INTRODUCING / AN INDEPENDENT LEARNING INITIATIVE
Deeper foundations. Further possibilities.
A three-month initiative to deepen our mathematical understanding of AI through close reading, first-hand derivation, open-source practice and community dialogue. Build understanding through practice. Let real questions lead the next step.
THE FOUNDATION
Project Bedrock is a three-month initiative for deep study and practical work in AI. Starting with the mathematics behind large models, it connects the history of deep learning, close reading, derivation and open-source practice. Sharing through X and Xiaohongshu will provide a continuing record of ideas, questions and feedback. We aim to build stronger foundations through work we carry out ourselves, find meaningful problems through community dialogue, and develop representative open-source contributions.
Launch preparations are complete. LANCER is examining the feasibility of the learning and practical directions, working together with CIE and other collaborators. Mathematical study, close reading and hands-on building will run throughout, producing understanding that can be explained, revisited and developed further.
Return to definitions. Understand the relationships.
Read the reasoning. Work through it yourself.
Put understanding into working code.
Let shared work lead to better questions.
LEARN · DERIVE · BUILD · SHARE
Return to definitions. Understand the relationships.
Study the mathematical foundations through concrete questions about large models, deep learning and neural networks. Examine how assumptions support a derivation, how representations shape learning and which conditions a conclusion requires, building the ability to explain and assess methods.
Read the reasoning. Work through it yourself.
Connect the history of deep learning, foundational papers and current research, recording how questions and methods evolved. Work through important formulas and use notes and small computations to check understanding, leaving knowledge that can be revisited and discussed.
Put understanding into working code.
Study how existing open-source projects connect theory, implementation and evaluation. Build and improve around practical problems, working toward representative projects of our own with clear documentation, development records and inspectable results.
Let shared work lead to better questions.
Plan to share knowledge, paper readings, progress and practical experience through X and Xiaohongshu. Bring questions from readers and developers back into study and implementation, continuing the accounts and their public record beyond the intensive phase.
THREE MONTHS · A LASTING FOUNDATION
The following is a planned path. Scope and pace will evolve with feasibility work and practical progress.
Map the mathematical foundations and central questions, connect the history of deep learning with representative papers, and establish a reading and derivation path with an initial set of notes and open questions.
Turn key derivations into small implementations, study open-source projects in depth, and record the differences between theory and engineering. Use practical difficulties and community feedback to identify directions worth pursuing.
Prepare learning and practical work for public use, advance representative open-source projects, and develop their documentation. Use feedback to define the next questions and carry the work forward through continued sharing and maintenance.
BUILT TO BE SHARED
We plan to establish an X account and share knowledge, paper readings, practical experience and progress alongside Xiaohongshu. Real questions from the community will help identify directions to investigate and improve. The accounts and sharing will continue beyond the three-month intensive phase.
Planned public outputs
PROGRAMME UPDATES
This is the continuing record of Project Bedrock: study and derivation notes, paper readings, open-source practice, and questions and progress emerging from community dialogue.
Preparations are complete. A three-month programme of mathematical study, open-source practice and community dialogue begins.
Read the update →THE WORK BEGINS
Project Bedrock is underway. We will continue to document progress in reading, derivation and practice, and publish access to the associated accounts and open-source work as they become available. A concrete question is a welcome place to begin a conversation.