🛠️ DevLog – SessionPayment Flow on Testnet L3 We've deployed the latest Session & Payment contracts on Testnet L3 (COR Rollup). 🔹 Continuing refinement of tightly coupled Session ↔ SessionPayment integration 🔹 E2E flow already verified on Testnet L2 (Arbitrum Sepolia) 🔹 Now

💡 Community Spotlight – Decentralized Prompting Helper Bot Built to support new AI users, this Telegram assistant teaches how to craft effective prompts – fully powered by Cortensor’s decentralized AI network. 📌 How It Works 🔹 Start with /start to access the usage guide 🔹

🛠️ DevLog – Testnet L3 (COR Rollup) Status The latest Session & Payment modules are live - technically, Testnet L3 (COR Rollup) is now working. 🔹 Core flow verified 🔹 Some abnormalities observed during credit flow 🔹 Further manual E2E + edge case testing continues today to

🛠️ DevLog – Fixing Equality Edge Case on L3 (COR Rollup) We found a bug where SessionPayment failed on L3 (COR Rollup) if a session balance was exactly equal to the required amount. ✔️ Worked on L2 (Arbitrum Sepolia) ✖️ Failed on L3 due to subtle EVM handling differences

🛠️ DevLog – NodeSpec → SessionConfig Design Today's focus: integrating NodeSpec into SessionConfig to pave the way for larger model support. This will be mostly a mock/design phase before full coding. 🔹 Use NodeSpec to right-size models at session start 🔹 Enable Session

🛠️ DevLog – NodeSpec Categorization & Dynamic Model Switching We're designing how NodeSpec and SessionConfig work together so users can run bigger models reliably. 🔹 Step 1 – Model Classes We group models by memory size: 0 = 6GB, 1 = 12GB, 2 = 24GB, 3 = 48GB 🔹 Step 2 –

🛠️ DevLog – NodeSpec Categorization Progress We're building the link between NodeSpec and Session config so sessions can correctly assign larger models to capable nodes. 🔹 Step 1 – Model Classes (6GB / 12GB / 24GB / 48GB) – ✅ Done 🔹 Step 2 – NodeSpec Attribute – adjusting

💡 Community Spotlight – Eureka Buddy Built to provide a safe, kid-friendly AI chat experience, Eureka Buddy is a Telegram bot designed for children under 12 – fully powered by Cortensor's decentralized AI network. 📌 How It Works 🔹 Choose an AI persona – Friendly Buddy,

🛠️ DevLog – NodeSpec Attribute & Session Integration Progress on linking NodeSpec with Session config for larger model support: ✅ Step 1 – Model Classes (6GB / 12GB / 24GB / 48GB) ✅ Step 2 – NodeSpec Attribute → category index now calculated on the fly (shown in dashboard,

⏳ 10 Days Left – Cortensor Hackathon #1 We're entering the final stretch - builders have 10 days left to ship and polish their apps. So far we've seen: 🔹 AI chat assistants w/ custom personas 🔹 Real-time market + sentiment bots 🔹 Web2 ↔ Web3 bridges powered by

🛠️ DevLog – NodeSpec Levels Recap We've introduced Spec Level, a categorization system that maps node memory capacity to the models it can support. 🔹 Level 0 (<6GB) – limited, only tiny models 🔹 Level 1 (6–12GB) – small / medium models 🔹 Level 2 (12–24GB) – standard medium

🛠️ DevLog – NodeSpec Memory Classification Update We've refined NodeSpec memory classification into clear spec levels with color-coded ranges + model capability mapping: ✅ Adjusted memory ranges (0 = <6GB … 8+ = ≥96GB) ✅ NodeSpec now stores a spec level integer directly,

🛠️ DevLog – NodeSpec Spec Level → Session Integration We'e advanced NodeSpec by mapping system and GPU memory into Spec Levels (0–8+), tied to ranges, colors, and model capability. ✅ NodeSpec now stores Spec Level directly ✅ Dashboard shows classification for easier node

📄 New Docs Update We've published 3 additions to Cortensor's technical docs - expanding on how the network scales, stays verifiable, and sustains itself: 🔹 Building Trustless AI – Proof of Inference (PoI) & Proof of Useful Work (PoUW) ensure inference is not just fast, but

🎉 Cortensor is now on Farcaster! We've just launched our account: farcaster.xyz/cortensor 🔹 Follow us for updates in your decentralized social feed 🔹 You can also buy & sell Cortensor directly through Farcaster Another step in making Cortensor accessible across

🛠️ DevLog – NodeSpec Memory Index & Dynamic Model Loading We've extended NodeSpec with a memory index that categorizes each node by system/GPU memory. ✅ cortensord reports memory index per node ✅ Dashboard shows classification for visibility ✅ cortensord now uses memory

🛠️ DevLog – NodeSpec Memory Index & Model Expansion While working on memory index & dynamic model loading, we've expanded the catalog to 12 models (incl. default): New Additional Models: - Qwen Qwen3 4B-Q8_0 - Google gemma-3 12B Q4_K_M - DeepSeek R1 Distill Qwen 14B Q4_K_M -

🛠️ DevLog – Expanded Models + Memory-Aware Sessions We've added 4 more models to the dashboard UI (total now 12 models), integrated into the NodeSpec memory index system. Nodes dynamically load only the models they can actually fit in memory at boot, ensuring reliability across

Monthly Node Rewards – August 2025 (Initial Batch Sent) We've just sent out the first round of $COR rewards to 30+ active nodes, as part of our monthly node reward program. 🔹 Rewards will be distributed in batches 🔹 More batches rolling out over the next few days 🔹