@Replit 10 year anniversary
Build 6 -
@KevinBlumson collab
What if the universe is a tiny rewrite rule applied to a graph, trillions of times per second?
That's the Wolfram Physics Project premise. Their team has catalogued hundreds of small hypergraph rewriting rules and shown some produce structures resembling spacetime, quantum mechanics, and general relativity. The open question: which rule is the right one? The search space has over 125 million candidates, and verifying a single one takes millions of simulation steps.
We built Hypergraph Rule Explorer to go look.
Every browser tab that opens the app becomes a compute node. JavaScript calls into a full Python interpreter (Pyodide, via WebAssembly) to simulate rules locally. Each session leases a slice of unexplored rule space from a coordinator, runs verification, and ships hits back. No install. No account needed.
The verification pipeline runs four gates on each candidate: causal invariance (multiple update orderings must converge, the hypergraph equivalent of Lorentz invariance), growth class analysis, asymptotic dimension estimation (flagging rules that consistently produce ~3 1D structure), and branchial graph analysis for quantum-like branch merging.
Rules passing all four gates are recorded as discoveries, stored with a permalink, a downloadable reproduction recipe, and a rendered hypergraph thumbnail.
Novelty detection runs against a canonical registry of 21,000 known rules. Candidates are normalized to canonical form before comparison so no variable-renaming or isomorphism creates a false positive.
Stack: Flask, SQLite, PostgreSQL coordination store, Pyodide, Plotly.js, Jinja2, Authlib OIDC, all running on Replit with zero external infrastructure.
From idea to a distributed physics search engine with auth, a leaderboard, visualization, and an Atom discovery feed: one environment, one tab.
Mad science. Possibly the Theory of Everything. Built on Replit.
#Replit #AdeviousAI #BuildInPublic #Science