Towards a research community for better thinking tools

27 November 2021
27 Nov 2021
New York, NY
15 mins

As I’m thinking about defining more narrow focuses for my independent work next year, one area has stood out consistently as both personally exciting and more widely important: imagining and building better ways computers can help people do their best creative, thoughtful work, and in the process rethinking the relationship creative people have with the computer as a part of their work.

My projects and this blog have never had a clearly defined focus, but a big part of my project portfolio is tools I’ve built to help myself manage the information that flow through my life and work. This theme has emerged slowly but definitively in the last year with projects like personal search engines and a custom Twitter client. This year, two things have happened in my life to help me realize this is a field in need of more independent, dedicated, inventive research efforts.

First, there is an increasing availability of capital flowing into companies building on new ideas in this space. Startups like Coda, Remnote, Mem, and Notion are raising millions of dollars on first investment rounds, and they’re coaxing even the very large tech companies to invest seriously into more competitive tools for thinking and creating together. There has always been a die-hard group of thinkers who argued that the “computer revolution isn’t here yet” and that the best software creative tools were yet to come, but the new influx of capital validates financial interest and demand for tools that break new ground in a way that hasn’t been true in many years.

Second, I’ve personally invested more seriously in designing and building more ambitious projects, like a universal personal search engine and a web browser focused on knowledge work. In response, I’ve spoken with many other brilliant and creative people working on similar projects, both independently and as a part of early stage companies. These conversations have reiterated in my mind how much I enjoy working at this layer of tool building, designing the basic metaphors and building blocks out of which the rest of the world runs businesses, shares new ideas, and creates art.

As I reflect on these realizations, I’ve been wanting to invest more time and effort specifically into building these kinds of tools, and conducting more exploratory, open-ended research that can fuel new creative ideas about how to build better thinking tools. This also spurred some interesting conversations I’ve had with other folks in the field about what an ideal research community could look like. What kinds of people and companies would push the ecosystem in the right ways? How might these researchers and companies communicate new ideas with each other in a way that results in open, lasting progress?


What follows is a loosely structured collection of different building blocks from which I think we can build a good research community to push this space forward. Though I’ve spent the last couple of weeks thinking about this topic, there’s no doubt I’ve missed some important pieces. If I think of any other, I might come back and add to this list. A brief table of contents:

  1. Community and identity
  2. Communicating research
  3. Outward-looking problem discovery
  4. Proven models for sustainability
  5. Many small projects, building on each other

Community and identity

Great communities are made of the stories we tell about ourselves – why we do what we do, how we’ve done it, and from where those traditions came. These, alongside how we talk about ourselves and our work, form the identity of our field. I think it’s worth being conscious and deliberate about such an important facet of the community.

So far, the way we label our work (“tools for thought”, “knowledge graph”) and the way the world perceives this field have been defined mostly by the loudest and biggest companies to fill the room. This kind of ad-hoc identity is enough to bring people together, but these words we use today are laden with overloaded meanings and associations to existing products and ideas. Different corners of the community behind these tools also agglomerate around different focuses: Some really love to customize and polish their tools and workflows. Others are more interested in building on fundamental interaction and interface design research dating back to the dawn of the personal computing era. One name and identity cannot contain them all – I think a research community should recognize that, and gather around a more independent identity that can represent the way we want the community to treat its members, its history, its traditions, and its future.

If I had my way, I think this new identity should…

There are entire essays to be written on each of these points – probably soon to come.

There is already a loose emerging community around thinking- and creative-software research. The challenge is to bootstrap a more structured community from these seeds. By “community”, I emphatically do not mean some group chat or online forum where every member hangs out and post messages. That kind of a monolithic community approach does not scale, and does not last. I want a research community that feels decentralized, but runs in roughly the same direction, branching and merging where the ideas take us. I want there to be people I can run to with a new idea or a prototype and get honest, hard-hitting, thoughtful feedback, and that doesn’t require much shared infrastructure, only shared enthusiasm and identity.

Communicating research

There are emerging conventions in the community about how to communicate research, and I think they set good examples from which to build traditions: written communications and open-source prototypes.

Written communications

There are two broad categories of written research communication I can think of: reports on findings after specific projects, and periodical updates on progress. I think the best example of the former is Ink & Switch’s reports. My favorite examples of the latter are Andy Matuschak’s Patreon posts and Alexander Obenauer’s lab notes. Both feel important to a vibrant research community.

In-depth write-ups of research findings can serve as anchor points and canonical descriptions of important ideas. There is value in giving names to new ideas and describing them in detail for others to cite and build upon. Reports of this type can serve that role. Well-written research reports also clearly lay out findings that future research could extend, and approaches and experiments that didn’t work. In a productive research community, research builds on past research by referencing and working from findings shared in these in-depth reports.

By contrast, periodical updates can fill the time gaps in between long research cycles and form a kind of asynchronous “group chat” for the community. Often the most interesting parts of research work are the parts that are fuzzy and vague and incomplete in our own minds, and hardest to articulate properly at first. These ideas don’t make it to “official” publications or reports because they’re not ready to be cast into form yet, but speaking about these more soft ideas with like-minded people can lead to the clarity we need to move forward on them. I think periodical updates from researchers and builders in this space can become a place to share those more fuzzy ideas. These communications might also be an effective “heartbeat” for the community, to keep a sense of loosely coordinated momentum.

I also want to note a third kind of communication – pieces like Up and Down the Ladder of Abstraction and Using Artificial Intelligence to Augment Human Intelligence – that push the field forward on what effective research communication can look like. As people working to expand the way we communicate and share ideas, I think it’s worth investing into the ways we communicate our own ideas between ourselves to make it more interactive, taking full advantage of the software medium we seek to embrace.

Open-source, working prototypes

One of my favorite things about the way this field operates is that often research yields open-source prototypes and working code. Szymon Kaliski’s projects don’t come with a lot of writing but self-demonstrate ideas because they’re working code on GitHub. It seems the culture of this ad-hoc community we have so far values open source software, pushing some companies to build products on a platform of being open-source. I’ve also embraced this culture; almost all of my own projects are free and open source. A community culture of open-source-by-default could also enable more ratcheting progress in a field increasingly tucked behind startup NDAs.

Outward-looking problem discovery

One of my greatest concerns about this field today is that almost all of the problem discovery happens by a kind of self-interested navel-gazing process, where product builders take the quote “build things you would want to use” a little too literally, and build products for the small niche group of people interested in note-taking tools and processes. This leads to products that seem useful to a small group of other people who are also working in this space and familiar with its vernacular and concepts, but are unusable or unapproachable by most people outside of that small community. I think this is a dangerous failure mode.

Good industrial research can only happen in problem rich environments, where research questions are anchored to problems found in real workflows used by domain experts working in their respective fields to solve problems other than the challenge of building more thinking tools. This kind of work requires that the research community work in close collaboration with people who work outside of our field. This is a process that requires active effort – radiologists and human rights lawyers and artists and journalists are not going to seek out risky, new ideas in their knowledge workflows as a part of their daily work. I think active outward-looking problem discovery, where we dedicate research time to consult with other domain experts to seek out new questions to pursue, is critical to an impactful research community.

Concretely, this can happen in a few different ways. Bell Labs’s problem discovery was performed by the entire rest of its parent company, AT&T, as it tried to span telephone wires across continents and oceans and anticipate technical problems 5-20 years in the future. Companies like Retool building business-facing products have field-deployed engineers working with customers to discover new uses cases for their products and ensure the rest of the company is directed forward, not just inward. Independent researchers like Andy Matuschak augment their own personal experience with tools they build by testing rigorously with other end users in a variety of professions.

I’ve spent a lot of time over the years desperately trying to think of a “thing” to change the world. I now know why the search was fruitless – things don’t change the world. People change the world by using things. The focus must be on the “using”, not the “thing”. Now that I’m looking through the right end of the binoculars, I can see a lot more clearly, and there are projects and possibilities that genuinely interest me deeply. — Bret Victor

Tools don’t change the world alone; people do, in the way they use tools. And the way to build the right tools for those people is to focus on the people and their work first.

Proven models for sustainability

As I think about shifting more of my own work into research, one of my own big questions concerns financial sustainability. For talented, experienced people to consider contributing to a more open-ended research community rather than going to work at any of the hot startups of their time, there must exist well-established models for making a fair income doing such work. I don’t think it’s necessary for researchers to make salaries competitive with top-of-the-line compensation from large tech companies, but pursuing research must not be a financial gamble, the way it is today for most people.

Some people have found success with a crowd-funded Patreon-kind of funding model. Even though ostensibly making is showbusiness now, I don’t think this is the proven revenue stream we want everyone to pursue. Not everyone wants to turn their online presence into pristine, well-curated identities about their professional interests. Even for those who can manage taking on a creator identity, I think a crowd-funded patron kind of model can lead to undesirable power dynamics where researchers may be pressured to pursue questions that satisfy mass-market curiosities most, rather than following their expert intuitions.

So, what can we do?

For a healthy and impactful research environment in the thinking tools space, I think we need a mix of “concept car” projects (a phrase I lovingly borrow from Jess Martin) and “production-grade” tools (a phrase I’m adopting from Ink & Switch). Concept car projects explore the boundaries of current technologies or showcase what new designs and ideas enable. They are necessary to push the field forward, but usually too rough or incomplete for the rest of the world to depend on. Production-grade tools are tools that are battle-tested to be secure, reliable, intuitive, and polished enough to be load-bearing components of real-world workflows.

These two kinds of work likely need different models for financial sustainability.

Research fueling products, products motivating research

The industrial research lab Ink & Switch offers an interesting precedent for a research group bringing products and production-grade tools to market, like Muse and Automerge. Though the lab itself is a nonprofit focused on exploratory research, some of their work leads to production-ready products that the lab can spin out into profit-generating companies. This approach, of research labs becoming sustainable by bringing products to market, resonates with me. Industrial research sets itself apart from academic research by the fact that industrial research seeks to answer questions relevant for building new products and companies. This was the explicit framing of research work done by Bell Labs and Xerox PARC, and the framing that labs like Ink & Switch also seem to be following. From this perspective, it makes sense for labs to become sustainable through research that leads to commercially successful products.

If research fuels creation of new products, I think building products can in turn fuel further research by being active areas of problem discovery. Much of the difficulty of building new products isn’t in coming up with the initial idea or insight, but in the thousand different engineering and design refinements that need to be made before a prototype can be turned into something the average customer will be able to use to solve their problems. These range from technical challenges like building a high quality rich text editor to design problems like balancing UI complexity with customizability. Once these products enter the real world, new problems always emerge at the point of contact between real-world use and research findings. These new problems can then feed back into research, and the cycle can continue.

The downside of taking this approach is that there’s a constant need for start-up capital in the beginning, when researchers and labs won’t have products to sell. Grants and corporate research programs may fill that role.

Financial support and research grants

In a world where labs become sustainable by spinning out products, researchers need some way to de-risk their initial work, when they won’t have any new products or technologies to sell. I think this is an effective place for open-ended research grant programs.

We can frame the role of research grants in a few different compatible ways:

I’ve also observed more startups in the field opening up “Researcher in residence” programs. These roles can be another way for new people to step into research work without associated risks or lack of structure, but I think we should avoid a world where most research about these tools are done in-house by companies. Corporate research ultimately results in proprietary intellectual property that is harder for a community to build on, and it puts emphasis on companies and their products rather than individuals and their learnings. I think corporate research programs should follow the conventions of the community, and be careful not to shadow individual contributors' efforts.

Many small projects, building on each other

I think smaller projects that are faster to build are better for research in this space. Building many smaller projects rather than large ambitious ones have helped me because I avoid getting too attached to one particular idea or product, and with smaller-scoped prototypes I can try many more iterations against the same question or problem. It also lowers the barrier to entry to try more risky ideas – “I’ll try this for a weekend” is much easier than “I’ll have to shift my schedule the next couple weeks to fit this in; is it worth that?” A culture of shorter, more atomic projects will also encourage everyone to break down large ideas into smaller ones that are individually testable, which I think is a good practice regardless of whether those ideas are for a product or an experiment. On the other hand, cycles that are too short obviously run the risk of keeping us from trying more ambitious or complex ideas.

My gut feeling is that three-month “cycles” focused on specific research questions strike an ideal balance. Any longer, and we might find ourselves not chasing after a concrete enough question; any shorter feels too short to really dive deep into a problem and try as many iterations as might be necessary to find good answers. If we can establish a culture in the research community of 3-month cycles pursuing a single question, it might also be a good foundation on which to build timelines for things like research grant programs, community conferences, and collaborative projects.


Thanks to Karina Nguyen and Theo Bleier who offered feedback on past versions of this post, as well as the many independent researchers and writers whose opinions and work appear above.


AI as a creative collaborator

Notational intelligence

I share new posts on my newsletter. If you liked this one, you should consider joining the list.

Have a comment or response? You can email me.