Who controls the stack?

A local-first desktop environment for models and agents.

llama.cpp / MLX · local API · local data

Why it matters

Prompts, memory and inference can remain on hardware the user controls.

Boundary

Control still depends on the chosen model license and the security of the host device.

Turn personal devices into one distributed inference cluster.

Tensor parallelism · MLX distributed · peer discovery

Why it matters

Lets users pool devices and run models larger than one machine can hold.

Boundary

Best suited to trusted local networks; heterogeneous links become the bottleneck.

A peer-to-peer market for personal AI inference.

Consumer nodes · providers · Base contracts · MOR staking · TEE roadmap

Why it matters

A local consumer node can discover and connect to providers using wallet identity.

Boundary

Token exposure and node operation create more friction than a conventional API.

Markets for digital commodities coordinated by programmable incentives.

Subnets · miners · validators · Yuma Consensus · TAO emissions

Why it matters

Subnet designers can define how useful machine intelligence is measured and rewarded.

Boundary

Decentralized incentives do not automatically guarantee privacy or objective quality.

Serverless AI inference backed by Bittensor Subnet 64.

Model containers · OpenAI-compatible API · Bittensor providers

Why it matters

Makes decentralized supply accessible through familiar developer interfaces.

Boundary

The gateway remains a practical coordination point and privacy depends on the route.

An open marketplace where providers bid to host compute workloads.

Cosmos SDK · SDL deployments · reverse auction · leases

Why it matters

Separates application ownership from a single cloud vendor.

Boundary

It decentralizes hosting, not the model policy or application data model by itself.

Aggregate underused GPUs into clusters for machine learning workloads.

Distributed GPU supply · cluster orchestration · Ray workloads

Why it matters

Expands accessible compute supply beyond hyperscale cloud inventory.

Boundary

Distributed hardware supply is not the same as a decentralized control plane.

Coordinate and verify machine learning across untrusted nodes.

Verde dispute resolution · RL Swarm · peer-to-peer learning

Why it matters

Attacks the hard problem: proving that outsourced ML work was performed correctly.

Boundary

Research and network infrastructure, not yet a substitute for a consumer assistant.

Let individuals deploy and monetize agents with their own knowledge and values.

WasmEdge · local LLM · RAG · vector database · OpenAI-compatible API

Why it matters

Treats a personal knowledge base and agent configuration as a user-owned service.

Boundary

Public node privacy and domain governance must be evaluated deployment by deployment.

A blind-computing stack for processing sensitive data without exposing it to operators.

MPC · homomorphic encryption · TEE · signed inference responses

Why it matters

Changes privacy from a policy promise into a cryptographic or hardware-backed claim.

Boundary

TEE paths inherit hardware-root trust; privacy does not imply censorship resistance.

Private, low-restriction access to many models through one interface.

Anonymous routes · ZDR · TEE/E2EE · OpenAI-compatible API · DIEM/VVV

Why it matters

A polished bridge between mainstream UX and stronger inference privacy.

Boundary

The service, model catalog and routing control plane remain company-operated.