Project summary

Acquaintances are the biggest source of new opportunities (Rajkumar et al., 2022), at least when it comes to job mobility, but making the mental connection of who would be a good fit depends on recall from memory - with little follow-through because no one is paid to do it.

And even a Superconnector can only connect people they know, while reposting is highly ineffective (Goel et al., 2016).

Together this makes for poor throughput of information.

Enter the TrustWeb:

This lightweight infrastructure aims to make it easy to record and access information, especially from and about acquaintances, with the people we trust. We can query the web of close friends; friends; and acquaintances, according to what information we have entrusted with whom.

Try our demo

What are this project's goals?

North star: every Effective Altruist can reach the opportunities, talent and resources that those who have met them want them to see - based on the degree of trust they have built with them and without anyone needing to be a full-time Superconnector.

At EAGxBerlin (Nov 2026), a beta of the open-source TrustWeb app reaches >30 attendees, who share actionable info in addition to contact details.

By EA Global London (May 2027), a second iteration is ready, and we publish the findings so any community can reuse the approach.

How will you achieve them?

We build on the prototype we already have, release a beta for EAGxBerlin in November and iterate on it for EA Global London next year in May.

The app lets each tier of a person's network (close friends; friends; acquaintances) access a more comprehensive part of the items each person has explicitly chosen to make visible within a given trust tier.

Example A: an acquaintance of mine who is looking for a job asks their network and learns that a friend of mine, whom they don't know, has just recorded in the app that they are looking to fill a new position - months after I have met either of them.

Example B: a friend working on an animal welfare project asks their network and learns that an acquaintance of my close friend has just opened an animal welfare fund. (Where my close friend trusted me to share information about their network with the people I regard as friends.)

This information will be populated primarily through notes taken after meeting people:

For example, after I meet someone for the first time, I record notes about them in the app and mark them as an acquaintance.

Or after my next conversation with a close friend, I mark them as a close friend and let them see whom I met at an event and what notes I took about them - and if for some reason they were to break my trust, I could just stop giving them this visibility.

Each tier of connection has a different kind of access to our notes:

1. close friends can see everything that we see

2. non-close friends can query, but whoever holds the information reviews the request - a decline looks exactly like "nothing found".

3. And acquaintances can only query for information that is deliberately reserved to be shared with this trust tier, such as that someone is “open to research roles”.

These sharing permissions can be changed or revoked as relationships and circumstances change.

The transparent Swiss LLM Apertus (hosted on Infomaniak's Swiss servers) will per default assist the note-taker by flagging when sensitive content - reserved for close friends or not to be shared at all - would be exposed to acquaintances. During the Berlin pilot, participants will be asked to make information about another person queryable only when they reasonably believe that person would want it shared in a warm introduction. Thus, ensuring that data is safe with those that we entrust it with.

The habit it replaces: at the end of a conversation, instead of scanning a Signal QR code that opens an empty chat, you show a TrustWeb QR code that carries your notes, in addition to your Signal link.

The other person wouldn't be forced to install the app to read the contact details and any mutually shared follow-up note from the conversation or to stay in touch, since they have your contact details. They would only need the app if they wanted to query the network of someone else - according to the trust tier that they have allowed them to see.

In this way, even with a single user this app will be valuable as a personal & visual CRM - that can be integrated with one's other apps through API or MCP - even before it becomes truly powerful for querying opportunities, talent and resources across one's extended network.

Importantly, the querying would work already with only a few 'weavers' using the app actively and giving each other and their close contacts, access to querying the information that they record.

And to make querying a strong habit, the user can choose to see other content that is relevant to their query - either what the community-organizers have curated, such as EA material, or search-engine results. Through seeing public content in addition to the query results, thus guaranteeing a 'reward loop' of always finding results to the query, even if their trust network is still growing.

TrustWeb intends to work alongside Swapcard. During the conference, attendees keep using Swapcard to book meetings, which the TrustWeb-Recorder can take notes about. Additionally, Swapcard content can be exported to fill in notes about the user's network. With the organisers' agreement, we will inform attendees of the possibility to use TrustWeb alongside Swapcard.

Security and standards before the EAGxBerlin pilot: We will run a rudimentary audit - a written threat model, a self-assessment against the OWASP ASVS Level 1, and dependency scanning. Until the stable release has passed a full independent review, the pilot is limited to low-stakes offers and needs (jobs, intros, events), not money or housing. We build on open and/or established standards rather than our own cryptography: W3C Decentralized Identifiers (DIDs) with a fresh identifier per connection, QR-code, NFC and Airdrop for the in-person connection, DIDs for connection-specific identifiers; DIDComm for secure messaging between devices; and OpenVTC-compatible trust-graph concepts for representing and querying relationship context, and WebRTC for direct, encrypted device-to-device connections, falling back to an encrypted relay.

Interoperable by design: Notes are stored as plain bullet-point markdown files on each person's own device, so they can be exported to any note-taking system and read by other tools through MCP. People connect using OpenVTC as open standards, so other apps that follow the same standards can join the same network.

The key mistake of the past has been that users were able to create fake accounts and try to cheat including by Sybil attacks (one person running many fake accounts that promote each other to gain influence). The TrustWeb can mitigate these threats in this manner:

Protection against fake accounts

• Every connection comes from meeting someone in person, so meeting in person is how each user confirms that the other accounts in their network belong to real people, which NFC/Airdrop and dynamic QR codes make much more likely.

• In addition, there is no vouching or rating in the default TrustWeb, which reduces the utility of Sybil attacks.

• Finally, if an account seems off, you can press a button to request the accounts of people you actually know to verify how you met the other account and if they have flagged them.

Adjacent projects have intended to build a complete alternative internet on this kind of infrastructure and although, we also believe that this is possible and desirable, we will focus on these two steps for now: making a tool that is valuable even if used just by one person and then making the querying useful with a few dozen users - we want to stay grounded and form the strong foundational momentum to let us become Superconnectors amplifying the power of weak ties.

What already exists

Main Demo

connecting in person with notes, adding to the inventory, querying interface

All Interactive demos

Multiple iterations with different backends: no server, encrypted relay, direct device-to-device, automatic fallback etc. Focusing on an early version of the consent flow.

Key repo on GitHub

One version of the TypeScript codebase in alpha

How will this funding be used?

Minimum viable pilot: USD 8,500.

If the minimum amount is funded by 11 October 2026, we will use the following month to take TrustWeb from its current demo to a beta ready for EAGxBerlin, 13–15 November. This includes

• completing the inventory recording and sharing-the-app experience

• pre-pilot security and privacy work, and

• alpha testing

Full project launch: USD 50,000.

If the full amount is funded, we will continue from the EAGxBerlin pilot through EA Global London in May 2027 with:

• analysing the initial findings

• iterating on the product with participants

• conducting a stable-release security review (by Danube Tech)

• and publishing the open-source code and pilot findings for reuse by other communities.

Funding between these levels extends the work in priority order:

1. first pilot security, privacy, and testing;

2. then additional implementation and iteration;

3. then the independent review and dissemination work.

Budget Line Items

Minimum: one month, to EAGxBerlin (USD 8,500)

• Stipend for Jakob (build, pilot, write-up): USD 2,500

• Rewards for paid advisors Day Waterbury, Zach Miltz and Andrew Hasse (launch retainer of USD 1,000 each): USD 3,000

• Paid alpha testers (5 people at USD 50 each): USD 250

• AI costs (Infomaniak-hosted models by default, Claude as fallback): USD 1,000

• Running costs (Hosting & subscriptions such as Tana): USD 250

• EAGxBerlin travel, ticket and accommodation: USD 700

• Buffer (~10%): USD 800

Full: until EA Global London (USD 50,000)

• Stipend for Jakob (build, pilot, iterate, write-up; 7 months at USD 2,500): USD 17,500

• Security review of the stable release by Danube Tech (Markus Sabadello): USD 10,000

• Rewards for paid advisors Day Waterbury, Zach Miltz and Andrew Hasse (launch retainer of USD 1,000 + USD 500 for months 2-7 each) : USD 12,000

• Paid alpha testers (two rounds of 5 people): USD 500

• AI costs (Infomaniak-hosted models by default, Claude as fallback): USD 2,500

• Running costs (hosting & subscriptions for 7 months at USD 250): USD 1,750

• Travel, tickets and accommodation for EAGxBerlin and EA Global London: USD 3,000

• Buffer (~5%): USD 2,750

Anything between the two amounts extends the work month by month.

Who is on your team?

Jakob Possert-Bienzle: Software Advisor & Builder with a background in organizational development.

I have been involved in the broader technical field of creating the 'Web of Trust' for more than a year and have been the driving force to stop talking and start building.

My previous positions include that I was a Supporting Co-founder for another web of trust enabling community and for an app that was intended to coordinate talent & resources in the global NGO space, in addition to a stint as Meta-Curator for the Internet Archive, as in coordinating those who have thought deeply about what needs to get done for a better internet and beyond.

Before that I have worked as an advisor to a funder, software builder for brainzmagazine.com/post/resolve-emotional-patterns-with-the-emres-app-your-path-to-emotional-freedom among others and organizational development consultant.

This unique blend lets me easily take the different perspectives in the ecosystem and how to support the transformation that a new tool can allow for an organization or community.

Advisors

Day Waterbury: Internet Archive; connected to web-of-trust and decentralized-identity circles. (paid advisor)

Andrew Hasse: documentary filmmaker, storytelling strategy expert for tech projects (paid advisor)

Zach Miltz: product manager, systems architect, and community weaver designing knowledge and coordination tools that help communities connect people, opportunities, and resources. (paid advisor)

Kevin Triplett: philanthropist, film-maker and funder of projectweave.tech

During the launch period, paid advisors will support participant recruitment, onboarding feedback, consent-flow testing, and a structured retrospective on adoption barriers.

From Month 2 through EA Global London, each advisor’s monthly retainer covers a monthly review, asynchronous feedback on the current build, and agreed recruitment, testing, or documentation tasks.

Supplier

Markus Sabadello: technical expert in decentralized identity (Danube Tech).

What's your track record on similar projects?

I have organized this team to start building these prototypes and have created a portfolio of related tools including QuestHub an early stage platform to organize resources for 'quests' that the world needs to get done. For clients, I have built a Somatic Therapy app and a Systems Transformation course-companion.

Day has been working with Internet Archive on credential system and has been in the field for a long time including with the First Person Project and has recently built a chat functionality on DIDs.

Zach combines product strategy and systems architecture for community knowledge and coordination tools. His work includes co-designing a community knowledge schema, connecting personal and shared knowledge architectures, and developing governance practices that distinguish sourced claims from inference. He also brings hands-on delivery of a community knowledge agent and five years in enterprise technology consulting, progressing to product manager.

Andrew has been filming and working on the storytelling of projects that support human flourishing through technology.

Kevin has been funding and bringing together the different stakeholders in the larger Web of Trust vision that this can ultimately lead to.

Markus has led projects for DIDs, a necessary component to reduce capture risk.

What are the most likely causes and outcomes if this project fails?

The most likely cause for the querying to fail is that there aren't enough people recording useful items to be queried. And for also the solo-player version to fail is that recording notes after conversations and the visual personal CRM itself is not useful enough.

One associated risk is that while the note-taking habit might stick, the querying the network might not.

I have described above how to mitigate this risk by the combined search.

And the outcomes would be that users have recorded information from meetings that they could export to their note taking systems - because notes live on each person's own device, no platform is left holding anyone's data.

And if it fails, we publish what did and didn't work, so the next attempt starts further ahead.

The app itself will stay open-source and the code stays usable in other contexts.

How much money have you raised in the last 12 months, and from where?

An early backer supported 'Web of Trust'-related work with €2,000, and the complementary project of simple in person recordings with ~€3,500 earlier this year.

An application for a grant to the city of Vienna (EUR 10,000) is under review.