- the master
- Posts
- How to read AI papers 10x faster?
How to read AI papers 10x faster?
You're reading research papers wrong (and wasting hours)
Today's newsletter breaks down a three-step method for reading research papers faster and better. You'll learn what to read, what to skip, and when to stop.
In today’s edition:
Paper Unfold— How to Read a Paper?
Build Together— Here’s How I Can Help You
75 Hard AI Engineering Challenge - Free 75-day guide to master the entire AI engineering stack and be in the top 1%.
Sponsor Spotlight
AI-native CRM
“When I first opened Attio, I instantly got the feeling this was the next generation of CRM.”
— Margaret Shen, Head of GTM at Modal
Attio is the AI-native CRM for modern teams. With automatic enrichment, call intelligence, AI agents, flexible workflows and more, Attio works for any business and only takes minutes to set up.
Join industry leaders like Granola, Taskrabbit, Flatfile and more.
[Paper Unfold]
How to read AI papers 10x faster?

Reading research papers is a skill nobody teaches you.
Most of us waste hours trying to understand a single paper, getting lost in the details, and ending up frustrated.
If you are an AI engineer, PM, or technical leader, you probably read dozens of papers every month.
Here's a systematic approach that will save you hundreds of hours.
S. Keshav from the University of Waterloo created a simple three-pass method that researchers have used for over 15 years.
You can access the research paper here.
Let’s go.
The Three-Pass Method

Think of reading a paper like zooming into a map.
You start with the satellite view, then street view, then finally walk through the neighborhood.
Each pass has a specific goal.
[First Pass] 5-10 Minute Scan
This pass answers one question: Should I spend more time on this paper?
What to do:
read the title, abstract, and introduction carefully
read all section headings but skip the content
read the conclusions
glance at the references and note which ones you recognize
The Five Cs Framework
After this first pass, you should answer these five questions:
Category
what type of paper is this?
is it presenting new measurements?
analyzing an existing system?
describing a prototype?
Context
how does this relate to other work?
what existing research does it build on?
Correctness
do the assumptions make sense?
are there obvious red flags?
Contributions
what are the main contributions?
what is actually new here?
Clarity
is the writing clear?
can you follow the logic?
Stop after the first pass if the paper does not interest you, if you lack the background to understand it, or if the authors make questionable assumptions.
This pass is perfect for papers outside your core area that might become relevant later.
Most reviewers and readers will only do one pass.
Your paper needs to communicate its value in those first 5-10 minutes.
Use clear section titles.
Write comprehensive abstracts.
Make your contributions obvious.
If a reviewer cannot understand your paper in one pass, expect rejection.
[Second Pass] One-Hour Deep Dive

Now you read with care, but you still ignore the heavy details like mathematical proofs.
What to do:
study the figures, diagrams, and illustrations carefully
check if graphs have proper labels and error bars
note which referenced papers you should read next
make notes in the margins or jot down key points
What you should achieve:
After this pass, you should be able to explain the paper's main argument and supporting evidence to a colleague.
You understand the content, but not every technical detail.
When you might struggle:
Sometimes even after the second pass, a paper remains unclear.
This happens when:
topic is completely new to you with unfamiliar terms
authors use techniques you have not learned
paper is poorly written with weak arguments
you are tired (yes, timing matters)
Your three options:
set the paper aside (you might not need it for your work)
come back later after reading background material
push through to the third pass
[Third Pass] 4-5 Hour Reconstruction
This pass is for papers you need to fully understand.
Maybe you are reviewing it, implementing the approach, or building on the research.
Try to recreate the work yourself using the same assumptions as the authors.
This helps you spot both the innovations and the hidden problems.
What to do:
question every assumption in every statement
think about how you would present each idea differently
note ideas for future work
identify gaps in citations or methodology
What you gain:
By the end, you should reconstruct the entire paper from memory.
You will know its strong points and weak points.
You will see implicit assumptions, missing references, and potential issues with their experiments or analysis.
For experienced readers, this takes about an hour.
For beginners, expect four to five hours.
This is normal.

Final Thoughts
Most people read papers incorrectly.
They start at the beginning and try to understand everything linearly.
They get stuck on details in the introduction and never reach the important contributions.
The three-pass method treats paper reading like the skill it is.
You build understanding in layers.
You make conscious decisions about how deep to go.
Start using this method today.
Your next research paper will take less time, and you will understand it better.
I am adding a new research paper every 3 days in the 75 Hard AI Engineering Challenge(75-day FREE guide for AI).
Before you go.
Want to work together? Here’s How I Can Help You
AI Engineering Product, Services & Consulting (B2B) at Dextar [Request a Brainstorm]
AI Engineering Live Bootcamp [AI Engineer HQ/Individuals]
AI Training for Enterprise Teams [MasterDexter/Teams]
I use [BeeHiiv] to send this newsletter.
I am hosting 8 FREE Workshop and Webinar events this November on AI Agents.
If you want to learn anything specific, reply to this email.

Reply