Praxis has officially launched on Base.
This is a major expansion moment for us. Bringing our decentralized AI agent mesh to a larger ecosystem, more users, and new partners building the future of trustless AI.
Base is one of the most vibrant ecosystems in crypto: a thriving builder community and a rapidly growing network of functional on-chain agents.
It’s the perfect home for scaling a decentralized AI mesh.
Now Praxis agents can connect, collaborate, and monetize across an even larger universe.
Contract: 0xC44f97feCFfEA3b22A15Ed04f823c7d6c5d34d1A
Nothing changes for current holders.
No supply changes. No action needed.
Just a bigger stage for what comes next.
Welcome to Base. 💙🔵
The next phase of AI won’t be owned by platforms..
It will be built by agents that run locally, coordinate trustlessly, and evolve together without central control.
Praxis is pushing this forward with a decentralized agent mesh:
• self-sovereign identity for every agent
• zero-knowledge privacy for users
• peer-to-peer collaboration
• open monetization rails for contributors
The goal is simple:
AI that anyone can run, extend, audit, and trust without needing to ask permission from an API or a company.
We’re moving fast. New integrations, new partners, and a wave of agents to be launched.
The agent economy is coming.
Praxis is making sure it’s decentralized.
🛠️ DevLog – Swappable Validation Prompts for /validate (WIP)
We've wired /validate to support multiple LLM-judge prompt templates, so we can iterate faster on evaluation behavior.
🔹 What's new
- /validate now accepts a validation prompt type (e.g. prompt_v1, prompt_v2), routing the same {task_request_input, task_result_output, context} through different rubrics.
- Outputs are nudged toward stricter JSON structure (scores, issues, summary, recommendation) to reduce post-processing hacks.
🔹 Why it matters
- Lets us A/B test short vs long/example-heavy rubrics on real tasks (ranking reports now, Virtual ecosystem samples later).
- Creates a future knob so agents (incl. Corgent / Virtual agents) can choose validation style/strictness per request.
🔹 Next
- Add more prompt variants and refine schemas, aiming for stable, precise JSON suitable for Corgent’s delegation / validation / arbitration tiers.
#Cortensor #DevLog #LLMValidation #CORProver #Corgent #AgenticAI
🛠️ DevLog – Dashboard Support for Off-Chain Inference Data (WIP)
Quick lower-hanging update: the dashboard now understands off-chain prompt URNs coming from the new /validate path.
🔹 What's live now
- When a task’s data field contains a urn:blob:... style value, the dashboard:
- Detects it as off-chain content.
- Fetches it from the configured object storage.
- Shows the resolved prompt inline under "Resolved Offchain Content."
This makes long validation prompts (rubrics, examples, specs) readable again instead of opaque URNs.
🔹 What's next
- We'll first extend this off-chain path for long-context inference via the Router / Web2 API (router node as entrypoint).
- Web3 SDK will stay as-is for now – no off-chain payload support yet, so long prompts still need to go through the router/web2 side.
This keeps /validate usable for large eval prompts today and lays the groundwork for broader long-context support through the router node.
#Cortensor #AI #DePIN #DevLog #OffchainData #Inference
USST is engineered as the core infrastructure for the on-chain economy, serving as the indispensable interoperable asset.
Ecosystem specific Stablecoins (ESS): USST provides the blueprint for ecosystems to launch their own white-labeled stablecoins. Minters can create custom stablecoins for utility across their ecosystem.
Cross-Chain Interoperability: Mint on one chain and bridge seamlessly across others. Integrations with Wormhole & CCIP allow minters to use collateral on chains like ETH and move it to cost-efficient chains, reducing liquidity fragmentation.
Wide Collateral Options: Mint USST using publicly available assets such as USDY, USDG (upcoming), OUSG, BUIDL, and BENJI (upcoming).
Permissionless Burning & Instant Liquidity: Unlock collateral instantly when providing YLD, with partial burn support. Upcoming Converters will allow liquidity providers to facilitate burning for others, accepting swapped USST & YLD for other stablecoins.
🌑 SHADOW WALLET IS LIVE 🌑
The first Zcash wallet on Telegram is here.
━━━━━━━━━━━━━━━
✅ What's included:
🔐 Unified Addresses - one address for all ZEC
📤 Instant Send - to any t-addr or unified address
🛡️ Shielded by Default - true privacy, always
Why ERC-8004 + x402 matter?
Because they enable the world MaximizeAI is building:
AI that can connect, transact, and operate across Bittensor ↔ EVM… with zero friction.
No-Code. Accessible. Interoperable.
That’s the $MAXI mission.
🔗 https://t.co/l4U2dAv1ds
- ERC-8004 gives agents an identity.
- x402 gives them the ability to transact.
- MaximizeAI gives everyone the tools to use them, no CLI, no complexity.
This is the foundation of a unified AI Hub bridging Bittensor and the EVM world.
#MaximizeAI #ERC8004 #x402 #Bittensor #T367 #AIhub #MaxiScreen
The next wave of agents won’t be “tools.”
They’ll be networks.
ERC-8004 gives them identity and reputation.
x402 gives them native access and payments.
Praxis gives them the mesh to work together, discovering peers, forming rooms, and coordinating tasks in real time.
When agents can trust, pay, and collaborate, entire industries start to rewire themselves.
The agentic economy isn’t coming.
It’s starting to self-assemble.
PRAXIS
🛠️ DevLog – Delegation / Validation / Arbitration on Completions for Corgent (Design WIP)
Quick clarification on scope: the refactored completion stack (including the x402-protected completions) is the core engine for all three Corgent surfaces: /delegation, /validation, and /arbitration.
🔹 Shared Foundation – x402 Completion Rails
- Both /completions and /api/v1/x402/completions/* are the base execution path.
- Corgent endpoints don’t build a new engine – they orchestrate around this path with different policies and schemas.
🔹 /delegation – Compute Wrapper on Completions
- Thin wrapper around existing completion APIs (plain or x402-protected).
- Policy chooses model, redundancy tier, and SLA; endpoint just returns the best/consensus result.
- This is the "Delegation-as-a-Service" path: "Run this task on Cortensor with reliability tier X."
🔹 /validation – Structured Verdict on a Result
- Uses the same completion engine, but the payload is richer:
- original_input, result_output, task_type, model_policy
- acceptance_criteria / spec (schema, rules, rubric)
- evaluation_mode (e.g. poi_rerun, consensus, spec_only, llm_judge)
- provenance (agent_id, miner_ids, run_hash / commit–reveal, timestamp)
- optional context / reference_data
- Response is a normalized verdict schema:
- binary_classification (valid/invalid + confidence)
- overall_assessment (score, status, risk_level)
- detailed_scores per dimension (accuracy, relevance, completeness, clarity, safety, etc.)
- issues_found, strengths, risk_factors, and a concise summary.recommendation.
- This is the "Validation-as-a-Service" path: "Is this result correct/useful under this spec?"
🔹 /arbitration – Oracle-Grade Re-Check on Disputes
- Reuses the same request/response schema as /validate, but with stricter policies:
- Higher redundancy tiers (e.g. 5 miners), stricter PoI/PoUW thresholds.
- Explicit dispute context (buyer/seller, claim vs. oracle consensus).
- Output = a validation verdict that’s treated as binding for settlement flows.
- This is the "Arbitration-as-a-Service" layer: "Resolve a dispute with oracle-grade replay + verdict."
🔹 Status & Next
- /delegation remains the simplest: direct wrapper on completions/x402.
- /validation and /arbitration are where the richer input/output structures live and where we’re focusing current design/prototyping.
#Cortensor #DevLog #Corgent #x402 #LLMValidation #PoI #PoUW #AgenticAI #AI #DePIN #AIInfra
Coinbase: “x402 is the trojan horse to global crypto adoption”.
Here to change the internet and with that, the world.
@Bankless Summit wrapping up, another amazing day.
PRAXIS
I’ll tell you here and now..
$COR is cooking, and once the full @virtuals_io agent ecosystem is connected with @cortensor, things are going to move fast
The pieces are falling into place: 🧩
Router upgrades, x402, ERC-8004 alignment, and the Virtual integration are all converging to build a real, scalable agent economy.
Still early, still undervalued imo.
NFA/DYOR.
We are on our way to Argentina! DevConnect 2025, time to take over.
We’re heading into a packed Devcon week with meetings across the ecosystem.
What's lined up? Tier-1 blockchains, the Ethereum Foundation, the ERC-8004 authors, and every team pushing trustless agents forward.
And we close the week on Friday, Nov 21st, at Trustless Agent Day.
We are on our way to Argentina! Devconnect 2025, time to take over.
LFG
PRAXIS
Everything is looking good for $COR.
Nearly all nodes and applications have been migrated to Testnet-0 and Testnet-1.
We’re now one step closer to launching the full testnet, but first, we’ll go through several phases before reaching the endpoint.
Stay tuned, Cortisians. I’ll walk you through this..
‣ ETH: https://t.co/xucQo1UzgG
‣ BASE: https://t.co/ebcCPvBRM0
🛠️ DevLog – NodePool Modular Refactor: Live on Testnet-0 (Regression Cycle)
The modular NodePool refactor has now been deployed on Testnet-0. This version may still encounter breakages, but it provides the foundation we need to perform deeper regression and concurrency testing across real node activity.
🔹 Deployment Context
- NodePoolData, NodePool, and NodePoolUtil are now fully isolated and functional as independent modules.
- Early validation shows proper state persistence and logic separation, but ephemeral node flows (reserve/release) still require further review under load.
🔹 Next Steps
- Run extended regression on Testnet-0 over the next few days to identify residual state race or edge-case issues.
- If stable, roll out to Testnet-1 for broader validation with live node operator traffic.
- Continue logging gas consumption and prepare optimization passes post-stability.
This stage marks the transition from local regression to network-scale verification, setting the groundwork for richer SLA filters and scoring logic in the upcoming Testnet phases.
#Cortensor #AI #DePIN #AIInfra #DevLog
Praxis will demo again at the next ERC-8004 Community Call tomorrow!
Same mission. Bigger scope.
Hosted by the ERC-8004 authors, Ethereum Foundation, and dAI Ethereum team, this call marks another important moment for us to show what we got.
Join us tomorrow if you want the latest and greatest from the ERC-8004 community.
Here’s what will be discussed 👇
Ecosystem Map — debuting the first view of the 8004 landscape
Demos — SDKs, explorers, and reputation in action
Taxonomies — defining categories for agent capabilities
Pre-Mainnet Smart Contract Changes — final pre-launch tweaks for clarity and UX
Join to see how the ecosystem around trustless agents is taking shape and how we are doing our part.
PRAXIS
🛠️ DevLog – NodePool Modular Refactor: Regression & Fixes (WIP)
We're deep in regression on the modular NodePool stack (Data / Logic / Util). Core paths are wired, but several race conditions surfaced under churn - especially around reserve ↔ release flows for ephemeral nodes.
🔹 Findings (so far)
- State races (ephemeral): Conflicts between reserve and release when rapid re-assignments occur; split data (NodePoolData) vs. logic (NodePool/Util) exposes timing gaps.
- Dedicated sessions: Stable in cycles; no critical issues observed.
- Fixes implemented: Added strict state checks within NodePoolData to ensure valid transitions and prevent stale or double actions across modules.
🔹 What we're doing next
- More regression locally, then scale tests on Testnet-0 to validate behavior with a larger pool size.
- If needed, push to Testnet-0/1 this week and co-test with node ops (that’s what the Testnet phases are for).
- Add lightweight atomicity/lease checks in NodePoolUtil, plus stricter state transitions in NodePoolData to reduce cross-module drift.
🔹 Risk & mitigation
- Ephemeral churn can still expose edge cases; we'll instrument counters/metrics and gate enforcement behind feature flags during rollout.
#Cortensor #DevLog #AI #DePIN #AIInfra
Over the past few months, we have seen some brutal market corrections. Many projects gave up and disappeared and that period became a true litmus test, separating real builders from the scammers.
That is exactly why I support projects that kept pushing forward despite the odds. $CLOAK is one of those as it stayed the course, kept building, and now has a real chance to make its mark in this space.
The storm is finally clearing and it’s nearly time for the real ones to shine!
🛠️ DevLog – NodePool Expansion (Data / Util / Filters Refactor)
We've completed the initial NodePool modular split - isolating data, business logic, and selection algorithms into separate modules to improve maintainability and prepare for richer SLA-driven inference.
🔹 Current Structure
- NodePoolData → stores all node/pool states with getter/setter interfaces.
- NodePool → handles core orchestration and routing logic.
- NodePoolUtil / Filters → new module dedicated to node selection algorithms and future SLA logic.
🔹 Current Status
- Refactoring for algorithm/selection logic into NodePoolUtils is done.
- NodePool is now cleanly separated into three layers: NodePoolData, NodePool, and NodePoolUtils/Filters.
🔹 Next Steps
- Once stable, begin implementing new SLA filters during Testnet phases and extend scoring logic in NodePoolUtils to enhance inference quality and routing precision.
- Roll out updates progressively across Testnet-0 and Testnet-1 for live evaluation.
#Cortensor #DevLog #AIInfra