How Recommendations Get Smarter Over Time

The first time you open ZiNote and start swiping through papers, you might think: "Some of these are spot on, but others feel a bit off." That is completely normal. In fact, it is exactly how the system is supposed to work on day one.

ZiNote is designed to learn from you. Every paper you swipe right on, every paper you skip — each of those tiny decisions teaches the system a little more about what you care about. Over the course of a few days, your feed transforms from a broad set of keyword matches into a finely tuned stream of papers that feel like they were picked just for you.

This guide explains how that learning process works, what you can do to speed it up, and how to steer your recommendations when your interests shift.

Every Swipe Is a Signal

At its core, ZiNote pays attention to two things: what you save and what you skip.

Swipe right (save) tells the system: "I am interested in papers like this. Show me more along these lines." The system takes note of the topics, themes, and styles of papers you save and starts looking for similar ones.
Swipe left (skip) tells the system: "This is not what I am looking for right now." It is not a harsh rejection — it simply helps the system understand where the boundaries of your interests lie.

The beauty of this approach is that you never have to fill out a preferences form or rate anything on a scale. Your natural reading behavior is all the input the system needs. Just swipe the way you normally would — quickly past things that do not grab you, and save the ones that do.

Over time, the system builds a richer and richer picture of your research taste. It remembers every single decision you have made, and it continuously refines what it shows you next. The result is a feed that feels less like a search engine and more like a knowledgeable colleague who knows exactly which papers to drop on your desk.

Getting the Best Start: Cold Start Tips

The first few sessions matter more than you might think. Here is how to give ZiNote the strongest possible foundation to learn from.

Swipe Generously on Day One

Your first session is the most important one. Try to swipe through at least 20 to 30 papers in your first sitting. It does not need to take long — most papers only need a glance at the title and abstract before you know whether they are relevant.

Why does volume matter early on? Think of it like this: if you only swipe through five papers, the system has five data points to work with. That is not much to go on. But 25 or 30 swipes? Now it has a real pattern to detect. The more you give it in the beginning, the faster it catches on.

Do not overthink each swipe. Go with your gut. If a title looks interesting, swipe right. If it does not, swipe left. Speed and honesty are more valuable than deliberation.

Choose Precise Keywords

Your initial keywords set the stage for everything that follows. The more specific your keywords are, the better the starting pool of papers will be — and the less noise the system has to sort through.

Too broad: "machine learning" — this will pull in thousands of loosely related papers
Better: "federated learning privacy" — this narrows the field to a specific intersection
Best: "differential privacy federated learning healthcare" — this gives the system a very clear starting point

You can always adjust keywords later, so do not stress about getting them perfect. But a thoughtful first set of keywords means your day-one experience will already feel more relevant.

What to Expect: A Timeline of Improvement

Recommendations do not flip from generic to perfect overnight. The improvement is gradual, but you will notice it.

Day 1: Keyword-Based Discovery

Your feed is primarily driven by the keywords you set. Papers match your topics, but the selection is broad. You will see a mix of highly relevant papers and some that are only loosely connected. This is the system casting a wide net, and your swipes are teaching it where to tighten things up.

Day 3: Adapting to Your Direction

By now, the system has started to pick up on patterns. If you have been consistently saving papers about a certain methodology or from a particular subfield, you will notice those themes showing up more often. The obvious mismatches start to fade. Your feed begins to feel like it "gets" you.

Day 7: Precision and Surprise

This is where things get exciting. After a week of regular swiping, your feed is noticeably sharper. Not only are the recommendations more precise, but you will also start seeing surprise discoveries — papers you would never have found on your own but that turn out to be exactly what you needed. These cross-pollination moments are one of the most valuable things ZiNote offers, and they only happen once the system truly understands your interests.

The improvement does not stop at day seven, of course. The system keeps learning for as long as you keep using it. But most users report that the first week is when the biggest leap in quality happens.

How to Correct Your Recommendations

Sometimes your research direction shifts, or you realize the system has picked up on a pattern you did not intend. Here is how to steer things back on course.

If Recommendations Are Drifting

If your feed starts showing papers that feel off-topic, revisit your keywords. Remove or refine any keywords that no longer match your current focus. The system will pick up on the change quickly and adjust what it surfaces.

You do not need to start over. Just a small keyword adjustment is enough to nudge the system in a new direction while preserving everything it has already learned about your broader preferences.

If You Want to Explore New Territory

Curious about a new subfield? Add a new keyword. The system will start pulling in papers from that area, and your swipes will teach it how deeply you want to go. This is a great way to dip your toes into adjacent research areas without committing to a full literature review.

Adding new keywords does not erase your existing preferences — the system learns to balance your established interests with the new ones. Over a few sessions, it figures out how much weight to give each topic.

When You Run Out of Papers: Smart Keyword Suggestions

Here is something that catches new users by surprise: sometimes you will swipe through all the papers for a given keyword. When that happens, ZiNote does not leave you with an empty feed.

Instead, the system intelligently suggests new keywords based on what it has learned about your interests. These suggestions are designed to expand your research horizons in directions you might not have considered — adjacent topics, emerging subfields, or methodological intersections that align with your saved papers.

You are always in control. You can accept a suggested keyword, modify it, or ignore it entirely. But many users find that these suggestions lead them to some of their best discoveries. It is like having a research advisor who says, "Based on what you have been reading, you might also want to look into this."

The Bottom Line

ZiNote is built around a simple idea: the more you use it, the better it gets. Your swipes are not just saving or skipping papers — they are actively shaping a recommendation system that becomes your personalized research engine.

The key takeaways:

Day one will not be perfect, and that is by design. Swipe through 20 to 30 papers to give the system a strong start.
Be specific with keywords to get better results from the beginning.
Trust the process. By day seven, you will notice a real difference.
Adjust keywords anytime to correct or expand your recommendations.
Follow the smart suggestions when keywords run dry — they often lead to your best finds.

Ready to start training your feed? Open ZiNote, set your first keywords, and start swiping. Your future self — the one with a perfectly curated research library — will thank you.