Heptabase August 2024 AMA

I hosted an AMA in Heptabase 's Discord community last week. This post includes my responses to the top upvoted questions, which cover the following topics:

Heptabase's position in the PKM market and its attitude toward AI.
Heptabase's attitude toward interoperability.
Heptabase's plans for team expansion.
Heptabase's plans for upcoming feature direction.
Heptabase's plans for diagramming.
Heptabase's plans for long-form writing.
Heptabase's plans for whiteboard collaboration.
Heptabase's daily use case.

Heptabase's position in the PKM market and its attitude toward AI.

There are many competitors in the crowded PKM space, several that offer similar functionality around infinite canvas. How are you differentiating yourself from them to justify a more expensive and premium solution? 39 upvotes

I'd love to know more about how you think about Heptabase as a general personal knowledge management tool. With the current gaps around quick capture and AI integration that other apps have been more focused on, is that something you see as a part of Heptabase's future? Or is it focused more on complex ideas and a different space in a user's workflow? 68 upvotes

About AI, could you talk more about hepabase and AI integration, like, what AI features do you think heptabase should have? what’s your plan about when will AI come to heptabase? how you balance privacy and integrate AI into notes app… 33 upvotes

Before discussing Heptabase's position and differentiation in the PKM space, I would like to emphasize one thing:

We're not trying to build the best PKM product (if such a thing even exists). We're trying to build the best product for people who need to work on complex projects, learn complex topics, and solve complex problems to acquire a deep understanding and conquer complexity. These use cases aren't limited to "personal knowledge management" by nature. Many complex projects are collective projects, and many complex topics are better learned through discussions and guidance from others. Heptabase will address these scenarios in the upcoming iterations.

Within this mindset, if you compare the current version of Heptabase with other PKM products, I believe the differentiation is in the way we put together different PKM features into a coherent design that’s optimized for an opinionated workflow to help you extract atomic knowledge from different knowledge sources and reconstruct it into your understanding of a complex topic. Since this workflow is opinionated, people who like it stick with it and find Heptabase 10x better than other alternatives, while people who don’t like it will go on to use other PKM products.

Regarding AI, we believe there are many ways we can integrate it into Heptabase in a way that can help you better conquer complexity. Here are some ideas that I’m personally excited about:

Using AI to help you extract insights from lengthy knowledge sources 10x faster.
Creating an AI professor with deep domain knowledge to work with you on a whiteboard and constantly provide feedback and ask questions that are relevant and useful.

There are definitely some common use cases of PKM where AI can be helpful, such as transcribing voice memos for quick capture, fixing grammar during writing, integrating AI into search, asking questions to your knowledge base, etc. These are more like low-hanging fruit that we might implement over time, but their priority will not be higher than the ones that actually give our users a better capability to deal with complexity.

Of course, if we ever introduce AI-related features, you should be able to decide what data you want to send to those AI models based on your level of trust in companies like OpenAI, Anthropic, and Google, who all claim they will not store your data or use it to train models in their terms of service.

Heptabase's attitude toward interoperability.

What is your plan for ensuring interoperability and data exchange between Heptabase and other applications that users use? Currently Heptabase is like a walled garden with little ability to interact with information in it with other apps. Is this how it’s intended to be long term, with your team building in-app functions to replace other apps or are the current interoperability limitations just a matter of development time? 61 upvotes

We will eventually support APIs so that other apps can better use the data of Heptabase, and vice versa. The real question is timing: it won’t happen this year, and probably not next year. Not because we don’t want to, but because our system is still constantly iterating, and releasing APIs too soon could lead to frequent breaking changes, high maintenance effort that will slow down our product development speed, and expose more attack surfaces to bad actors. These are the things that we don’t want to worry about at the current stage. Before we support APIs, we will implement integration with other apps on a case-by-case basis, depending on the number of requests we receive through our in-app support.

Regarding “building in-app functions to replace other apps,” I think it will happen no matter what, with or without interoperability limitations. As we stated in our vision article, the entire Heptabase system is designed on the premise to support future users to build and publish their own MetaApps on top of a shared ontological card system. To support future users in doing so, we need to first support our own engineers in shipping useful MetaApps to our users. The MetaApps that we’re shipping, as stated in our vision, will address the problems our users face in the five stages of the knowledge lifecycle (exploring, collecting, thinking, creating, sharing) and ensure that knowledge can be seamlessly passed from one stage to another.

To be more precise, in the long haul, the real problem we want to solve is not just “make data interoperable across MetaApps.” The real problem we want to solve is to unbundle data, computations, and view components from an app, where end users can easily create and publish useful computations and view components (with the help of AI) on top of Heptabase’s card database. MetaApp is purely a form of bundling computations and view components that are related to the same context.

Heptabase's plans for team expansion.

In the field of note-taking and knowledge management, competitors are moving quickly. Especially Tana. They seem to have a quite large team because I see many different people from the team replying to user’s questions. Do you have any plans to accelerate the implementation of new features by expanding the size of your team? If so, what is the plan? 26 upvotes

We recently expanded our team from 6 to 10 people. I didn’t expand the team because the competitors have a large team. I only expand the team when:

We actually need more people to support the current volume of users and the next phase of development.
We have enough recurring revenue to hire more people while maintaining profitability.

To be noted, one of the biggest mistakes a startup can make is scaling the team prematurely, which forces the founders to:

Focus on management and internal communication instead of product innovation and customer research.
Increase the speed of burning money leads to the situation of "we need to either raise VC money in X months or lay off employees; otherwise, we're out of business."

I focus more on NOT making such a mistake. And I couldn’t care less about how many people other competitors have. We have a vision, a product that’s constantly improving toward that vision, and a small team that has full control of the company’s direction. I will try as hard as I can to keep this state as long as possible.

Heptabase's plans for upcoming feature direction.

Could you think out loud what features (planned or ideas) that you see as key for improving heptabase further? Currently on the roadmap there are multiple strains catering for different users (PDFs, collaboration, mobile, AI), while more "basic" features (such as live query, property-filter, integrations, deep link, block embeds) are further down. What are the areas that you think Heptabase are currently lacking, and what are the main audience that will benefit the most from the direction heptabase is heading. 31 upvotes

There are two categories of features we’re planning:

Features that create new capabilities, make the product more powerful in helping users solve complex problems.
Features that improve usability, make the product easier and more delightful to use, and lower the friction users face.

Regarding capability-driven features, as stated previously, the principle of the product is “breaking complex things down to acquire deep understanding.” So what we’re going to build includes:

Annotation Layer: Breaking down knowledge from all types of media, such as audio, video, image, epub, etc.
Communication Layer: Breaking down long discussions into connected atomic threads.
Application Layer: Breaking down complex projects into atomic pieces of computational problems that can be solved by citizen programmers, citizen scientists, citizen mathematicians, AI, etc.

The main audience who will benefit the most from this direction is those who have a complex project/problem to work on or a complex topic to learn.

Regarding usability-driven features, we prioritize based on the requests we receive through our in-app support. Our focus in the upcoming months includes the mobile and iPad experience (share sheet, handwriting, quick capture, task app, PDF highlight), global search accuracy, and mindmap usability.

Regarding live query, property-filter, integrations, deep link, and block embeds, I don’t think these are considered more “basic” features. They are still catered to the needs of a specific group of users. For example, my mom is a Heptabase user and I don’t think she will ever need any of these features. I do think deep link and block embeds have higher priority than others, according to our data in the in-app support system. For deep link, my current thought is it’ll be implemented after multiple workspaces and collaboration. I think that’s a better time to formalize our link structure. For block embeds, we’ll implement it after we rebuild our in-house editor library. This is an internal project that we’ve been working on for a while and requires a lot of care.

Heptabase's plans for diagramming.

Any idea that where we can make diagram like draw.io as element in the Heptabase card? Or make it a new type in the vault? 22 upvotes

We’re pretty sure we’ll build a diagramming feature, but unsure about which routes of implementation we’re going to choose (diagramming on the whiteboard vs. having an editable image card for diagramming). This probably comes after we have implemented features related to image annotation.

Heptabase's plans for long-form writing.

I would like to know the plan for long-form writing support i.e. a single document that will be built from multiple cards e.g. manuscript, book (chapters), blog post, Youtube script etc. If this is part of Heptabase plan, how is Heptabase going to approach this and any time estimate i.e priority on the road map. 22 upvotes

I personally have a huge interest in solving this problem, as I spend a lot of time writing long documents and articles. For me, a good writing solution is not just about assembling multiple cards. It’s about helping users deepen their understanding of a topic through an iterative writing process that involves self-questioning, logical validation, concise explanation of ideas, and structured composition. I have some preliminary ideas on this, and I believe it’ll be prioritized after we solve most of the problems in the “quick capture” stage.

Heptabase's plans for whiteboard collaboration.

Could you expand more on your vision for the upcoming roadmap item Whiteboard: Collaboration and Discussion? I would love to know how far you plan to expand the collaborative capabilities of this app, especially given your pricing model. Will it be more focused on presentation style collaboration? Will it be host/guests? Will we see new types of cards that work in a similar vein to sticky notes or comments? 21 upvotes

The reason we’re building the Whiteboard: Collaboration and Discussion is the same as the reason we build other feature: we want to better serve our users who’s working on a complex project, which a lot of times involves a lot of discussions and feedback loops that result in a high volume of text responses that are hard to trace back and make use of in the future. We’re less interested in solving the problem of “presentation” and more interested in helping our users gradually build “structure” on top of all the discussions they have around a topic.

There will be two phases of implementation. The first phase is rather trivial, which is laying out the technical foundation of the permission model where you can invite other people to your whiteboard, collaboratively edit this whiteboard, let other people bring their own cards to your whiteboard, etc. The second phase is more important, which involves introducing the discussion feature that has a lot of interplay with the whiteboard.

Heptabase's daily use case.

I'm curious about how you use Heptabase throughout the day. For instance, do you utilize it for tasks such as project planning, brainstorming, reading, or conducting research for writing technical specifications? I'm particularly interested in how you use it beyond the wiki use case. 23 upvotes

I always start my day by deciding what I want to work on and writing down my to-dos in the Journal. For example, today I’m focusing on researching “application layer design.” My to-dos include the things I want to read, the questions I want to answer, etc. I’ll then go to the “Product” tab group and enter the pinned whiteboard called “Application Layer” to stare at it and think, add new content to existing cards, add new cards, add new connections, rearrange cards, break down cards, etc. During the process, I might clip some internet content or add some PDFs I found to the whiteboard, break it down into atomic cards, delete useless cards, rewrite the headers of useful cards in my own words, etc. That’s pretty much it.

Over the course of a few weeks, the whiteboard will go from empty to messy to structured, and then eventually I might write a document that synthesizes my research on this whiteboard and share this document with my team. When drafting this document, I always open it on the right sidebar of the whiteboard and start by adding relevant links from the whiteboard to the document (using the copy link feature), so I can click those links to navigate the whiteboard and replace those links with the words I write.