Introducing Fathom, a platform for decentralized educational content creation, built on StarkNet.
We aim to accelerate the rate at which teaching and learning material forms for and around domains on the cutting edge. As an extreme but illustrative example, there is arguably no proper full and contemporary course on ZK-STARKs, StarkWare Cairo, and Practical Probabilistically Checkable Proofs (PCPs) for Blockchain Scaling and Delegated Computing. There are myriad related materials such as lectures, docs, wiki pages for subtopics, etc floating around, but they have yet to be assembled into a more structured, guided course. This is what we endeavor to create in our first pilot.
More generally, we and our hyper-curious friends struggle to find courses to satisfy our personal curiosity w.r.t. the cutting edge of, e.g.:
Distributed Systems for Crypto
Clean Energy Storage Technology
Carbon Capture
Quantum Information Theory and Quantum Computing with Today’s Noisy Systems
Large Language Models and ML At Scale
Small Data Regime and Sample-Efficiency Critical Reinforcement Learning
Mechanism Design and Opensource Governance for Web3 and OpenSource Software → The Tokenomics Course
To double click on a single example, a course discussing the key ideas undergirding opensource governance and examining case studies spanning crypto protocols, open-source software (Linux, Hadoop, Android), projects such as Wikipedia, etc would be intellectually intriguing and informative. However, no thorough and up-to-date course exists.
A lot of education is compulsory, so educators (e.g. college professors teaching Computer Networking) don’t have to necessarily agonize over how to make the material engaging and productive. Educational material is often a Bed-of-Procrustes.
Likewise, the professor can make strong assumptions about the prerequisite material that the students know and can restrict access to the course on this basis.
To return to our ZK-STARKs example, a course could be intended for varied demographics, e.g.:
Practitioners who plan to build decentralized systems (my go-forward term for open-sourced/crowdsourced projects) on top of StarkNet
Investors (large holders of ETH or a hypothetical future StarkWare token) who want to understand the technology and where it fits within the broader picture
Scientists and Researchers, endeavoring to understand the underlying math and tech so we can push it forward and expand it to new applications on the frontier (e.g. Advanced Machine Learning on StarkNet)
Each potential course would share a great deal of foundation, but the details, depth, and emphasis across dimensions of the undergirding material would vary greatly. In a bottoms-up and decentralized course creation scheme, many niches could spawn and thrive.
Decentralized Course Creation Schemes could also more effectively surface the-often implicit-knowledge dependency tree: e.g., ZK-STARKs are a spire atop a vast structure consisting of deep ideas from the vanguard of complexity theory, cryptography, and beyond. To choose a simpler example, though, let’s briefly examine ML. To master ML, one must learn a respectable amount about Linear Algebra, Probability and Information Theory, Numerical Computation, Convex Optimization (not even to mention all of the different regimes within ML).
“Composability is to software as compound interest is to finance” - Chris Dixon
Anyone who has been a TA or knows a teacher understands how much time teachers spend creating class notes, drills, assignments, practice problems, tests, etc. Many of these disappear. An ideal market mechanism would enable sharing and would surface the absolute best resource for each learning goal.
It would also allow a broad swath of people to collaborate on a shared tree, with macro-contributions of full branches (e.g., ML sub-course on the most important ideas from Probability to understand for Deep Learning) and micro-contributions (e.g., a single case study example regarding the distributed systems architecture of BitTorrent for “The Computer Science of Web3”). These sub-trees can be combined and recombined to develop content for a vast array of topics and learning goals.
The world’s best teachers of a topic should be everyone’s teachers. Everyone should have access to the material these people generate and these people should be rewarded for their excellence. As an example, The Michael Jordan and Tiger Woods of early childhood math education should be on every computer screen in the world. MIT OCW and other online education initiatives have paved the way here and we have a ton of admiration for what they have done.
However, the economics of education is still abhorrent, generally. Many world-class teachers can barely make a living. Great online educators, who contribute to Wikipedia, YouTube, etc. also often cannot earn a living off of their contributions (such that they can do this work full-time) via the ads or donation-based model.
The very first pilot of Fathom will be hands-on administered directly by our team of contributors, in collaboration with StarkWare and a panel of experts. However, we subsequently intend to codify our learnings in a scalable platform for creating broad forms of content at scale.
The theme of the first variant is “Learning Paths”. Under this theme, as members of the community deepen our knowledge of ZK-STARKs, Cairo, and the underlying technologies and ideas, we can leave a trace of materials we found helpful along the way. These chronological traces can be reconciled and compiled into learning lists that new members, just entering the community, can follow.
The endeavor will unfold in two phases:
Learning plan construction and team formation Learner participants generate a learning plan/outline and form teams. The goal of this is to get people to formulate a learning plan, to think critically about their goals and objectives, etc, and then to form sub-groups around common goals. (~2 weeks)
Learning path generation via “lead learning” Learners go deep to master these topics, all the while saving a list of links to resources that they found helpful using a chrome extension. This chronological list of links serves as a “learning path”. (~6 weeks)
The learning paths generated will then be reconciled and synthesized into master lists of top resources by our team and our expert advisors. This master list will be made available to the community and contributors will be rewarded with tokens based on the quality and uniqueness of their contributed resources/links.
We are excited to work with the community to deepen our collective mastery and to help us all make the future happen a little bit faster.