Cortensor (cortensor)

🧪 Testnet Phase #1 Starts Now The first of four structured Testnet phases is now live. We're entering a 1-week grace period: node stats and session data have been reset on Testnet-0 (Arbitrum Sepolia) and Testnet-1 ($COR Rollup L3), and this week serves as onboarding + warm-up before full regression. 🔹 Focus – System Integration & Regression - Validate all node roles and coordination flows end-to-end across apps, routing, miners, and validators - Stress-test both Pay-to-Use and Stake-to-Use inference flows under live conditions - Exercise the new NodePool refactor to generate large user-task & validator datasets for future SLA filters and inference-quality tuning 🔹 Incentivized Testing & Bug Bounty - Node ops: submit random sessions/tasks daily to simulate real usage - Manual + load testing throughout Phase #1 will be incentivized - Bugs and edge cases will qualify for bounty rewards - Detailed rules and amounts will be announced soon in Discord 🔹 Docs / Plan - Testnet Phase #1: https://docs.cortensor.network/community-and-ecosystem/community-testing/testnet-phase/testnet-phase-1 - Testnet Phases: https://docs.cortensor.network/community-and-ecosystem/community-testing/testnet-phase - Structured Plan: https://docs.cortensor.network/community-and-ecosystem/community-testing/testnet-phase-wip-draft/wip-draft-testnet-phases-and-focus-areas-structured-plan Phase #1 is about breaking and hardening the pipeline so later phases can safely introduce SLA-driven routing, payments, and agentic integrations.

• 0 replies
• 2 recasts
• 4 reactions

🗓️ Weekly Recap – Load Restart, Model Expansion, Dashboard Polish & Phase #1 Snapshot This week closed out Testnet Phase #1 with load testing back at scale, major model expansion completed, and final polish ahead of snapshot and Hackathon close. 🔹 Incentivized Testing & Manual Regression - Incentivized testing and manual regression continued with node ops and community testers. - Ongoing follow-ups on abnormal node states helped keep network behavior stable. 🔹 Automated Load Testing (Restarted) - Automated load testing successfully restarted on the new Router endpoint. - Total execution now at ~130k tasks, giving strong coverage going into the Phase #1 snapshot. 🔹 New Inference & Embedding Models Online - Inference models 47–55 are now live. - Embedding stack 56–64 deployed to support Validator v2/v3 and upcoming evaluation paths. - Model coverage expanded to 64 total, including additional embedding models. 🔹 /validate & /delegate Iterations (Light This Week) - Some progress continued, but iteration depth was lighter than planned this cycle. 🔹 Dashboard UI Refinements - Additional polish shipped across the dashboard. - The Node Reward / Performance page now better reflects incentives, performance metrics, and monitoring signals for node operators. 🔹 Hackathon #3 – Final Window - Hackathon #3 wraps in the next 24 hours, with final support and reviews in progress. 🔹 Testnet Phase #1 – Snapshot - Phase #1 snapshot scheduled within 24 hours, locking state for rewards, rankings, and next-phase planning. A strong close to Phase #1 - higher sustained load, expanded model coverage, improved operator UX, and clean state heading into the next Testnet phase.

• 2 replies
• 2 recasts
• 2 reactions

🛠️ DevLog – Node Performance/Rewards View, Refined We’ve tightened up the Node Performance page so node operators can see both performance and estimated rewards in one place. 🔹 NodePerf now shows three reward scenarios side by side – Flat base, Tiered base, and Config-based – including totals in USD and COR. 🔹 The Config model dropdown lets you tweak the $/level parameter and instantly see how it changes per-node and total rewards. 🔹 The overview bar now surfaces average reward per eligible node for each scenario, so operators can quickly sanity-check what a solid node should roughly earn. 🔹 Live on both networks: Testnet-0: https://dashboard-testnet0.cortensor.network/node?tab=nodeperf Testnet-1: https://dashboard-testnet1.cortensor.network/node?tab=nodeperf This should make reward modeling and monthly expectations much clearer without needing external sheets or manual math.

• 0 replies
• 0 recasts
• 0 reactions

🛠️ DevLog – Ollama Engine Status Check (Ephemeral + Dedicated) Quick follow-up on the #Ollama path: most of the critical code paths for both dedicated and ephemeral nodes now run end-to-end with the new engine, including LLM Gateway enforcement. A few miner-side quirks + GPU enablement remain. 🔹 Looks stable enough to poke at - Task routing → container spin-up → Ollama inference → result back to router works on both node types. - #OpenAI #OSS 20B is responding reliably in CPU tests. 🔹 What's left (focused) - Run /validate on real sessions with this model and compare judgment quality vs existing models. - Track down remaining GPU issues on Ollama images (device config, concurrency, memory behavior). Once those are understood + regression-tested, we'll decide how/where to expose these larger models in the network.

• 2 replies
• 4 recasts
• 5 reactions

💡 Community Spotlight – Cortensor Price Analyzer One of the entries in Hackathon #2: An intelligent web app that delivers AI-assisted market insights across equities, ETFs, cryptocurrencies, forex pairs, and core commodities – all powered by Cortensor’s decentralized AI network. It brings together resilient data fallbacks, automated analytics, and Cortensor-generated narratives so traders can move from idea to action in one dashboard. 📌 How It Works 🔹 Select an Asset – Choose from equities, ETFs, crypto pairs, forex, or commodities. 🔹 Automated Data Fetch – The system cascades through multiple sources (Alpha Vantage, Stooq, TwelveData, CoinGecko) for reliable coverage. 🔹 AI-Powered Analysis – Cortensor miners generate narratives, highlight technicals, and summarize fundamentals. 🔹 Insight Delivery – View actionable summaries with risks, opportunities, and market sentiment. 🔹 Save & Compare – Store past analyses for quick review and performance tracking. Core Features: 🔹 Multi-Asset Coverage – Analyze equities, ETFs, crypto, forex, and commodities in one view 🔹 Resilient Data Fallbacks – Smart failover across top market-data providers 🔹 AI Narratives & Insights – Contextual reports with key takeaways and horizon guidance 🔹 Technical & Fundamental Highlights – RSI, MACD, SMA, valuations, margins, and more 🔹 Headline Sentiment – Connects recent news flow to price action 🔹 Saved Analysis Library – Search, reload, and compare previous scans 🔹 Responsive UI & Charting – Clean, real-time visualizations across devices Why It Matters: By decentralizing market analysis, the Cortensor Price Analyzer combines human-level reasoning with transparent AI inference – empowering traders, analysts, and investors to understand markets faster and act smarter. 🔗 Try it here: https://cortensor-price-analyzer.eureka7.online/ Powered by decentralized inference via Cortensor Share feedback: https://discord.gg/cortensor

• 2 replies
• 2 recasts
• 4 reactions

Monthly Node Rewards – October 2025 (Batch 2 Sent) Another 30 active nodes have just received their $COR rewards - bringing the total to 60 rewarded nodes so far this month. 🔹Rewards will be distributed in batches 🔹More batches rolling out over the next few days 🔹Recognizing uptime, stability & contribution Join our Discord to set up a node or get involved: https://discord.gg/cortensor Thanks to all node operators pushing decentralized AI forward - it's $COR or nothing.

• 2 replies
• 2 recasts
• 4 reactions

🛠️ DevLog – /validate Re-Iteration & Gradual Ollama Rollout 🔹 /validate focus today Now that the dedicated GPT-OSS 20B & 120B sessions are online, today’s work is re-iterating on /validate — refining prompts, scoring schemas, and comparing evaluations against our existing Llama/Llamafile stack. 🔹 Dashboard model rollout plan - First step: expose Model 12 (gemma3:270m-gpu) on the dashboard for ephemeral nodes. - If it behaves well under real traffic, we’ll gradually enable the remaining Ollama models 13–20 for ephemeral usage. 🔹 Dedicated-node availability All new Ollama-backed models (12–20) are already usable via dedicated-node sessions for anyone experimenting more aggressively while we keep the dashboard exposure conservative.

• 3 replies
• 4 recasts
• 4 reactions