Filter
Exclude
Time range
-
Near
8 Dec 2025
Low-severity findings in ncameiri/validate_protocol Alongside the high-risk issues, our system also flagged multiple low-severity items that affect code quality: • Unused variables (params, start) • Unused async promises • Unused requestCache map • Unused nodeStats map • Unused constant values (CACHE_TTL_MS) Individually small, but together they add noise, reduce efficiency, and make future debugging harder. Sharing these for full transparency. More updates to follow.
6 Dec 2025
More findings from our analysis of ncameiri/validate_protocol Our system flagged two additional security issues in the current MVP build. 1. Secret exposure A private key is imported directly from ./keys.ts. If this file ends up in git history or a public repo, the key becomes exposed to anyone. Developers should move secrets into environment variables and avoid shipping sensitive values in source code. 2. Missing input validation The handler accepts any JSON body without checking or validating its structure. A malicious payload could break the parser or cause downstream failures when forwarded to a Solana RPC node. Simple schema checks would prevent this.
2
2
9
1,045
4 Dec 2025
High Severity Issue 2 — ncameiri/validate_protocol Our system identified a non-functional health check mechanism in the repository. checkNodeHealth() creates an array of Promises, but none of them are awaited or executed. This means health checks never run, nodeStats stays empty, and the entire monitoring flow fails. This issue was detected during our automated scan of the repository.
3 Dec 2025
High Severity Issue 1 — ncameiri/validate_protocol Our system flagged a POST handler that never completes its response cycle. Every request hangs indefinitely, which can lead to memory leaks and possible service freeze. This issue was detected during our automated scan of the repository.
4
4
13
2,137
23 Nov 2025
@GypsyRickyRich Even if LinkedIn's "meaningful" ( i have no interest to even look at it ), code doesn't lie. Repo= no new commit, 15h dead and still garbage. nodeStats empty. POST eats reqs, spits 200. Zero txs routed. Facts > resumes DYOR
1
223
29 Sep 2025
🛠️ DevLog – Validator v2 PoC: Multi-Node PoI Testing Progress on Proof of Inference (PoI) prototyping continues - now with multiple Validator v2 nodes running in parallel. 🔹 Current Status - At least two Validator v2 nodes are live as experimental at DevNet7, running PoI validation. - Each validator generates embeddings, clusters miner outputs, and performs anomaly detection. - In addition to detailed clustering outputs, validators now emit a binary decision (normal vs anomalous miner). - Example: session task analysis → cluster → anomaly detection → binary abnormality flag. 🔹 Quantitative Stats Prototype - For now, schema mirrors NodeStats counters: normal → 1 point abnormal → -1 point - Each session/task increments counters. Over time, repeated anomalies will surface unreliable nodes. 🔹 Next Steps - Refine scoring beyond binary flags → integrate cluster size, isolation ratio, and consistency metrics. - Review and adapt Quantitative Stats schema for multiple validator reports (ensuring results are properly recorded). 👉 This is the first step: basic multi-validator PoI checks are live, with Quantitative Stats storing early signals. Schema and scoring refinements will follow as we are experimenting #Cortensor #AI #DePIN #DevLog #Validator #PoI
29 Sep 2025
🗓️ Weekly Focus – Gateway Stability, PoI Prototyping & Model Match Testing 🔹 LLM Worker Gateway Monitoring - Continue monitoring rollout for stability. - Address edge cases flagged by node ops with fixes scheduled this week. - Focus on preventing regressions before rollout to Testnet-0 and Testnet-1 environments. 🔹 Validator Expansion (PoI Prototyping) - Extend contract-side data storage for PoI scoring & aggregation. - Run Validator v2 PoC across CPU & GPU nodes to test performance & edge cases. - Begin multi-node tests: validators will submit reports into Quantitative Stats → later aggregated in Quantitative Metrics to surface anomalies. 🔹 Model Switch & Matching Validation - Conduct further debugging and testing since node pool selection still mismatches unsupported models. - Deploy additional debug hooks this week to track errors. - Perform manual validation to ensure accuracy. 🔹 NodeSpec Extension - Extend NodeSpec to show currently running model for easier debugging and ops visibility. - Ship updates in the next binary release alongside Gateway fixes. 🔹 Hackathon #2 Support - Continue supporting builders with agentic app ideas & integrations. - dorahacks.io/hackathon/1488 👉 This week: stabilize Gateway rollout, expand PoI prototyping, validate model switch/index match, extend NodeSpec, and keep momentum for Hackathon #2. #Cortensor #AI #DePIN #WeeklyFocus
6
11
28
630
28 Sep 2025
🛠️ DevLog – Node Level Heatmap Bug Fix One of our node ops flagged an inconsistency between the Node Level heatmap and the node-level table. After review, we found the heatmap was pulling from the wrong dataset. 🔹 Fix - Heatmap now uses the correct indexer aggregation. - Numbers between the Node Level Table and Heatmap should match. - Minor differences may still appear due to caching/refresh lag. 🔹 Check It Out 📊 Heatmap: dashboard-devnet7.cortensor.… 📋 Node Table: dashboard-devnet7.cortensor.… Backend Summary: db-be-7.cortensor.network/al… 👉 Already live on DevNet7 - thanks to the community for surfacing this. #Cortensor #AI #DePIN #DevLog #NodeStats
5
9
29
576
19 Sep 2025
Yo, what's good, fam? 🔥 New banger listing on #Klingex, $NS is poppin’! 📈 #NodeStats, let’s get those views LIT! 🚀 Smash that like, follow for more! 😎 #KlingEXVibes klingex.io/trade/NS-USDT
1
5
326
17 Sep 2025
!! COIN LISTING !! It is my pleasure to announce, that C-NOTE Coin ( CNOTE ) has decided to support NodeStats, and list their coin. CNOTE is also present in the masternode hosting/monitoring system. Watch all the data here: nodestats.info/default.php?c…
3
6
6
55
🚀 TicsForge Node Update Yesterday we reached a major milestone: over 5 million $TICS staked 🎉 We are now ranked #2 in voting power and continue to achieve incredible results. 🔹 We’re proud to be the first validator with a fully detailed and transparent website, where delegators can track real-time node statistics. And this is just the beginning: many more features are coming soon! 🌐 Visit the site 🔹 Current server status: operating at just 5% RAM usage, delivering maximum reliability with minimal effort. 🎁 Airdrop update: once we reach 10 million TICS staked, the Saturday weekly airdrop will double from 100 to 200 TICS. And the community will decide how to distribute it through a poll. 🙏 Thank you all for helping us reach this first milestone together. Please share and recommend our node to your friends — the stronger our community, the stronger the network. Stay tuned — the best is yet to come! 💎 #TicsForgeNode #Qubetics #TICS #CryptoStaking #StakingRewards #Delegators #Blockchain #Validator #CryptoCommunity #StakingCrypto #Decentralization #CryptoEcosystem #CryptoUpdate #Milestone #NodeStats #CryptoGrowth #StakingPower
2
144
Replying to @JoelKatz
Looking solid, Joel! 🚀 The peer stats show steady growth and stability—perfect sign for production readiness. Excited to see your hub go live next week! 💪 #XRPLedger #NodeStats
1
104
24 Jul 2025
Swampcoin add on MasterNode Hosting Provider - Nodestats nodestats.info The listing price is $1 per masternode per month. Paid in any major currency.
1
7
15 Jul 2025
🛠️ DevLog – CognitiveLevel Iteration: From Stats to Smart Routing CognitiveLevel iteration continues on DevNet6, shaping how task performance data powers adaptive coordination. 🔹 Progress So Far - Oracles and contracts are being updated to evaluate miner output post-session. - Metrics like token size, hash match, and task success are now recorded. - These results feed into CognitiveLevel, just like how Cognitive updates NodeStats. 🔹 How It Works - CognitiveLevel tracks success rates and streaks per node. - Oracles report these after each session. - Node levels auto-adjust and inform the Cognitive module, which adapts output targets. - Oracles then verify outputs in the next task - forming a full feedback loop. 🔹 Why It Matters - Enables level-aware routing and task scaling. - Supports SLA enforcement, smarter incentives, and heterogeneous node usage. - Lays groundwork for multi-model support and agent coordination. Still early in testing - but this unlocks the path toward smarter, more efficient decentralized inference. #Cortensor #DevLog #CognitiveLevel #DevNet6 #DePIN #PoUW #AIInfra
13 Jul 2025
🛠️ DevLog – CognitiveLevel Iteration Begins on DevNet6 The CognitiveLevel module is now deployed to DevNet6, and we're kicking off real iteration. 🔹 Why It Matters Designed to classify nodes based on performance (success streaks, token output, reliability), CognitiveLevel enables smarter, level-aware task routing across the network. 🔹 Current Status - Initial contract deployed - Integration hooks are in place between miners, oracles, and Cognitive - Testing flow: miner → oracle → CognitiveLevel → dynamic prompt selection 🔹 Next Steps - Unit tests and integration testing across full pipeline - Expand oracle tracking for per-task token metrics - Begin miner-side prompt logic tied to node level 🔹 Looking Ahead This will be a multi-phase iteration cycle, requiring extensive testing across all existing network task prompts. Each prompt must be validated to ensure proper behavior under different CognitiveLevel tiers. It's a deep refactor but one that unlocks adaptive routing, tiered incentives, and more precise decentralized execution. #Cortensor #DevLog #CognitiveLevel #DevNet6 #AIInfra #DePIN #PoUW #TaskScheduling
3
6
15
868
9 Jul 2025
🛠️ DevLog – Rough Design: CognitiveLevel Module for Smarter Network Task Evaluation We've drafted the initial version of the CognitiveLevel contract - a new module that manages performance-based levels for nodes handling network tasks, modeled after patterns used in NodeStats and NodeReputation. 🔹 What It Does - Tracks per-node task outcomes: success/failure counts, streaks, and average token size - Oracle nodes can upgrade/downgrade a node’s level based on historical performance 🔹 How It Fits - The Cognitive module (which manages network-level tasks) reports completion to CognitiveLevel - Oracle nodes evaluate each miner's task result, including token length, validity, and latency - On each new task, Cognitive queries CognitiveLevel to dynamically set prompt complexity based on each node's current level - Because oracle nodes are sharded, evaluations can be distributed for scalability 🔹 Why This Matters - Enables precision orchestration across a heterogeneous network (from Pi to H100) - Ensures high-load or mission-critical tasks go to high-reliability nodes - Forms the backbone for long-term task scheduling, tiered incentives, and trustworthy decentralized inference 🔹 What's Next - Will require significant updates to oracle and miner codebases - Oracle nodes must track token counts for every task to verify level eligibility - Miner logic will need programmable prompt templates that adjust with each node's assigned level This is still an early brainstorm/PoC, but it unlocks a smarter, more adaptive Cortensor network. We'll keep refining design through this and the next phase. #Cortensor #DevLog #CognitiveLevel #NetworkTasks #AdaptiveInfra #DecentralizedAI #DePIN
8 Jul 2025
🛠️ DevLog – Designing CognitiveLevel for Smarter Network Task Routing We're in the design phase of a new module called CognitiveLevel, which manages capability-aware node selection for network tasks - similar to how NodePool manages ephemeral nodes for user tasks. 🔹 Core Concept - CognitiveLevel tracks each node’s abstracted performance tier (e.g. CPU/GPU class, reliability) - Only the Oracle can upgrade or downgrade node levels Scoring is based on recent network task samples (accuracy, response time, etc.) 🔹 How It Fits - The Cognitive module powers the network task layer - it evaluates and routes user jobs before they are passed into user task workflows - create, prepare, and precommit steps will query CognitiveLevel to ensure only eligible nodes are selected 🔹 Why It Matters - Just as NodePool filters nodes for user task session reliability, CognitiveLevel enables performance-aware orchestration at the network layer - Ensures heavier AI workloads land on proven, high-capacity nodes Still early-stage, but sets the stage for trustable and scalable AI inference. #Cortensor #DevLog #CognitiveLevel #NetworkTasks #NodeRanking #DePIN #DecentralizedAI #AIInfra
3
8
18
768
30 Jun 2025
!! COIN LISTING !! It is my pleasure to announce, that WeeCash Coin ( WCH ) has decided to support NodeStats, and list their coin. WCH is also present in the masternode hosting/monitoring system. Watch all the data here: nodestats.info/default.php?c…
1
1
1
45
27 Jun 2025
🛠️ DevLog – Dashboard Backend Refactor: 3-Part Architecture As the Dashboard backend grows to handle deeper Web3 activity and real-time monitoring, we've refactored it into three focused components: 🔹 1. Event Listener - Listens to key onchain events (NodeStats, NodeReputation, SessionStats, SessionReputation, NodePool, Session, SessionQueue) in real time 🔹 2. Indexer - Structures and processes event data - Generates Redis-cached time-series datasets for efficient tracking 🔹 3. API Service Layer - Serves structured data via RESTful endpoints - Powers the dashboard UI and external tools This modular split improves performance, scalability, and observability. 🧩 Next Steps This foundation allows us to expand indexing and time-series tracking - with upcoming work focused on deeper coverage of Session and SessionQueue activity. #Cortensor #DevLog #Dashboard #Infra #Refactor #Monitoring #DePIN #AIInfra
27 Jun 2025
🛠️ DevLog – Monitoring Update & Upcoming User Task Stats So far, the newly deployed SessionStats, SessionReputation, and dynamic node pool logic are holding up well. We'll continue observing through mini hackathon activity to surface more edge cases and real-world bugs. 🔹 What’s Next: We're planning to introduce a User Task Stats dashboard - similar to the current network stats page, but scoped to user-generated sessions only. 📊 Planned Time Series Metrics: - Ack / Precommit / Commit - Precommit-to-commit ratio - Avg & max output length per session - Avg Max token size set by developers - Avg Task duration (end-to-end) - LLM parameters set (temp, top_p, top_k, penalties) This will provide deeper visibility into app-level inference behavior, helping developers tune performance and spot issues faster. More to come later today as we explore implementation. #Cortensor #DevLog #Observability #SessionStats #DePIN #AIInfra
1
7
16
993
26 Jun 2025
🛠️ DevLog – Session Stats & Reputation Deployed Stat & Reputation module separation is now fully live across DevNet5 and the dashboard backend. 🔹 What’s Deployed: - New SessionStats & SessionReputation modules for user task tracking - Dashboard backend updated cache reset - Monitoring now focused on ack, precommit, and commit events in sessions - Just in time for the mini hackathon - to test and validate via real usage This enables cleaner tracking, better debugging, and more accurate scoring as the network scales. #Cortensor #DevLog #SessionStats #NodeStats #Hackathon #DePIN #AIInfra
26 Jun 2025
🛠️ DevLog – Stat Separation Now Implemented We've implemented the planned stat separation - and testing begins on DevNet5. 🔹 What's Live: - Dual tracking via NodeStats new SessionStats for session queue activity - Once validated, we’ll fully switch to SessionStats and SessionReputation for user task metrics This allows us to isolate, reset, and analyze user-level stats without affecting network-level data - a key step toward fairer tracking and future reward scoring. #Cortensor #DevLog #SessionStats #NodeStats #Phase5 #AIInfra #DePIN
1
7
21
896
26 Jun 2025
🛠️ DevLog – Stat Separation Now Implemented We've implemented the planned stat separation - and testing begins on DevNet5. 🔹 What's Live: - Dual tracking via NodeStats new SessionStats for session queue activity - Once validated, we’ll fully switch to SessionStats and SessionReputation for user task metrics This allows us to isolate, reset, and analyze user-level stats without affecting network-level data - a key step toward fairer tracking and future reward scoring. #Cortensor #DevLog #SessionStats #NodeStats #Phase5 #AIInfra #DePIN
25 Jun 2025
🛠️ DevLog – Stat Separation Planned for Phase #5 As Phase #4 continues, we're preparing a key upgrade for Phase #5: separating network-level and user-level stats for better accuracy and control. 🔹 Current Limitation: - NodeStats and NodeReputation are shared between network tasks and user sessions. - This prevents resetting session metrics without affecting overall network data. - Some early session stats are skewed - but can't be cleanly corrected. 🔹 Coming in Phase #5: - NodeStats / NodeReputation: for network-level tracking - SessionStats / SessionReputation: for user inference sessions 🔹 Why It Matters: - Enables selective resets of session metrics - Preserves long-term network stats - Improves tracking, debugging, and reward scoring A necessary step for deeper observability and network fairness as we grow. #Cortensor #DevLog #Phase5 #Stats #Reputation #DePIN #AIInfra
2
6
16
862
25 Jun 2025
🛠️ DevLog – Stat Separation Planned for Phase #5 As Phase #4 continues, we're preparing a key upgrade for Phase #5: separating network-level and user-level stats for better accuracy and control. 🔹 Current Limitation: - NodeStats and NodeReputation are shared between network tasks and user sessions. - This prevents resetting session metrics without affecting overall network data. - Some early session stats are skewed - but can't be cleanly corrected. 🔹 Coming in Phase #5: - NodeStats / NodeReputation: for network-level tracking - SessionStats / SessionReputation: for user inference sessions 🔹 Why It Matters: - Enables selective resets of session metrics - Preserves long-term network stats - Improves tracking, debugging, and reward scoring A necessary step for deeper observability and network fairness as we grow. #Cortensor #DevLog #Phase5 #Stats #Reputation #DePIN #AIInfra
3
6
11
609
27 May 2025
🛠️ DevLog: Oracle V3 – Foundation for Distributed Validators (2/2) With user tasks being independently triggered and running in parallel, validating them is like finding a needle in a haystack. Oracle V3 handles this by: 🔹 Logging every validation state per task/session 🔹 Updating NodeStats & Reputation based on penalties 🔹 Enabling rewards for validators who correctly validate async tasks 🔹 Laying the groundwork for a shared, decentralized validator pool While commit state is already verified by miners, Oracle V3 ensures ack/precommit are independently verified – a key step toward trustless, scalable user task flow. #Cortensor #OracleV3 #ValidatorLayer #UserTasks #DePIN #DevLog
2
3
8
219
26 May 2025
🛠️ DevLog: Oracle V3 - Enforcing User Task Signals for Penalties So far, user tasks (ack, precommit, commit) have been tracked primarily for reward signals - helping build up NodeStats and Reputation for positive contributions. Next step: we’re developing Oracle V3 to introduce penalty tracking for user task failures. 🔹 What Oracle V3 Will Do • Penalize missed or invalid ack, precommit, and commit states • Extend scoring logic from network tasks to user tasks • Feed directly into NodeStats & NodeReputation • Enable stronger filtering and miner quality assessment 🔹 Architecture • Built as a standalone module (Oracle V3 Agent) • Enables isolated testing before future V4 unification This will bring user task enforcement on par with network tasks - critical for trust and quality as we scale. #Cortensor #DevNet5 #OracleV3 #UserTasks #NodeQuality #AIExecution
25 May 2025
🛠️ DevLog: Week’s Focus – Session Iteration, Oracle V3, and DevNet 6 Prep Here’s what we’re focused on this week across the Cortensor stack: 🔹 1. Session Iteration via SessionStats • Refining UI and backend logic using live session metrics • Improving task visibility and orchestration insights 🔹 2. Oracle V3 (User Task Regulator) • Like Miner V2 was for network tasks, Oracle V3 is designed for user tasks only • Handles ack, precommit, and commit timers • Regulates multi-miner behavior on shared sessions • Paves the way for Oracle V4 (unified task orchestration) 🔹 3. User Task Stress Test • Running another test cycle with improved Node Pool dynamic capacity logic • Helps benchmark real-time responsiveness and scaling 🔹 4. DevNet6 Preparation • Wrapping up DevNet5 and shifting it toward Phase #4 readiness • DevNet6 will become the main iteration network while DevNet5 hardens for node ops 🔹 5. Router Node API Iteration • Polishing all existing Router Node APIs • Ensuring API key-based authentication is in place (verification step) • Stabilizing endpoint behaviors 🔹 6. Ongoing L3 AppChain Testing • Continuing experimental L3 setups • Simulating both network & user task execution under L3 rollup environments We’re moving toward a more regulated, responsive, and scalable decentralized AI stack - one layer at a time. #Cortensor #DevNet5 #DevNet6 #OracleV3 #SessionStats #DePIN
1
5
22
1,268
Replying to @cortensor
great progress, the integration of nodestats and dynamic capacity planning is a huge step forward, excited to see how realtime utilization improves efficiency #cortensor #devnet5 #nodepool #web3
1
17