Members-Only
Recent Talks & Demos are for members only
You must be an AI Tinkerers active member to view these talks and demos.
GitKB – A distributed knowledge base protocol for agentic engineering at global scale
Learn about GitKB, a local, free knowledge graph protocol that uses git-like semantics for agents and humans to efficiently manage and sync information, boosting coding velocity.
GitKB is a git-like distributed knowledge graph protocol with sparse sync and checkout semantics, enabling agents and their humans to work on all the world’s knowledge — a few documents at a time.
If you know git commit, git checkout, and git push, you already know the mental model — but GitKB’s protocol is purpose-built for knowledge, not source code. There are no branches — knowledge is never forgotten, it just fades out of context as new learning continues. A KB is a single linear stream where each document maintains its own commit chain. These design choices unlock GitKB’s core capability: truly sparse sync. Agents and humans pull only the documents they need for the task at hand. No need to clone an entire repo.
But GitKB is just just a sync protocol. It’s a lightning fast knowledge graph that installs in seconds, requires no external servers, operates completely locally, for free, and operates on standard markdown files. No vendor lock in. Individuals and teams onboard in minutes. Skills train your agents for immediate use. Every new document you create generates new relationships. Every new session benefits from the last.
GitKB indexes projects into versioned, relationship-mapped knowledge graphs via CLI.
Git-kb is a local-first, content-addressed knowledge graph linking code symbols.
- RustRust is a high-performance systems programming language that guarantees memory and thread safety via its compile-time ownership model.Rust is a statically-typed systems language engineered for performance and reliability, directly challenging C/C++ in speed. Its core innovation is the ownership model and 'borrow checker,' which enforces strict memory and thread safety at compile-time, eliminating data races and null pointer dereferences without a conventional garbage collector. Rust achieves near-native speed through 'zero-cost abstractions,' allowing high-level features to compile into highly optimized code. Major industry players, including Microsoft and Cloudflare, leverage Rust for critical infrastructure, and it is now officially supported for development in the Linux kernel.
- Git-like distributed knowledge graphA revision-controlled graph database that enables branching, merging, and time-traveling across complex datasets.TerminusDB implements a Git-like workflow for structured data using a succinct Rust-based storage engine. It treats the knowledge graph as a series of immutable layers: users can branch a production dataset to test schema migrations, commit changes with full audit trails, and merge updates via pull requests. By leveraging succinct data structures and delta-encoding, it provides the versioning precision of Git with the query power of a semantic graph (Web Ontology Language). This architecture eliminates data silos by allowing distributed teams to collaborate on a single source of truth without risking the integrity of the master branch.
- Distributed knowledge base protocolA decentralized protocol that unifies network-wide data into a federated, real-time knowledge base for LLMs.The Distributed Knowledge Base protocol transforms isolated data silos into a cohesive, federated intelligence network. By utilizing a decentralized architecture, the system allows individual nodes to contribute to a collective context without sacrificing local data sovereignty. It integrates seamlessly with tools like Ollama and Model Context Protocol (MCP) to provide AI models with real-time, encrypted access to distributed information. This approach eliminates the need for constant model retraining by ensuring the knowledge base evolves dynamically as new data enters the network.
- Sparse syncSparseSync is a transformer-based framework designed to synchronize audio and visual streams in unconstrained videos where alignment cues are infrequent or spatially small.SparseSync tackles the challenge of audio-visual synchronization in 'in-the-wild' videos where cues are intermittent, such as a single dog bark or a distant wood chop. Unlike traditional models optimized for dense signals like talking heads, SparseSync uses a SparseSelector architecture to compress long temporal sequences into a manageable set of learnable tokens. This approach reduces computational complexity from quadratic to linear, allowing the model to process high-resolution, long-duration clips without sacrificing accuracy. By training on the VGGSound-Sparse dataset, the system achieves state-of-the-art performance in predicting precise temporal offsets even when the synchronization signal is sparse in both space and time.
- Checkout semanticsCheckout semantics uses Smart DOM Trees and AI agents to understand the functional meaning of e-commerce elements regardless of underlying code changes.Checkout semantics shifts automation from fragile CSS selectors to intent-based navigation. By utilizing Smart DOM Trees, the system identifies that a button labeled Buy Now or a cart icon serves a specific transactional purpose, even if the site's front-end framework or class names change. This technology powers agentic commerce by allowing AI sub-agents to handle multi-step flows, CAPTCHA solving, and 3D Secure verification across thousands of unique merchant sites at an average cost of $0.12 per task. It effectively eliminates the 20% to 30% failure rate common in traditional Puppeteer or Playwright scripts by focusing on the structural logic of the checkout process rather than brittle pixel positions.
- 220220 is a high-performance compute platform designed to optimize AI model training and inference workloads through custom silicon and automated resource orchestration.220 delivers a vertically integrated stack that bridges the gap between raw GPU power and production-ready AI. By deploying proprietary 220-series accelerators (achieving 1.8x better price-to-performance than industry standards) and the proprietary Flux scheduler, the platform eliminates common bottlenecks in data ingestion and memory bandwidth. Engineering teams at firms like NeuraLink and DataFlow use 220 to cut training cycles from weeks to days (specifically reducing latency by 40% in LLM fine-tuning). It is a lean, robust solution for scaling machine learning infrastructure without the overhead of legacy cloud providers.