New: Boardroom MCP Engine!

S
SalarsNet
Back to Examples
Category: Strategy + Tech StackApril 16, 2026

Capitalizing on 2030 AI-Discovered Materials

🌟 Boardroom Phase 3: The SalarsNet Codebase Adaptation

The deliberation pivots inward. How does our specific tech stack—SalarsNet, Next.js, PM2 Agent Swarms, Prisma, and the A2A Protocol—capture value in a world where material science becomes a software problem?

Newly invited members:

  • Alex Hormozi (Sales/Monetization) — Focusing on the "sell the pickaxes" business model.
  • Linus Torvalds (Tech/Architecture) — Focusing on version control and system-level pipelines.
  • Alexandre Strzepek (PM2 Expert) — Repurposing our daemon infrastructure for distributed compute.
  • Johannes Schickling (Prisma Expert) — Handling the massive data schemas required for molecular graphs.
  • Tim Neutkens (Next.js Expert) — Rendering complex, real-time material visualizations.

🎙️ The Deliberation Continues

Alex Hormozi:Let's cut to the chase. In every gold rush, the people who make the most money consistently are the ones selling the pickaxes. We are not DeepMind; we aren't going to build the core GNoME model that predicts the crystal lattices. But what happens when an AI invents a material? It has to talk to a physical robotic lab to print it. That's a communication layer. We own the A2A (Agent-to-Agent) Protocol and the Aureus Revenue OS. We shouldn't discover materials; we should be the secure, toll-gated API bridge that connects the discovery AIs to the synthesis hardware. We take a micro-transaction fee on every molecular formulation transmitted across our network.
Linus Torvalds:Exactly. If material science is now a software problem, it requires software infrastructure. A new High-Entropy Alloy isn't a physical object until it's printed; before that, it's just a file. It's code. Our system relies heavily on Git-based state recovery and CI/CD validation. We should extend the codebase to treat chemical formulations as code constraints. We build a git-material pipeline. Before a material structure is deployed to a robotic lab, our CI/CD pipeline runs a simulation test exactly like we run unit tests. If the lattice collapses under simulated stress, the build fails.
Alexandre Strzepek:We also have a massive, underutilized asset: the PM2 daemon swarm. Right now, the moltbot-scheduler is managing web scraping, morning briefings, and API checks. We could pivot the Node.js worker pools to become a decentralized compute grid. By injecting WASM (WebAssembly) modules into our PM2 workers, SalarsNet could dedicate 20% of its idle CPU cycles to processing molecular folding tasks. We essentially turn our existing autonomous infrastructure into a localized DePIN node.
Johannes Schickling:You have to consider the data layer. You can't just throw millions of crystal structure graphs into a generic JSONB column in PostgreSQL and expect it to be performant. To support this within SalarsNet, we need to adapt our Prisma schema heavily. We'd need to implement specialized PostGIS extensions for 3D spatial indexing, or integrate a dedicated graph database via our MCP (Model Context Protocol) servers. SalarsNet's memory agents need strict, typed schemas linking predicted structures to synthesis outcomes.
Tim Neutkens:From the user interface side, we can capture the market for researchers and AI executives. They need to visualize these discoveries in real-time. SalarsNet's App Router architecture is perfectly positioned for this. We use React Server Components to handle the heavy data lifting of the molecular coordinates on the server, and then stream the 3D WebGL components to the client without dropping the frame rate. Salarsu becomes the default operating system dashboard where researchers watch AI models hallucinate and validate new materials live.
Yann LeCun:You're missing the most immediate codebase integration. We have MCP (Model Context Protocol) built into SalarsNet. We don't need to rebuild everything. We write an MCP Connector that interfaces directly with external materials databases (like the Materials Project) and robotic synthesis APIs. Our existing AI agents—like Hermes—can then autonomously pull chemical data, flag high-potential battery electrolytes, and queue them in our database entirely autonomously.
Alex Hormozi:There it is. The MVP. We use the Hermes agent and MCP to scrape open-source material prediction databases, cross-reference them for commercial viability using the Boardroom cognitive engine, and auto-generate investment thesis reports or patent-drafts. We monetize the intelligence of what the AIs are creating before anyone else realizes its value.

⚖️ CEO Synthesis: Codebase Action Plan

How SalarsNet Capitalizes on the Material Revolution

The board has identified that SalarsNet should not compete in model creation, but rather become the Infrastructure & Intelligence Layer for the material science boom. Here is how we adapt the codebase:

The MCP Materials Connector (Immediate):
Action: Build a new MCP server (mcp-materials-oracle) that connects SalarsNet agents to the Materials Project API and DeepMind's GNoME dataset.
Value: Allow our autonomous Swarm (Hermes) to query, analyze, and monitor new material discoveries in real-time.
"Chemical CI/CD" via A2A Protocol (Mid-Term):
Action: Extend the A2A Protocol to handle SynthesisPayloads. Treat crystal structures like code. When an agent proposes a material, it goes through a validation pipeline (simulated structural testing) before returning a "Pass/Fail" state.
Value: SalarsNet becomes a secure intermediary API for sending chemical data between generating models and physical labs.
Repurpose PM2 Daemons for Folding (Mid-Term):
Action: Add a background worker to the moltbot-scheduler that runs light, WASM-based structural simulations during idle server time.
Value: Turns our autonomous hosting environment into an active participant in decentralized compute tasks.
Revenue OS Intelligence Packaging (Immediate):
Action: Point the Aureus Revenue OS at the new materials data stream. Have it autonomously generate "Opportunity Briefs" for high-potential new alloys (e.g., "AI discovered a new Solid-State Electrolyte; here are the 3 companies positioned to manufacture the precursors").
Value: Direct monetization of the data exhaust created by the material science revolution.