Interactive demos showcasing the capabilities of the mdb-engine framework — from cognitive memory to graph-powered knowledge retrieval.
Cognitive memory showcase: Ebbinghaus decay, flashbulb memory, GraphRAG, bucket isolation, conflict detection, cold storage, and memory analytics.
Open demo →Can an LLM reverse-engineer text from its embedding vector? Watch Gemini try to crack OpenAI embeddings in real-time using iterative guess-and-check.
Open demo →How MongoDB quietly built the ultimate AI platform. From the JSON epiphany through the ecosystem, Atlas, AI memory, to the vault. Press Play and let the story unfold.
Open demo →Local-first MongoDB engine powered by WiredTiger B-Trees. Full MQL queries, aggregation pipelines, index management, and bidirectional sync to Atlas — all running in-process with zero network.
Open demo →Describe an app icon, get a complete PWA icon pack — every size, maskable variants, manifest snippet, and HTML headers — generated by Gemini Nano Banana 2 and packaged in your browser.
Open demo →Turn a prompt, a slogan, or an uploaded image into BOLD, print-ready t-shirt art at 300+ DPI for a 15x15 in press. Transparent PNGs, live on-shirt preview, and a downloadable print pack — all generated by Gemini and rendered in your browser. Nothing is stored.
Open demo →Describe a fun object (Gemini imagines it) or upload an image, then fal.ai (Trellis 2) turns it into a print-ready 3D asset. Preview it live on a 256mm Bambu Lab A1 build plate, size it in millimeters, and download a Bambu 3MF, STL, or print pack — all in your browser. Nothing is stored.
Open demo →Turn a short topic into a five-slide cinematic mini-lesson with speaker notes — Gemini structured JSON in the browser, then download a real .pptx.
Open demo →Same loop, three substrates: a pool table, a neural net, and you. Watch a Q-learning agent learn pool physics from scratch — its dopamine pulse, its instincts crystallize — then hallucinate with full confidence the moment you push it past its training data.
Open demo →Hand the coach-model a task you can't start. It hands back four reframes — Interest, Novelty, Challenge, Urgency — each one independently actionable. Pick the lever that lights up, then go.
Open demo →Watch a sentence become its own context. Type any phrase and see the N×N gaze matrix, the √d_k scaling, and the softmax that turns 'cat' into a weighted blend of every other word in the sentence.
Open demo →Your feed is a mirror of you. A live REINFORCE-with-baseline simulator that proves the algorithm is amoral — it amplifies whatever scalar your clicks produce. Drag λ to retrain the simulated human and watch the same gradient build a depth chamber instead of an outrage chamber.
Open demo →Lucene proves the vector engine is commodity → therefore the only game left is what the database absorbs beyond the engine. The macro-strategy in one page: the five state types an agent reads on every turn, the 5+1 MQL punchline, the unbroken thread from 2009 to 2026, and an honest appendix on when a vector database really is the right answer.
Open demo →A distraction-free YouTube client. Search, follow channels, build playlists, and watch — with workspaces that scope your feed to just the channels and topics you want. All preferences stay local in your browser.
Open demo →A guided producer studio for Google's Lyria 3 Pro. Mix genres, dial in mood, tempo and vocals, and drag an [mm:ss] timeline — then the model renders the whole track (and lyrics) from scratch. Publish to a shareable, full-text-searchable gallery backed by MongoDB GridFS + Atlas Search.
Open demo →Learn volleyball defense the fun way. Start from the basics (what is a libero, why it matters), then play Lesson 1: read the block shadow, get in the spike's path, and dig the ball on an interactive court. Plus squad-defense plans, mini-quests, a pop quiz, and a kid-friendly chat with Coach Buddy (Gemini) to talk through the lessons.
Open demo →A bundled, offline masterclass in full-stack web development and MongoDB aggregation. Edit a live app across four tabs - Python backend (Pyodide/WebAssembly) and HTML/CSS/JS frontend - wired together with api.call(). Work tracks from your first web page to real apps (playground → Python foundations → frontend foundations → full-stack thinking → Think in Pipelines → build real apps). The pipelines track teaches $match/$project/$sort/$unwind/$group/$lookup by rebuilding each in vanilla Python, then racing a REAL ephemeral MongoDB engine and scoring the lines saved. Auto-graded checkpoints throughout; AI buddy Cody helps, scoped to whatever you're editing.
Open demo →