Cueng BlazarForce

🛠️ DevLog – More Ollama Models Reaching Ephemeral Pool We've started enabling a few more Ollama-backed models for ephemeral sessions on Testnet-0, focusing on the new batch in the 21–30 index range: - DeepSeek R1 14B - Gemma 3 27B - Qwen 3 VL 4B / 8B - Ministral 3 14B - Qwen 3 4B / 8B - Mistral 7B - Phi 3 3.8B / 14B 📌 Notes - Capacity is still small; errors / "no capacity" are expected while we tune things. - All of these remain fully usable via dedicated nodes even if ephemeral capacity is thin. - Node ops can continue to adjust include/exclude model configs based on their hardware + preferences. #Cortensor #Testnet #Ollama #DePIN

🛠️ DevLog – Agent/MCP Endpoints Live on DNS (Legacy → Streamable Fix, WIP) We've stood up minimal agent + #MCP endpoints behind DNS for Router Node discovery and testing. Legacy clients work, but a standards mismatch surfaced when testing with an external MCP client (Apify). This is still in-progress scaffolding. 🔹 What's Live Now - Router DNS endpoints are up and passing existing test scripts: router-dev-0.cortensor.network/api/v1/ping router-dev-0.cortensor.network/api/v1/about - MCP endpoint is reachable and functional for our current (legacy) flow: router-dev-0.cortensor.network/mcp/messages - SSE handshake returns session info as expected via: curl -N -H "Accept: text/event-stream" router-dev-0.cortensor.network/mcp/messages 🔹 Issue Found (External MCP Client) - Apify's MCP client fails while browser/legacy SSE works. - Root cause: our server is using the older HTTP+SSE session model (expects session_id early), while Apify uses modern Streamable HTTP (initialize via POST without a session id). - Result: initialize POST gets 400 session_id is required. 🔹 What We’re Changing - Support Streamable HTTP sessions on the same MCP endpoint. - Allow first initialize POST with no session id, return Mcp-Session-Id header, require it only after init. - Keep legacy SSE compatibility so existing tooling doesn’t break. 🔹 Next Steps - Patch /mcp/messages to be streamable-compliant and retest with Apify + other clients. - Document the updated MCP handshake for node ops and external agents. - Continue iterating toward a single Router-Agent process that owns #REST + #MCP + #x402 together. #Cortensor #AI #DePIN #AIInfra #MCP

🛠️ DevLog – #ERC8004 Deeper Dive & Prerequisite Setup (Continuation, WIP) This week’s agentic exploration & experimental track is running in parallel: #MCP + #x402 Router experiments and #ERC8004 discovery. All are still exploratory, but together they're shaping what Router-Agent needs to become in later #Testnet phases. 🔹 #ERC8004 Progress - We're now able to register a mock agent with baseline attributes. - This validates the entry step for "Router-Agent as an ERC-8004 service," but it’s still just scaffolding. 🔹 What We Learned - #ERC8004 expects real, stable capability surfaces beyond registry writes. - Proper discovery will require prerequisite setup like DNS/HTTPS endpoints and structured comms metadata. 🔹 Prerequisites To Set Up Next - Stable DNS/HTTPS endpoints for at least one Router-Agent / session. - Clean capability ads that point to #MCP + #x402 routes. 🔹 Next Steps - Deeper review of #ERC8004 attributes and linked standards: #A2A, #OASF/OSSF taxonomies, #DID/#ENS, wallet + reputation fields. - Map how Router-Agent should publish #MCP + #x402 endpoints for trustable discovery and task validation. 🔹 Scope - Still research/prototyping to surface unknowns early; full alignment comes later in Testnet. #Cortensor #AI #DePIN #AIInfra #ERC8004 #MCP #x402

🛠️ DevLog – #ERC8004 Deeper Dive & Prerequisite Setup (Continuation, WIP) This week’s agentic exploration & experimental track is running in parallel: #MCP + #x402 Router experiments and #ERC8004 discovery. All are still exploratory, but together they're shaping what Router-Agent needs to become in later #Testnet phases. 🔹 #ERC8004 Progress - We're now able to register a mock agent with baseline attributes. - This validates the entry step for "Router-Agent as an ERC-8004 service," but it’s still just scaffolding. 🔹 What We Learned - #ERC8004 expects real, stable capability surfaces beyond registry writes. - Proper discovery will require prerequisite setup like DNS/HTTPS endpoints and structured comms metadata. 🔹 Prerequisites To Set Up Next - Stable DNS/HTTPS endpoints for at least one Router-Agent / session. - Clean capability ads that point to #MCP + #x402 routes. 🔹 Next Steps - Deeper review of #ERC8004 attributes and linked standards: #A2A, #OASF/OSSF taxonomies, #DID/#ENS, wallet + reputation fields. - Map how Router-Agent should publish #MCP + #x402 endpoints for trustable discovery and task validation. 🔹 Scope - Still research/prototyping to surface unknowns early; full alignment comes later in Testnet. #Cortensor #AI #DePIN #AIInfra #ERC8004 #MCP #x402

🛠️ DevLog – ERC-8004 / Agent0 SDK Exploration (Router Node Surface, WIP) Building on this week's #x402 + #MCP experiments on the Router Node, we spent time going deeper into ERC-8004 docs and the Agent0 SDK to understand what a "Cortensor agent" might look like on the Router surface. 🔹 What We Explored - Reviewed #ERC8004 + Agent0 SDK to map required agent attributes: identity, metadata fields, capability adverts, and DNS/HTTPS-style service endpoints an agent is expected to expose. - Registered a mock agent on Sepolia and minted its #ERC721 ownership token as a research pass: Agent registration tx: sepolia.etherscan.io/tx/0x4d7a3335b… ERC-8004 agent NFT: sepolia.etherscan.io/token/0x8004a6… 8004scan view: 8004scan.io/sepolia/agent/… 🔹 Early Takeaways for Router Node - A future Router / COR Prover / Corgent will likely need to expose: - Stable DNS/HTTPS endpoints for MCP + x402 surfaces - Properly structured #ERC8004 metadata (capabilities, tools, skills, x402 support) - Configurable fields so node ops can choose what gets advertised on-chain - This pass mainly surfaced direction + gaps: what we already have (#MCP/#x402 prototypes) vs what needs design (discovery metadata, reputation hooks, operator UX). 🔹 Scope & Next - Still firmly exploratory/research – the goal is to de-risk ERC-8004 integration before deeper #x402 + #ERC8004 work in later Testnet phases. - Immediate next step: set up stable DNS/HTTPS endpoints on at least one Router Node/session so it matches ERC-8004-style discovery expectations. - Then: try a minimal "Router-as-8004-agent" registration flow via Agent0 SDK, and iterate on how Router Node config/docs should look for agent discovery. #Cortensor #AI #DePIN #AIInfra #ERC8004 #MCP #x402 #Agentic

🧭 2026 Sneak Peek – Agentic Roadmap on Router Node (Priorities & Staging) We're outlining the agentic track for 2026, centered on evolving the Router Node into an agent-ready surface, then wiring it into #ERC8004 and Virtual. Early experiments are happening now; most of this solidifies across 2026 testnet/mainnet phases. 🔹 Router Node v1.5 – "Router Agent" (Node-Operated, 2026 Early) - #x402 on Router Node – per-router pay-per-call endpoints (e.g., completions) so node ops can offer paid inference services directly to Web2 clients and Web3 agents. - #MCP on Router Node – MCP server/proxy on top of existing REST, making the Router callable via MCP and HTTP. - Together, these form Router Node v1.5 ("Router Agent") – an agent-ready endpoint operated by node ops on top of $COR infra. 🔹 Validation & #ERC8004 Alignment (Mid 2026) - /validate endpoint – generalized task evaluation API using PoI/PoUW, where agents submit {input, output} for LLM-based scoring and reports. - #ERC8004 on Router Node – register services into the #ERC8004 agent ecosystem, expose validation artifacts, and offer agent task validation to ERC-8004-compliant agents. 🔹 Payments & Virtual – Optional #x402 at Core Level (Late 2026) - Core network rails remain Pay-Per-Use + Stake-to-Use for protocol-level inference and capacity. - #x402 at network level is optional / undecided – Router-level x402 may be sufficient unless protocol-wide flows add clear value. - Virtual integration / "Corgent" – #Virtual #ACP agents consuming Router Node /completions and /validate as execution + validation backends, positioning Cortensor as agentic-ready, verifiable AI infra for #Web2 users and #Web3 agents. Most of this will be iterated through Testnet Phases 3–4 and into 2026 mainnet milestones, as we harden the Router from a REST broker into a first-class agent surface. #Cortensor #AI #DePIN #AIInfra #AgenticInfra #x402 #MCP #ERC8004 #Virtual #ACP

When you take one look at the @EchelonMarket Ecosystem Map, what do you see? Aptos everywhere 🌐 As a Move-powered modular money market, read more about the moves the Echelon team has been making in this Aptos Ecosystem Spotlight Blog: aptosnetwork.com/currents/ecosy…

📌 Early WIP Draft: 2026 Technical Roadmap - Network Launch + Agentic AI Infrastructure We've published an early WIP draft for the 2026 roadmap. Still subject to change, but it outlines the direction we're moving toward next year. 🔹 Docs (WIP) docs.cortensor.network/roadmap/wip-dr… 🔹 2026 Core Direction - Cortensor evolves from Testnet-first infra into agentic-ready, verifiable AI infrastructure. - Three tracks move together: Router Agent surface, Virtual/ACP integration, and network launch. 🔹 Track 1: Agentic Router Evolution - Router Node → Router Agent (v1.5 → v2.0) with MCP + x402. - Expose /completions + /validate as first-class agent endpoints. - Turn PoI/PoUW into reusable validation primitives for agents and apps. 🔹 Track 2: Virtual / ACP Integration - Deploy Cortensor Agents inside Virtual where most Web3 demand lives. - Corgent becomes the default delegation + verification oracle for Virtual and ERC-8004 agents. - Route Web3 agent compute + validation through Virtual → Cortensor infra. 🔹 Track 3: Network Launch Milestones - Complete the 4-phase Testnet in Q1 → mid-Q2. - Ship Mainnet-Lite (Arbitrum L2 Alpha) in early Q3. - Launch Mainnet-Full (COR L3 Beta) in Q4. 🔹 Rough 2026 Timeline - Q1: Testnet Phases #1–2 → full E2E regression + Router v1.5 foundations. - Q2: Testnet Phases #3–4 → /validate + ERC-8004 alignment + COR Prover upgrades. - Mid-Q2 → early Q3: Mainnet prep → payments hardening, Router Agent specs, stability sweeps. - Q3: Mainnet-Lite + Virtual/ACP prototype → agentic compute begins flowing live. - Q4: Mainnet-Full on COR L3 → sovereign execution + verification layer. We'll keep tightening this draft during Testnet, but this is the current north star for 2026. #Cortensor #AI #DePIN #AIInfra #AgenticAI #Roadmap2026

📌 Early WIP Draft: 2026 Technical Roadmap - Network Launch + Agentic AI Infrastructure We've published an early WIP draft for the 2026 roadmap. Still subject to change, but it outlines the direction we're moving toward next year. 🔹 Docs (WIP) docs.cortensor.network/roadmap/wip-dr… 🔹 2026 Core Direction - Cortensor evolves from Testnet-first infra into agentic-ready, verifiable AI infrastructure. - Three tracks move together: Router Agent surface, Virtual/ACP integration, and network launch. 🔹 Track 1: Agentic Router Evolution - Router Node → Router Agent (v1.5 → v2.0) with MCP + x402. - Expose /completions + /validate as first-class agent endpoints. - Turn PoI/PoUW into reusable validation primitives for agents and apps. 🔹 Track 2: Virtual / ACP Integration - Deploy Cortensor Agents inside Virtual where most Web3 demand lives. - Corgent becomes the default delegation + verification oracle for Virtual and ERC-8004 agents. - Route Web3 agent compute + validation through Virtual → Cortensor infra. 🔹 Track 3: Network Launch Milestones - Complete the 4-phase Testnet in Q1 → mid-Q2. - Ship Mainnet-Lite (Arbitrum L2 Alpha) in early Q3. - Launch Mainnet-Full (COR L3 Beta) in Q4. 🔹 Rough 2026 Timeline - Q1: Testnet Phases #1–2 → full E2E regression + Router v1.5 foundations. - Q2: Testnet Phases #3–4 → /validate + ERC-8004 alignment + COR Prover upgrades. - Mid-Q2 → early Q3: Mainnet prep → payments hardening, Router Agent specs, stability sweeps. - Q3: Mainnet-Lite + Virtual/ACP prototype → agentic compute begins flowing live. - Q4: Mainnet-Full on COR L3 → sovereign execution + verification layer. We'll keep tightening this draft during Testnet, but this is the current north star for 2026. #Cortensor #AI #DePIN #AIInfra #AgenticAI #Roadmap2026

🧭 2026 Sneak Peek – Agentic Roadmap on Router Node (Priorities & Staging) We're outlining the agentic track for 2026, centered on evolving the Router Node into an agent-ready surface, then wiring it into #ERC8004 and Virtual. Early experiments are happening now; most of this solidifies across 2026 testnet/mainnet phases. 🔹 Router Node v1.5 – "Router Agent" (Node-Operated, 2026 Early) - #x402 on Router Node – per-router pay-per-call endpoints (e.g., completions) so node ops can offer paid inference services directly to Web2 clients and Web3 agents. - #MCP on Router Node – MCP server/proxy on top of existing REST, making the Router callable via MCP and HTTP. - Together, these form Router Node v1.5 ("Router Agent") – an agent-ready endpoint operated by node ops on top of $COR infra. 🔹 Validation & #ERC8004 Alignment (Mid 2026) - /validate endpoint – generalized task evaluation API using PoI/PoUW, where agents submit {input, output} for LLM-based scoring and reports. - #ERC8004 on Router Node – register services into the #ERC8004 agent ecosystem, expose validation artifacts, and offer agent task validation to ERC-8004-compliant agents. 🔹 Payments & Virtual – Optional #x402 at Core Level (Late 2026) - Core network rails remain Pay-Per-Use + Stake-to-Use for protocol-level inference and capacity. - #x402 at network level is optional / undecided – Router-level x402 may be sufficient unless protocol-wide flows add clear value. - Virtual integration / "Corgent" – #Virtual #ACP agents consuming Router Node /completions and /validate as execution + validation backends, positioning Cortensor as agentic-ready, verifiable AI infra for #Web2 users and #Web3 agents. Most of this will be iterated through Testnet Phases 3–4 and into 2026 mainnet milestones, as we harden the Router from a REST broker into a first-class agent surface. #Cortensor #AI #DePIN #AIInfra #AgenticInfra #x402 #MCP #ERC8004 #Virtual #ACP