I Built a Tool for My Girl. Now a Harvard Professor and Fortune 500 CEOs Use It.

An AI tool frustrated my doctor girl, so her frustration became my next project. Here’s how………….

She was pursuing postgraduate studies. Her days already involved a lot of reading, revision, and trying to connect what she was learning. NotebookLM had become part of that routine: a place to work through research papers, ask questions about difficult topics, and prepare for academic discussions.

Over time, she had built up more than 100 notebooks around papers, subjects, and topics she wanted to return to.

There was effort inside each one. A paper she had taken the time to understand. A topic she had worked through. Material she had collected because she knew she would need it again.

But on the home screen, that structure was difficult to see. Everything sat together. Opening a notebook brought it back towards the top, while other work slipped further down. Related topics could be separated by whatever she had opened most recently.

Returning to something meant remembering where she had put it, recognizing a title, or opening a notebook to check whether it was the right one.

What stayed with me was the repetition. She had already done the difficult part of collecting and studying the material. Yet the next session could begin with another small search for her own work.

Her syllabus already gave her enough to remember. The location of a notebook should not have needed its own revision.

I cared about that because I cared about the person dealing with it. Making this one thing easier felt like a useful way to show up for her ( Bare Minimum )

So I started looking more closely.

The studying was organized. The notebooks weren’t.

Inside a notebook, she could ask questions, work through explanations, and follow citations back to the sources. The difficulty appeared when she needed to move between topics or return to an earlier session.

Her work had a structure: subjects, research areas, material for revision, and things she wanted to revisit. The workspace gave her a growing list of notebooks without a useful way to express those relationships.

At a small scale, remembering where everything lived was manageable. Beyond a hundred notebooks, that memory became another thing she had to maintain.

I began thinking about the return journey. What happened when she came back to a topic after a week? Could related work stay together? Could the notebooks she needed regularly remain easy to reach?

That helped me define the problem more clearly: people needed a dependable way to return to research they had already invested in.

Before deciding what to build, I wanted to understand how far that problem extended.

Reddit gave me the confidence to build

I went into NotebookLM discussions on Reddit to understand how other people were managing their work.

I read about research spread across notebooks, repetitive imports, difficulty moving answers into other documents, and requests for more control over the workspace. As the research continued, I paid attention to the workarounds people described.

Someone had turned a large collection of emails into PDFs to get the material into NotebookLM. Others wanted to export an entire conversation instead of copying answers individually. Teachers wanted mathematical notation to remain readable when they shared material with students.

These examples showed how much effort surrounded the main research task. People were collecting, rearranging, copying, and preparing material so they could use the tool in the way their work required.

Reddit gave me confidence that there was a wider need worth building for. I could recognize parts of the same struggle I had seen in her workspace, expressed by people using NotebookLM for very different purposes.

For the first time, I could see a real chance of this becoming a product people would choose to use.

Alongside those discussions, I researched the market and compared approaches to organizing, importing, and reusing information. I looked at where people repeated actions, where a workflow became complicated, and which improvements could fit naturally into NotebookLM.

The research opened up plenty of possibilities. I still needed a clear starting point.

I kept coming back to her screen: more than a hundred notebooks, and no useful home for the ones that belonged together.

One problem. A focused first version.

The first version became FolderLLM, later renamed FolderLM.

I started with folders, colour-coding, pinned favourites, and moving notebooks into groups. Drag-and-drop and bulk movement made it practical to organize an existing collection, including the work that had already accumulated.

The goal was easy to picture. Open the workspace, recognize the subject or project, and get back to the right material.

I wanted the controls to feel familiar. Someone who already understood folders and favourites should be able to use the extension without learning a new system of organization.

That influenced the design as much as the feature selection. I worked towards a calm interface that fitted the NotebookLM experience, with clear actions and enough structure to make a large collection easier to scan.

The starting problem gave me a practical way to judge the first version: would this make it easier for her to come back tomorrow and continue where she had left off?

From an idea to a working extension

I took ownership of the product from research and prioritization through interaction design, implementation requirements, review, and launch.

I used AI-assisted development and defined the behaviour of each workflow before refining its interface. A feature needed a clear starting point, an understandable sequence of actions, and a result the user could recognize.

That became especially important as FolderLM grew beyond folders.

Take merging notebooks. A user selects related work, chooses a name, and starts the merge. The process continues in a NotebookLM tab, but the person may still be looking at the FolderLM dashboard.

I designed the experience around that whole journey: show that the merge is running, prevent repeated submissions, preserve the chosen name, and make the resulting notebook visible when it is ready.

The same attention went into everyday actions. Moving or selecting something should preserve your place in a long list. Changing the interface language should update the controls while keeping your own notebook titles and prompt content intact.

I also made decisions about what could wait. Advanced controls and additional organization options had to earn their place alongside the core workflow. Each release needed to remain understandable to someone installing the extension for the first time.

The original goal stayed useful throughout those decisions. I was trying to remove a repeated interruption from someone’s day. Every additional feature had to justify the attention it asked for.

Getting it into people’s hands

Reddit had helped me understand the demand. I also used it to market FolderLM and reach people already working with NotebookLM.

That meant thinking about how to explain the value before someone installed anything. I worked on the store title, description, localized copy, and screenshots, giving the main workflows a clear place in the story.

As the product expanded, I updated that explanation. Folders remained the starting point, while bulk import, merge workflows, and reusable prompts became part of what someone could use it for.

The question I wanted the listing to answer was simple: where would this make your work easier?

Eventually, other people began explaining FolderLM to their own audiences. It appeared in Korean and German LinkedIn posts, and in a YouTube tutorial.

There is something moving about a product reaching a person whose language or daily routine is different from yours, while the problem still makes sense to both of you.

The thing I had started because of one person was finding a place in other people’s work.

Then a Harvard professor recommended it

Lawrence Lessig, a professor at Harvard Law School, recommended the extension.

In the recommendation dated 25 January 2026, he wrote:

“FoldersLM is a brilliant complement to a brilliant product (NotebookLM). Grateful for the insight that made it possible.”

That meant a lot to me.

The word “complement” captured what I had been trying to build. People already valued NotebookLM. FolderLM could make an important part of working with it easier.

A problem that had first caught my attention in my girlfriend’s study routine had become something a Harvard professor appreciated.

That was a much bigger audience than the one I had imagined at the beginning.

Users started shaping the roadmap

As people organized more of their work with FolderLM, their expectations grew.

One request was especially revealing: they wanted their folders to follow them between computers. They had invested time in creating a useful structure and expected to find it when they returned on another device.

On 25 March 2026, I announced sync support in response to requests for that capability.

That request changed how I thought about the value of organization. A folder could hold part of someone’s routine: how they separated projects, what they needed to revisit, and where they expected to find unfinished work.

I continued gathering feedback through Reddit, store responses, and feature requests. I weighed improvements by how closely they supported the main workflow, how often the problem appeared, and what it would take to deliver a complete experience.

For future imports, I compared the scope of a user-provided file with the work required for an account integration. Permissions, implementation effort, and the usefulness of a simpler first version all affected the sequence.

That helped me decide which ideas belonged in the next release and which needed more investigation.

The roadmap became a continuing conversation with the people using the product.

The product kept evolving

As the workflow grew, FolderLM developed into a broader companion for research.

Folders, colour-coding, favourites, and bulk movement help organize notebooks. The dashboard gives that collection a more manageable structure.

Bulk import helps bring multiple sources into the work. Merge workflows help combine related material when research has spread across notebooks. Source tools make those growing collections easier to manage.

The prompt library lets people reuse instructions for recurring tasks. Full-screen mode gives reading and writing more space. Copying answers with citations helps carry source context into the next document. Multilingual support lets people use the interface in their preferred language.

Together, these capabilities support the journey from collecting material to doing something useful with it.

I have also been evaluating whole-chat exports, better handling of mathematical notation, bibliographic formats, and research-writing workflows for future versions.

Each possibility brings me back to the same decision: how much repeated effort would this remove for the person using it?

Then Google addressed the same need

Google AI later announced Collections in Gemini Notebook, giving users a way to group notebooks and include a notebook in more than one collection.

For me, that brought the importance of organization into focus again. It also raised a practical question about FolderLM’s future: where could it continue to add value as the platform improved?

I had to keep evaluating the work around research—preparing sources, combining material, reusing prompts, and carrying answers into the next task.

A product built around a changing platform needs that kind of attention. The problem you started with can change, and the reasons people keep using your product can change with it.

Building FolderLM taught me to keep revisiting those reasons. A useful first release creates an opportunity to earn the next visit.

From one person’s workspace to thousands

As of 29 September 2026, FolderLM’s Chrome Web Store listing displayed 40,000 users and a 4.2 out of 5 rating from 54 ratings.

Between 1 January and 29 September, the listing recorded 110,397 views—more than one lakh listing views. Visitors came from Spain, the United States, Taiwan, South Korea, Peru, and other countries.

Organic Search accounted for 29,877 listing sessions, or 32.5% of the total during that period.

Those figures gave me a way to see the reach of the work. The product had moved from a problem I could describe through one person’s routine to a public tool that thousands had chosen to install.

Getting there involved the full journey: understanding the need, researching the market, choosing a scope, building the experience, finding users, and continuing to make decisions after launch.

I’m proud of how far FolderLM has come. There is sadness in telling its story, too.

When I think about the beginning, I still remember her crowded workspace and the simple wish behind the first version: I wanted to make her day a little easier.

That memory has become bittersweet. The product has given me reasons to celebrate, but returning to its beginning also means returning to someone I once imagined a future with.

It is difficult to hold both feelings at once. I can be happy about what FolderLM became and still wish the relationship that started it had worked out.

Building something for someone can hold a lot of care. A relationship needs that care to continue beyond the thing you made. There is no point at which the effort is finished and everything simply takes care of itself.

Eventually, on 1 June 2026, we went our separate ways.

It’s been 121 days since then.

Apparently, shipping a useful feature doesn’t guarantee retention.

I’d built something to make her life a little easier. The relationship ended, but FolderLM kept finding people who needed it.

You have to keep listening, making an effort, and occasionally introducing something new. You can’t expect the first release to keep someone interested forever.

Here I’m not talking about product management.

Thanks for reading. I build products with passion. Hire me.