Make agentic work explicit enough to trust

beyond10x is building an open, product-neutral stack for work that agents can plan, execute, and prove without hiding authority or evidence inside a prompt transcript.

The stack begins with Agentic Principles grounded in observed work. Entity Runtime turns state, rules, operations, and events into deterministic decisions. AEP governs engineering work, while ESS makes system intent executable and testable. Harness runs model and tool loops; Metaharness observes and compares them. Substrate, services, and connectors bound the effects that reach the outside world.

The reading order is conceptual, not a claim that every project depends on the next. Every technical detail remains owned by its source repository and is displayed here at an immutable revision.

What should be different​

• Decisions, permissions, evidence, and failure are data rather than implicit convention.
• Deterministic mechanisms carry guarantees that do not belong in model behavior.
• External effects are attributable, bounded, inspectable, and recoverable.
• Public documentation tells readers what exists, what is experimental, and what remains planned.

Choose an outcome-led path or browse the ecosystem reference.