# Memory Meets Motion: The FalkorDB Standouts > Source: > Published: 2026-08-18 06:58:31+00:00 # Memory meets *motion*: the FalkorDB standouts On August 3, 2026, builders gathered at Frontier Tower in San Francisco for Memory Meets Motion, an eight-hour hackathon built around one mandate: every project had to make real, load-bearing use of four sponsor technologies: FalkorDB as the memory layer, RocketRide.ai as the execution and orchestration engine, Guild.ai for multi-agent coordination, and LaserData as the real-time event stream. **Memory Meets Motion Hackathon** Frontier Tower, SF Aug 3, 2026 01 ## Atrium 1ST PLACE OVERALL Atrium turns a single search into a full curriculum: it crawls the web for sources, drafts cited lessons, then runs those lessons through a simulated classroom before a real student ever sees them. Its cleverest move is graph-native: rather than lumping every student who scores 40% into one remediation bucket, Atrium models each wrong mental model as its own node in FalkorDB, so two students who got the same grade for completely different reasons get grouped, and taught, differently. That grouping is a two-hop Cypher traversal that a flat table or a vector search simply cannot express. Firecrawl supplies the web research layer alongside the FalkorDB memory graph. 02 ## MuscleMemory A simulation platform where one fixed humanoid robot, MM-01, learns household delivery tasks across changing apartment layouts. The robot’s body never changes, only its accumulated experience does, and FalkorDB is what makes that experience compound: it links failures, corrections, and lessons across simulation runs so later training curricula target real recurring mistakes rather than starting from scratch. LaserData handles the live sensor and telemetry stream alongside it. 03 ## AgentCourt Precedent Every risky AI action gets put on trial: a Prosecutor, Defense, and Judge agent, all running for real on Guild.ai, argue over whether to allow it, while FalkorDB stores prior rulings as precedent so the next similar case cites what happened before. A deterministic Bailiff, not an LLM, has final authority to require backups or human approval regardless of what the Judge recommends, and RocketRide executes only the approved action. 04 ## Continuum: The Open Loop OS Continuum tracks Open Loop Debt, meaning all the meetings, briefs, and renewals that quietly never get closed, and measures it in dollars at risk. FalkorDB mirrors the durable graph of people, projects, and decisions, while a wider bench of vendors shows up here too: RocketRide for pipeline execution, LaserData for durable events, Guild.ai for experiment logging, Linkup for live web research, and even Snyk for dependency scanning. 05 ## Facet Facet is a news feed that reshapes the same underlying facts differently depending on how a specific reader learns best: a visual map for one person, an evidence-dense deep dive for another. FalkorDB stores the graph of readers, personas, preferences, and topics that drives that reshaping, while a Guild.ai editorial council of five agents separately handles curation, research, and source verification, and RocketRide coordinates the delivery pipeline. 06 ## Drift A simple, sharp idea: decisions change in meetings, but GitHub issues do not know that. Drift listens to meeting transcripts, stores every decision in a FalkorDB graph, and when a later meeting contradicts an earlier one, it posts a what-changed comment on the affected issue through a governed Guild.ai agent with a human approval gate. Built solo in an afternoon, notably using Gemini alongside the FalkorDB and Guild.ai combination. 07 ## Smart Home Orchestrator When two people in the same room want conflicting things, this system has their AI advocates argue it out, and it gets fairer over time because FalkorDB remembers who conceded last time. The fairness logic is explicitly a multi-hop graph query, room to occupants to history between them, which is the kind of relationship that is awkward to model outside a graph database. LaserData feeds the live sensor stream, Guild.ai runs the advocate agents, and RocketRide carries out the final decision. More Standouts Worth a Look 08 ## Hopper: Security Notification Triage This Hopper answers a very specific question: not is this dependency vulnerable, but does this vulnerability cost me a customer, and by when do I legally have to tell them. It walks a FalkorDB graph from advisory to package to dependency to repo to service to customer to contract clause in under 100 milliseconds, and treats no path found as a real proof of safety rather than an empty result, deliberately skipping vector search because a negative result needs an actual traversal, not a similarity score. RocketRide stores its execution pipelines as graph nodes, LaserData supplies real-time telemetry, and Guild.ai gates the one action, customer notification, that carries legal risk behind a human click. 09 ## Hopper: Adaptive Distribution Network A different, unrelated project that also chose the name Hopper: a supply-chain simulator where two agents negotiate cost versus urgency when a demand spike hits a hub, then write the resulting decision back into FalkorDB so the next disruption benefits from precedent instead of starting cold. The negotiation runs as an actual RocketRide pipeline with an Anthropic-powered agent handoff, and LaserData carries the live demand-spike signal. 10 ## Onari: Ambient Context Capture A voice-first ambient assistant that captures raw, half-formed thoughts the moment you have them, bound to whatever window and moment you were in, and later stitches a whole work session together against long-term memory to propose and execute a real action. FalkorDB holds that accumulating context graph, LaserData carries capture events in both directions, Guild.ai consolidates and gates human approval, and RocketRide is what actually executes the resulting task. 11 ## Nous A personal assistant that catches the promises you make in conversation before they are forgotten, and, more interestingly, watches you handle routine tasks such as paying an invoice once, then compiles what it learned into a reusable RocketRide pipeline it can run on its own next time. FalkorDB stores the personal life graph of people, commitments, and learned protocols, LaserData is the life-event stream, and Guild.ai’s distiller and critic agents decide whether a learned skill is safe to arm. 12 ## DealBench An AI-powered M&A cross-examination simulator, complete with a courtroom interface where counsel, judge, and witness agents argue over a fictional acquisition dispute using a pre-built case-document corpus. FalkorDB backs the memory service behind the courtroom, RocketRide runs the actual courtroom-turn pipeline, and Guild-compatible agents fuel the underlying inference. Notably, this team built it so a live run fails loudly rather than quietly falling back to canned dialogue. 13 ## Oto.ai Oto.ai watches a developer’s work session, builds a live FalkorDB task graph of the steps, decisions, and blockers as they happen, and, if the person steps away mid-task, has a governed agent pick up exactly where they left off and open a pull request explaining what it inherited and what it did. LaserData supplies the replayable event backbone, Guild.ai handles session governance and PR risk review, and RocketRide’s node graph runs the actual fetch, edit, and test-retry loop. [Speak with an Engineer at FalkorDB](https://www.falkordb.com/contact-us) [Start Free](https://app.falkordb.cloud/signup) ## Author - Guy Lubovitch serves as a Pre-Sales & Post-Sales Engineering Leader at FalkorDB, bringing with him over a decade at Redis, where he held a series of senior customer success and solutions architecture roles. Most recently Senior Director of TAM Americas & Cloud, he built and scaled Redis' Global Cloud Customer Success organization from the ground up, driving it to support Fortune 500 customers and multi-million dollar ARR deals. Earlier in his career at Redis, he led a team of solution architects for large enterprise and strategic accounts. Before Redis, Guy spent several years at GigaSpaces as Director of Solutions Architecture. He studied at Bar-Ilan University and is based in the San Francisco Bay Area.