mapped in one image.

This is a whole paper, mapped in one image.

In order to build such a diagram, or teach an AI to do it for you, you need to define a “visual language”.

1. Define Your Visual Language

Note a few key facts about the diagram above:

• Every box is a concept (or a “noun”).

• Every arrow is a relationship (or a “verb”).

• The colours and shapes carry meaning I defined once.

Result: I can read the paper’s logic at a glance and immediately see where the mechanism has holes.

This works because the diagram has a key property:

Every path through the diagram yields a valid sentence. Example:

Walking the top arch, you could say “Plant traits determine the intrinsic growth rate (of a plant), which is used as an estimate for its Fitness”. That’s a claim someone can be making.

Next, you need to define what the colors and shapes mean.

Remember: A box is a noun, a concept. An arrow is a verb, a relationship. Give each type a color and a shape you can remember.

Now, instead of writing the verb on the arrow, you rely on the shape/color, leaving the label for a reference.

For example, in the diagram above: “Media bias causes misinformation (Miller 2020)”

See how this already resembles academic writing? We are transforming academic writing into a visual representation.

This is the visual language! The AI can’t invent it for you.

So experiment manually on a few papers to derive a preliminary language.

Here is an example of the visual language used for a diagram like he one above:

2. Build a few examples with Draw.IO

AI performs best with examples. So to get a really good skill that resembles your thinking, you need to build a few examples.

It’s fun! Just read a paper and try to map it out. When you fail to express yourself properly, adapt your visual language, legend, and retry until you are proud and happy with the result.

Here is how one of my hand-made diagrams look.

Keep in mind that the AI will not only copy the language but also the style of connections you use (curved vs angled), how far apart the boxes are spaced etc.

Make it beautiful!

3. Build the Skill

Since the visual language is personal. You will need to build a skill and can’t just download it.

That’s what AI enables anyway: Personalisation. Use it to build the perfect solution for you.

Here’s how it works:

1. Create a draw.io template.

Put your legend on one page and your example diagrams on another. The more detailed these are, the closer the output.

2. Give the AI the draw.io format. AI needs to know how these files are built technically. To teach it just install a draw-io diagram skill as the baseline (there are a few available). This part never changes.

I used this one: https://github.com/github/awesome-copilot/blob/main/skills/draw-io-diagram-generator/SKILL.md

3. Build your skill.

Upload your legend file (1), link the technical skill (2), and describe your goal (3). State what goes in and what should come out. Don’t script every step. Modern models figure that part out.

4. Test and iterate. Pin the conversation that created the skill. When it produces an error or an ugly diagram, go back, show the problem, and let the AI fix it. Mine got much cleaner after one round of feedback.

5. Finish it by hand. AI gives you a first draft, not a final answer. Double-check every connection yourself. That’s where the understanding actually happens.

A serious word of warning

Any monkey can use a skill to create something that looks impressive (and scream loudly on X about it…)

Don’t be that monkey.

Use these skills as a starting point to build understanding. What happens here is that you “translate” text into a visual representation, and a picture can say more than a thousand words. So essentially, you are using a shortcut to reading.

But even if you are using a shortcut, you still have to walk.

Critically engage with the material and don’t fall into what is called “cognitive surrender”. This describes a process where we surrender to the AI because it is too confident, the subject matter is too complex, or we are too lazy.

Paired with the possibility that the skill can make blunders, cognitive surrender can kill an academic career and erode your ability to do science.

So put in the effort and engage with what you are learning.

4. Finding a Research Gap with Mind Maps

If instead of a paper, you map out the entire domain (this takes a while), you unlock a bonus ability: “Spotting research gaps”.

There is no single recipe to find a research gap, but once you have mapped out everything you know, what you don’t know becomes quite apparent.

Here are some ways to spot a research gap:

• Missing connections are the biggest indicator of research gaps, especially if they can be replaced with a hop over multiple other connections. This could help explain why the gap exists and what the solutions might be.

• Connections often have associated studies. A lack of studies indicates that something might be speculative.

• Similarly, if the studies are too old, you could reassess the findings and contrast them with what is known today.

Becoming an Effortless Academic

This is one of many strategies and tutorials I teach at Effortless Academic.

It’s all about using digital tools and AI responsibly to struggle less in academia.

This nature study found that most PhD students work 40+ hrs.

Ilya Shabanov@Artifexx·Jul 4The problem with PhDs is not the workload, but the lack of skill at structuring information.

Unis/PIs don’t teach that.

It’s like running a bank, but having no clue about accounting.

I’ve finished my PhD in ~2.5 years and built a business on the side. It’s possible.Quotenature@Nature·Jul 2Can a PhD be pursued 9-5?

https://go.nature.com/4b5pPZP11194K

→ I worked closer to 20hrs during my PhD, still finished in 2.5 years, built a business on the side and had 5 first-author papers (3 published) at the end.

The secret: Digital information management with Obsidian and some AI (I started in 2023…). In 2026, you can do so much more.

Subscribe here (for a free 21-day crash course) if you wish to learn more: https://effortlessacademic.com/academic-quick-start-guide/