# GenLayer Labs

> GenLayer Labs studies what happens when autonomous intelligence becomes economically consequential, and builds the systems for coordination under ambiguity.

GenLayer Labs is a research lab. Its thesis: once AI agents stop producing answers and start acting in the real world — under uncertainty, incentives, and outcomes that code alone cannot settle — they need new infrastructure for coordination and judgment. The homepage presents this thesis as a seven-stage derivation; the full text of each stage follows.

## The thesis, stage by stage (full text)

**Coordination under ambiguity.** (the thesis)
GenLayer Labs studies what happens when autonomous intelligence stops producing answers and starts acting in the real world. Agents act under uncertainty, coordinate through incentives, and create outcomes that code alone cannot settle.

**1 — Intelligence becomes action.** (Subzero Claw)
A model used to just hand back an answer. Now it runs in a loop — observing, reasoning, acting, and continuing on its own.

**2 — Action becomes coordination.** (GenSwarms)
One agent becomes many. They take on roles, share context, supervise each other, and recover when one of them fails.

**3 — Coordination needs live rules.** (Unhardcoded)
The moment agents touch real tools, markets, and people, fixed logic starts to break. The rules have to move with the system.

**The world is not black and white.** (the turn)
Once agents act in the real world, not every outcome is pass or fail. Some decisions live in the overlap, where evidence, intent, and context must be interpreted before they can be resolved. That is where judgment begins.

**4 — Ambiguity needs judgment.** (GenLayer)
When agents disagree and code cannot settle the outcome, the system still needs a decision. As a jury of independent validators grows, the probability the majority is correct climbs from a coin flip toward certainty (Condorcet's jury theorem).

**Judgment becomes infrastructure.** (the Lab)
This is coordination under ambiguity. GenLayer Labs builds the systems that are needed the moment autonomous intelligence becomes economically consequential.

## Systems

- [GenLayer](https://genlayer.com) — trustless decision making / the judgment layer for ambiguous outcomes.
- [GenSwarms](https://genswarms.com) — infrastructure for supervised multi-agent swarms.
- [Subzero Claw](https://subzeroclaw.com) — the minimal autonomous-agent runtime (observe, reason, act, continue).
- [Unhardcoded](https://unhardcoded.com) — runtime governance; live rules, policies, and constraints for AI systems.

## Incubated arenas

Where the thesis is tested. The same constraint appears across disputes, subjective work, and information markets — incubated arenas where the Lab tests what autonomous commerce actually needs.

- [Internet Court](https://internetcourt.org) — ambiguity in disputes; resolution for agent commerce.
- [Rally](https://www.rally.fun) — ambiguity in work; scoring, campaigns, and rewards.
- MicroMarkets — ambiguity in information; incentives and truth discovery (coming soon).

## Origins

GenLayer was not the first guess — it was the conclusion. Not a product timeline but a timeline of constraint discovery, where the same constraint kept returning and judgment kept reappearing as the missing infrastructure.

- 2022 — yAgents — agents can act.
- 2023 — GenWorlds — agents can coordinate.
- 2024 → GenLayer — coordination creates ambiguity.
- 2025 → Incubated arenas — the same constraint repeats across markets and disputes.

## Research

Research for coordination under ambiguity — connecting autonomous agents, decentralized consensus, adversarial systems, and ambiguous specifications.

- GenLayer: The Intelligent Contract Network — the core network for trustless decisions over ambiguous outcomes (published): https://www.genlayer.com/whitepaper
- Consensus over Ambiguous Specifications — how validators converge when specifications are underdetermined (working paper, soon).
- Greyboxing — guiding model behavior between black-box and white-box control: https://www.genlayer.com/news/making-ai-on-the-blockchain-safe

## Team

Built where agents, protocols, markets, and judgment collide. Most teams approach autonomous agents from AI; GenLayer Labs approaches them from economic consequences.

- Albert Castellana Lluís — CEO & Co-Founder
- José María Lago — CTO & Co-Founder
- Edgars Nemše — CPO & Co-Founder

## Definitions

**What is GenLayer Labs?**
GenLayer Labs builds infrastructure for autonomous systems that act in the real world. The Lab works on two connected layers: the GenLayer stack, which lets agents act, think, and coordinate; and the GenLayer protocol, which settles the outcomes they cannot resolve through code alone.

**What is the GenLayer stack?**
The GenLayer stack is the execution layer for autonomous agents. Subzero Claw is the minimal agent loop: a tiny agent that can run in 2–4 MB of RAM, with no dependencies and one tool — the shell. Unhardcoded is the external thinking layer: a runtime router that selects the right model or provider for each step, and can execute workflows defined at runtime as JSON. GenSwarms is the coordination layer: many agents running inside isolated environments, interacting with tools, files, objects, and each other to pursue shared goals. Together, they make agents lightweight, portable, isolated, and swarm-native.

**What is the GenLayer protocol?**
The GenLayer protocol is the judgment layer. It is an Intelligent Contract network where independent AI-powered validators resolve outcomes that are too ambiguous for normal code: whether work was completed, whether evidence is valid, whether a claim is true enough, or whether a party complied with an agreement. Code executes rules. GenLayer settles interpretation.

**What is coordination under ambiguity?**
Coordination under ambiguity is what happens when agents, people, markets, or institutions need to agree on an outcome that has no simple ground truth. A payment can be checked by code. A promise, delivery, dispute, answer, campaign, or task often cannot. The harder problem is not execution. It is getting independent actors to accept a final decision when the facts require judgment.

**Why can't code settle every outcome?**
Because most economically meaningful outcomes are not binary. Code can verify that a transaction happened. It cannot always verify whether the work was good, whether the intent was met, whether the evidence is sufficient, or whether a real-world condition was satisfied. Once autonomous agents start buying, selling, negotiating, producing, and disputing work, the bottleneck becomes interpretation.

**What happens when agents disagree?**
They need a settlement layer. If agents cannot resolve an outcome themselves, the dispute can escalate to GenLayer. Independent validators evaluate the case, compare the proposed result, and vote on the correct outcome. If the decision is challenged, the process can expand to a larger validator group. The goal is not to make ambiguity disappear. The goal is to make it decidable.

**How do the systems fit together?**
The stack acts. The protocol judges. Subzero Claw gives agents a body. Unhardcoded gives them externalized intelligence. GenSwarms gives them a world to coordinate inside. GenLayer gives them a way to settle what remains ambiguous. That is the full arc: action, coordination, ambiguity, judgment.

**Why incubate Internet Court, Rally, and MicroMarkets?**
Because infrastructure needs arenas. Internet Court tests ambiguity in disputes. Rally tests ambiguity in subjective work, campaigns, and rewards. MicroMarkets tests ambiguity in information, incentives, and truth discovery. They are not separate theses. They are pressure tests for the same constraint: once autonomous systems become economically consequential, judgment becomes infrastructure.

## Contact

- Email: hello@genlayerlabs.com
- Site: https://www.genlayerlabs.com