Instill Science — Grant Report #2

Project Update

Instill is a community-driven project aimed at helping early- and mid- career scientists advance their research careers.

Initially, we were excited about becoming a micro-publications and for highly technical topics, like procedures for performing specific calculations, assays, and setting up experiments. Going into last year, we heard from many early researchers, especially those from the Global South, that they wanted some help getting their manuscripts cleaned up for journal publication. For researchers, getting a paper published in a decent journal can set the trajectory for the rest of the careers, so we pivoted to the “Peer Preview” system to help more researchers get published in better journals.

In academic publishing, once a raw “manuscript” submitted for publication, other scientists recruited from the field volunteer by reading and giving feedback for the author (this is called “peer review”). After a few back and forth emails between the peer reviewer, author, and sometimes journal editor, the paper is either accepted for publication or rejected from publication.

Our “Peer Preview” system intended as a quick “temperature check” to see if a paper was ready to be submitted for the official peer review process, which can take weeks to months to publication. We partnered with a nascent phage journal (Mary Ann Liebert’s PHAGE journal), and paired our community of experienced researchers with early researchers to Peer Preview their papers. The interest for Peer Previews was overwhelming, but we quickly realized that looking through papers before publication wasn’t very helpful. Many of the manuscripts we saw had fundamental shortcomings in experimental design and thesis, which required more than just copy editing.

Towards the middle-end of last year, we realized that we’d have trouble scaling up the more narrow “Peer Preview” system, since many papers needed more discussion on the scientific approach and experimental design. We instead decided to take the best parts of our learnings and lay them on top of familiar community tools like Stack Overflow and dev.to. Essentially, we realized that the way forward was to merge our v1 and v2 concepts, and build an Academia-centric system that is heavily inspired by existing tools. Once this platform becomes technically more mature, we will re-create but generalize the “Peer Preview” system to cover all sorts of abstract and scientific reviews.

In the latest iteration of the platform, currently live at https://instill.xyz, we built a concept we call a “super-portable micro-community” where users are able to share research, pose questions, and discuss science. The platform currently mimics popular forum-like interactions like Hacker News, but we’re using this platform as a launchpad to prototype better research-specific user flows, based on our past and future learnings.

Progress on objectives

Last year, we continued launching and supporting more academic, community events. Overall, we spent a lot of our time and efforts building out prototypes for peer review and hosting academic events to get an idea of how to support peer review at scale. We built a system that received hundreds of abstracts to be reviewed by a small group of peer reviewers.

Figure 1. Abstract submission portal and Retool-based Abstract Assignment dashboard (abstracts not shown due to personally-identifiable information)

Our system, based on the Peer Preview system built on top of our Airtable and Retool interface we submitted in our previous progress report, was able to handle 600+ abstract submissions and poster uploads from researchers around the world for VoM (https://isvm.org), a virus conference in Portugal of almost a thousand attendees. The system allowed a small committee of senior scientists to quickly approve and assign the abstracts into various tracks of the conference. We also successfully hosted a large virtual conference series for VoM called “iVoM” (1100+ registrants) in which we experimented with various community interactions (to various degrees of success).

The Super-Portable Micro-Community Tool

We also built and deployed our “super-portable micro-community” forum tool, which is currently hosted on our homepage (https://instill.xyz).

After the Elon-Twitter acquisition, there has been an explosion of Twitter-like, forum-like, and blog-like projects like Post News, Postcard, t2.social (among others), and an eagerness to try these tools out. There’s a large contingency of the “Science Twitter” crowd that we’re aiming to tap into.

Inspired by the movement towards decentralized spaces like Mastodon, we decided our community system and its data should be easily managed by the owner of the community. It should be easy to moderate and manage, and use familiar tools. Unlike Mastodon, we don’t necessarily the contents to be decentralized. Instead, we need to make it easy for many groups to create their own “flavors” of communities where they can control what goes on in their own community (not unlike old-school forums and bulletin boards).

Fig 2. Version 3 of the Instill community. The community currently focuses on the bacteriophage research and phage therapy community; the screenshot on the left has been cropped to be shorter, as the main page is too long because pagination hasn’t been implemented yet

We made some early decisions that this tool should be very easy for anyone to clone and launch as their own, with their own backend. This meant that we didn’t want complicated or expensive setups like Docker. We wanted it to be embeddable in as many places as possible, including places like a Notion page or as a mere iframe. We also wanted users to have full user-friendly control over the data, which is why it uses Airtable as the data source.

Of course, this does mean we will run into scaling issues very quickly, if we were to get lots of users. The benefit of being Academia-focused is that we’ll probably never run into scaling issues (joking aside, we will definitely be working on data adapters for more scalable databases like Cloudflare D1, Pocket Base, Supabase, LiteFS but maybe also even less-scalable ones like Notion and Google Sheets).

Currently, our system is live, and we’re slowly onboarding more users while squashing bugs and redesigning flows. Our current community is specifically aimed at the phage community, but we are happy to add communities more as we’ve fleshed out most of our user interactions. We’ve implemented a few interactions like like creating a Post, a Link, a Q&A-style “vote for the answer” flow, poll flow. We’re now making it easy to add and change interaction styles through a configuration file, so the community can easily morph from looking like Twitter to something like Hacker News or Instagram.

Towards the end of the year, we get to build and test our workflows for a couple of academic conference events, and a “Biobank Working Group” initiative, where we will use Instill to aid in scientific discussions and get a couple of manuscripts published.

Key activities

In the past year, we’ve made strides in both academic community-building and in developing Instill out as a community platform:

1. Built and launched a completely “portable, customizable, embeddable, micro” community concept that takes after inspirations like Hacker News, Stack Overflow and dev.to. Here’s what we mean with our terms:
   1. Portable means that any team should easily and cheaply create and manage their own community space. This led to decisions to create an API-first community framework deployed as a serverless Sveltekit project on Vercel (as opposed to using Docker, AWS, and so on). We also built the UI on top of “data adapters” that lets us connect tools like Airtable as the database (and eventually support “real” databases like Postgres). This portability eventually means other should be able to create their own Instill instances, using their own Vercel and Airtable accounts, completely separately from the core project, and with full control of their instance and data. We want the experience to be similar to launching old phpBB forums and Discourse, but without the technical overhead. In order to keep the data small and portable, we decided to keep the schema flat and simple, borrowing ideas from Reddit’s “RDF triple” architecture. This makes it easy to port the database into document databases like CouchDB in the future, but also unlikely places like Notion or Google Sheets. The main table tracks “Posts” — any comment, reply, or blog post is a “Post”. The second table holds User accounts and profile data. Events like “likes”, votes, and reactions are stored in the third table, which allows to relate any number of Events to any Posts. Events can be quantified, so the plan is to sync our IOU and tipping events to a ledger system like Formance or TigerBeetle, so that eventually they can be resolved and paid using Rafiki / the Web Monetization layer.
   2. Customizable means that different instances should be allowed to set the “interaction styles” to the owner of the community’s liking. Interaction styles change the way users use the community — e.g. Instagram, Substack, Twitter, Postcard, and Post.news all have different “styles” and user interfaces but are at their core the same interactions and data model. Customizations should allow a community to combine the best of Substack and Twitter, to fit their needs. To address this design goal, we’ve applied an “API-first” principle where every user interaction is handled by an API, in which interactions are sent to the database using “data adapters”. This allows us to design and prototype interactions (like reactions, upvoting, polling, marking as answer etc.) through the API and Events table, rather than having to add custom tables and columns in the database.
   3. Embeddable means that the tool should easily be embeddable (and fully usable) in any web environment. Similar to Disqus, Instill should be embeddable as a comment section to any content, like other Sveltekit projects, static web pages, or even in Notion pages. (I tried to get iframe embedding working in dev.to and the Interledger community, but couldn’t get it to work, since it doesn’t seem to allow for iframes). Getting embeddability to work meant having to create a slightly unfamiliar account sign up and sign in pattern, and then designing every interaction element around that pattern. For example, because the app needs to operate in iframes, we can’t rely on local storage or cookie data to persist — so any forum interaction like posting a topic or comment needs to potentially prompt the user for account information. Every time a user “posts” anything to the API, it needs to re-verify the account. We’re hesitant to add complex caching, key-value or redis on top, as that would make the system more complicated to deploy.
   4. Micro means that we want to design around small communities first, and being ok with using tools that aren’t scalable. For example, we want anyone to be able to make the community and data their own. This usually means that the data can’t be that large, and consequentially the community can’t be that large (for now). We want users to be able to “bring your own database” — hence the use of Airtable to store all data, including account information and encrypted passwords: using Airtable lets users easily moderate their communities. Though in the future we’ll adapters for other tools like Google Sheets and Notion, but also other database projects like Pocket Base, D1, and Supabase, for now the idea is to be not scalable, but be able to easily support a small community of ~300 active users, which is the size of our Slack group. Few academic communities ever grow past 1,000 users, so we don’t need to address the scaling issue at the moment.
2. In terms of attracting users and fulfilling our mission of making research more equitable, we’ve:
   1. Worked to encourage users to post about their lab issues and small findings. This has proven challenging, since asking the community questions implies that “you are junior and inexperienced.” The tech field has somehow managed to overcome this barrier with Stack Overflow, but much of academia still runs on appearances and status games instead of getting results. To get around this, we’re collectively creating a Biobank Working Group to address issues that generally everyone in the field has, and eventually posting questions as the Working Group. This should hopefully make people less afraid of posting questions.
   2. Helped researchers improve their “Peer Preview” submissions by giving them tips on how to improve their papers and research. Peer Previews are what we called our system for peer reviewing preprints (drafts of scientific manuscripts before they’re submitted to an academic journal). This proved tougher than we thought, as the relationship quickly turns into a “mentor / mentee” relationships, which becomes hard at scale. Even with the systematic peer review interface, we couldn’t get to “automating” the peer reviews because each paper was so different. In the end, we thought it might better to shift focus toward answering questions about basic scientific approach, rather than fix the paper. If we can change the way the researchers think about the approach and framing, we can catch problems before they make it on paper.
3. On our Science Communication side, we’ve:
   1. Prototyped and built a system for “massive peer review,” which received 600+ abstracts and let a core group of scientists accept and place each abstract into various tracks of the Viruses of Microbes conference in Portugal. The system also generates a database directory of all posters, and creates an abstract book printout of each abstract. (Screenshots in the previous section)
   2. Successfully hosted a virtual events series with almost 1200 registrants (and 400~ active participants across events) (Screenshots in the previous section)

Communications and marketing

In order to gain trust and relevance within academia (specifically the phage therapy space), we’ve collaborated with academic societies on several fronts. For our partners, we’ve built science education tools that cover lab basics; we built a scientific abstract and poster peer review system for a major European virus conference with almost a thousand attendees and we launched a couple of online science communications series.

The overall purpose was to establish strong relationships, and build trust within the established inner circles, understand the user needs (of both young and experienced researchers, as well as publishers and academic societies) within our space, and indirect marketing, to get our users familiar with using our tools.

By launching science-focused events and tools with our partners, we can both get valuable insights, and also embed ourselves as the technical and community partner of these influential organizations.

Viruses of Microbes: iVoM Virtual Phage Conference Series

https://ivom.phage.directory — We were central to organizing an academic-centric conference, posting videos online, eliciting questions and generating responses from the audiences.

Viruses of Microbes Poster & Abstract Upload & Review System for Portugal conference

https://vom.phage.directory — The goal of this system was to explore ways to scale up Instill v2 in a way that allows a small staff to review a large number (though shorter, in abstract form) of scientific content. The system handled more than 600 individual submissions from around the world, and using a panel of judges filtered and selected submissions into various tracks for the Viruses of Microbes conference in Portugal

What’s next?

For this year, we are continuing to build tools, host events, and experimenting with building community within academic research. In the last few years, we built one-off prototypes in tools like Retool to quickly iterate and learn from our users. This year, we’ll integrate these learnings into the core Instill community platform.

Most of our efforts this year will go into expanding features on the platform, and using the platform to accomplish a few use cases. We’ll also be expanding Instill to support hosting educational material and videos, and hosting live event streams before, during, and after academic conferences (like a small conference-specific Twitter thread).

Biobank Working Group

We are working with a few research labs and academic societies around the world (ISVM, Phages for Global Health, Phage Australia) to create a working group to discuss and establish standards on building phage and bacterial biobanks, which are collections of phages and bacteria. These discussions will involve things like lab standard operating procedures, lab safety, data collection, shipping and handling, paperwork, etc.

We are planning to host moderated and guided asynchronous discussions on Instill throughout the year, bi-monthly virtual calls, and an in-person meetup at the 2023 Viruses of Microbes meeting in Tbilisi, Georgia. The meeting notes, agenda, virtual meeting videos, and other materials will all be posted within the Biobank Working Group page on Instill, and every post will receive DOIs and permalinks, and will be citable units for the standards papers that will published by the group. All of the posts will be publicly available. If all goes well, we aim to publish our first paper either in Q3 or Q4 of 2023.

Phages for Global Health Online

In 2021, we partnered with Phages for Global Health to build a micro site with a curriculum designed around teaching lab skills for working with bacteriophages (https://open.phage.directory). We spent last year rethinking how to reorient the curriculum and content to be more community-centric, including encouraging user-uploaded videos, benchmarking tasks (e.g. different groups around the world perform a lab task like a plaque assay on the same phage and bacteria, and post their results with the rest of the group). We hope to relaunch PGH Online this year, with our new curriculum, towards this new direction. We are also building out the platform to support “playlists” in order to group content and video posts around various topics.

Evergreen Phage Meeting 2023

In 2021 we also created and hosted Evergreen, a hybrid phage conference in Olympia, WA. We experimented with some community features, but eventually moved all conversations to Slack. This time around, we will use Instill as the basis for before- and after-conference conversations. We’ll explore setting up conference presentations to be released recorded before the conference, so we can maximize face-to-face meetings during the conference. Just like last year, conference posters and abstracts will be searchable and tied to attendees presentations, but this year we want to add things like poster voting, polls for activities, and “pop quizzes” for different talks during the event, to make it a bit more interactive for those tuning in from elsewhere.

Expanding Instill Features

While we’re keen to build out our partner sites, we’ll be adding more community interaction styles, like short-form text posts and threads (e.g. tweets and tweet threads), image-only posts (e.g. Instagram), event sequences (e.g. similar to what The Verge posts during conferences or while covering Apple releases).

An area we are still excited to build out the points-based rewarding and bounty system we’ve outlined previously. Since we want to be end-users of Rafiki, we’ve been waiting for the system to become more stable and easy to integrate with. To allow us to develop tipping interactions separately, we’ve explored a ledger system like Formance (or TigerBeetle, a recently-launched project within the Interledger ecosystem). By using a database-backed, points-based ledger, we can separately test out tipping features without direct integration into a payments platform like Rafiki — instead, we can then “resolve” the transactions by letting users “pay out” the final amount, by connecting their payment pointers using the Rafiki and Web Monetization framework. Using a database ledger allows us to freely prototype and design user flows around payments with fake points, without worrying about running afoul of any laws. Ideally, we end up partnering with more mature groups using Rafiki, so we can connect our ledger system with their implementation (hopefully they’ve figured out the legal side of things as well).

Finally, we might look into launch another instance for an eager (and potentially non-academic) community to try it out (if you’re reading this and interested, send me a DM through Forem!)

What community support would benefit your project?

As we’re steadily adding use cases within academia and science, we’re eager to expand outside of academia. Specifically, we want to collaborate more with ILP, Web Monetization, and Coil-adjacent groups — in other words, this community!

We think the uptake for exploring the Events system, Formance ledger and ILP and Rafiki use cases will initially be much stronger and apparent in an Interledger community. Once the product reaches more stability, we’d love to work with other groups that have ILP and Rafiki figured out, so we can explore prototyping tipping and bounty interactions. Onboarding and getting to prototyping will be much faster with a community already familiar with Interledger; especially one already well-versed with Rafiki.

Towards the end of this year, we hope to:

• Work with a couple Interledger communities familiar with Rafiki and that might have community needs and a “test net” set up for us to plug into (Snake Nation comes to mind!)
• Find other Interledger projects which might roll out small-scale forum, Substack, or comments-like interactions for their products or platforms
• Connect with Interledger projects interested in working within academic research biotech and deep tech, since we’ve gained a lot of insights into how to work with and deploy new projects in this space

Additional comments & relevant links/resources

Our core project website and main deployed instance for v3 can be found at: https://instill.xyz.

• Github repo for v3 is here (still poorly documented, sorry!): https://github.com/janzheng/instill-v3

Our public-facing prototypes and projects from last can be found at:

• iVoM, a virtual conference series for VoM: https://ivom.phage.directory/
• VoM Abstract Portal: https://vom.phage.directory
• Sadly, the peer review dashboards contain too much personally-identifiable information to fully share. We’ll consider to create a “sample deployment” at some point, if there’s interest

Our partnership projects we’re launching this year with Instill-integration can be found at:

• Evergreen Phage Meeting 2023: https://evergreen.phage.directory
• Phages for Global Health Online: https://open.phage.directory. Phages for Global Health, the partner organization, can be found here: https://phagesforglobalhealth.org
• The Biobank Working Group will be a collaboration of members between the International Society for Viruses and Microbes (isvm.org), Phages for Global Health, and various other academic institutions (the list is still fluid)
• Phage Australia Network* (phageaustralia.org) will host discussions, topics, and share data between member labs on Instill towards the end of the year.