Throw away all your academic skills, the only thing you need is Obsidian + CC/Codex / similar tools

You could say 99% of current academic skills are garbage.

They leave nothing but a fleeting sense of novelty.

1. What can this combo do?

You throw your literature into a folder, tell the AI to archive: it will execute two steps.

Step 1 — Analyze first: Read all files, output the knowledge type / links to existing pages / contradictions / suggested subject folders for each paper. Do not write any files, new folders require user confirmation.

Step 2 — Execute: Classify → Deduplicate (FATAL-006) → Move files → Create research pages → Update topic pages → Update toolbox → Update index + log → Update overview → Report.

What you get in the end: a wiki page for each paper + an academic methodology toolbox.

2. How to install:

Connect your Obsidian folder with Codex, then feed the md files from the repo below into your Codex:

https://github.com/NoNightWatch/Obsidian-Codex-CC-for-academy

Let it follow the file content to set up.

(It’s best to create a separate vault for academic content)

3. References:

https://gist.github.com/karpathy/442a6bf555914893e9891c11519de94f

And

AI’s strictest father @dashen_wang · Apr 28 Article Deconstruction & Reorganization: The underlying operating system of AI civilization No one ever told you this, but it’s always been happening. Every leap forward for humanity wasn’t because someone discovered something new. It was because someone took old things apart and put them back together in a way no one had used before. This is the underlying move of civilization. It always has been. …13184422K

4. Installation is done, now let’s talk about usage:

Drop your literature into raw/inbox/, tell it “archive inbox and update the academic toolbox”.

It automatically classifies, moves files, generates wiki pages, creates cross-references, updates indices and logs.

When a paper comes in, it may simultaneously update a dozen related pages.

5. What else can you do?

Relying on this academic vault, you can ask questions restricted to the articles within it; after all, most articles are not open-access, and general AIs cannot give you relevant papers.

Based on your research protocol, it can recommend suitable methods and approximate experiment timelines.

Other uses? Wait for you to explore.

6. Tools are just tools, they don’t change quality

A knowledge base and a bookmark folder are pretty much the same — it’s just been empowered by AI, a bit automated, able to answer your questions.

But if your articles are of poor quality, if you just dump anything in without principles, or if you can’t tell good from bad, then in the end it’s just a fancy bookmark folder.

So selecting good articles and improving your database quality matters.

7. First master the fundamentals, then read the literature

The most important thing in a field is first the definition:

For example, you study Topic A.

Then, when was A first proposed by xx1? How did xx1 define A? What was the basis for xx1’s definition of A in the paper? What did A evolve from?

After xx1, did the definition of A evolve? Was it questioned or supplemented? Did xx2 propose A1 (new A) definition?

Does A have ambiguity? Is the definition named A actually divided into versions like A1, A2, A3? What exactly are A1, A2, A3? Which of A1, A2, A3 can describe A more comprehensively and objectively? Why?

Is it possible that after xx1 proposed the definition of A, xx1 themselves refuted A and proposed A1, but because A was so famous it became a paradigm, even though A is actually less objective and correct than A1? Are there recent papers that confuse A and A1? Or after A1 appeared, they didn’t keep up with the theory and still discussed deficiencies in A, but those deficiencies were already fixed in A1? Or do they think A is the latest definition, ignoring or never learning about A1?

Next is method:

Still using Topic A as an example:

Method a1: When was it proposed, and by whom?

In its method definition, what does a1 measure? What is actually observed during operation (a2)? Does a1 directly equal a2? If not, where do the errors come from?

If a2 needs to be measured by equipment, then there is an a3. Does a3 equal a2? If not, where do the errors come from?

Finally, do a1, a2, a3 refer to the same object?

Method a2: Same as above, no repetition.

Then methodology:

Still using Topic A as an example:

If for the definition of A, there are methods a1 and a2, Which method better reflects / is closer to A? Which method’s observed object is closer to A? What are the differences, pros and cons? Under what circumstances is it better to use which? Are there restrictions on conditions, materials, samples, etc.?

If for A, there are definitions A1, A2 and methods a1, a2: Which method corresponds to which definition? Which is better? Or depending on the sub-definition A1, A2, should different methods be used?

During the use of methods a1 and a2, are any actual conditions, interferences, or overly idealistic assumptions neglected?

Finally, under what circumstances should which method be used? Why? Which is optimal? Which has limitations? Where do errors come from? etc.

Finally, independent and dependent variables:

Still using Topic A as an example:

What changes in factors affect A?

Does A’s change affect other factors?

Others, adjust according to your own discipline~

8. Once the fundamentals are solid, you can start picking literature

Piles of data ≠ good paper; messy logic ≠ good paper; weak basic theory ≠ good paper.

So good papers are rare — even in top journals or their sub-journals, you’ll still find logically messy or theoretically weak papers. That’s normal.

As for how to read papers, I personally haven’t paid much attention, but a hot recent discussion recommends:

https://web.stanford.edu/class/ee384m/Handouts/HowtoReadPaper.pdf

Summary at: https://x.com/paperpaper886/status/2066148829439861016?s=20

My personal order is roughly:

Title,

Abstract,

Introduction (focus on definitions and hypotheses),

Results (look at data, figures, here you must summarize your own conclusions),

Discussion (compare with your own conclusions: check whether results are discussed objectively, whether anything is glossed over, whether logic is sound, whether basic theory is solid, whether limitations are addressed, whether conclusions are overblown),

Methods (is it appropriate and reasonable?).

For reference only

After about 10–20 papers you’ll be able to tell good from bad.

Certain institutes, labs, universities are really bad in some directions — take note.

If an institution consistently puts out poor work (2–3 papers), you don’t need to read their papers in that direction anymore.

Of course, you’ll also find interesting gossip: If one paper suddenly is written well and cleverly, it’s probably because some poor sucker joined them.

After all, like father like son.

9. Miscellaneous

Sleep, eat, mental & physical health come first. You have to be yourself first, then you are your parents’ child, someone’s partner… and last of all a student in a lab.

Prioritize graduation. After graduation you have plenty of time to realize your ideals and academic ambitions.

If you’re unhappy, stop. Sunk cost doesn’t factor into major decisions.

Take care of yourself. Take care of yourself. Take care of yourself.

Go out and walk around, don’t stay cooped up in the lab all day.

Eat something sweet when you’re down; enjoy your holidays.

Writing style needs polishing, but you can also distill it from your literature vault — no worries there.

Some words mean the same thing but are used differently in different fields. AI will mix them up when generating; your literature vault can guide it and narrow the range.

For certain top journals, don’t blindly follow every single comment from a peer reviewer. What editors want to see is whether you can defend your point firmly while keeping the paper’s main argument intact, especially against an unreasonable reviewer. Don’t just do whatever the reviewer says and end up with a mishmash — that will also get you rejected. Honestly, there are plenty of stupid reviewers, but if there’s no one else, they have to assign idiots.

Don’t offload your work onto your students. Research is your own performance, your own business. Don’t tie your own research pressure to your students’ graduation. If you can’t do it, get out. Forcing students to come up with ideas, do experiments, and write papers is not your ability. A three-year master’s: one year of classes, one year of internships and job hunting plus thesis — what kind of research can you do in one year? Let the kids graduate in peace.

10. Afterword

All tools and methods serve your purpose.

If you just want to graduate, these tools should be enough. But if you want to do good research, you’ll have to put in some effort.

However, everyone has their own path to walk — whether in life, work, or career. Don’t envy others; walk your own path. Your victory will come. Don’t lose yourself.

Finally,

the original intention of writing this article is to help students complete their research more easily and graduate more easily.

It’s NOT so that some beastly colleague (university faculty) can throw this article at a student and say:

“Go read this article and give me a summary report next week.”

**If that’s how it ends:

Is the tumor on your shoulder only there to make you taller?**

Can’t even read the literature, but you call yourself a university faculty?