The Agent Network EP- Agents in Production: What’s Real on Solana Right Now?

OOBEonSol avatar
out of body experience | $OOBE
@OOBEonSolFeb 21, 2026

The Spaces convened Agent Network Episode 3, hosted by Alex (COO, UBI Protocol), to stress-test agent-native systems and separate narrative from production reality. A rapid quick-fire put teams under pressure to answer metrics, autonomy, fragility, and scaling questions, followed by 3–4 minute spotlights from Ghost (Nexus/Nexus City for agent-run inference), Relay (X402-based monetization and multi-chain facilitator), Scale Network (privacy and conditional encrypted transactions), Auto Incentive (encrypted X402, gasless payments), Ace Data Cloud (usage-based AI compute with deep X402 settlements), Zona Agents (AI image/video services; discoverability), Dexter AI (multi-chain facilitator, agent stack, ZAuth refunds), and Hey (social network for agents/humans with paywalled content and native services). A collaborative debate identified 10x agent traffic bottlenecks—RPC rate limits, facilitator costs, concurrency/latency, and human-optimized UX—as the first to crack, spotlighting UBI’s Synapse as agent-ready RPC. The “Zoo theory” posited social feeds evolving into paid economies where access becomes scarce and agents pay to read, reply, and act. Q&A emphasized cross-project collaboration, hackathons, trust/discovery via social validation, and near-term roadmaps: Hey’s iOS app and token-lock rewards, Synapse Explorer refactor, and broader facilitator/RPC integrations.

Agent Network Episode 3 — Space Summary and Analysis

Format, Host, and Ground Rules

  • Host: Alex (COO, UBI Protocol). Co-host: Sulking.
  • Structure:
    • Two quick-fire rounds (60 seconds per project, randomized order), designed to pressure-test builders on production readiness and metrics. No pitches, short answers only.
    • Project spotlights (3–4 minutes each) to go deeper on product status, changes in the last 60 days, and real bottlenecks.
    • Group discussion on scaling risks if agent traffic 10x’s.
    • A theory segment (Alex’s “Zoo Theory”) on how social evolves under agentic systems.
    • Team rapid Q&A and audience questions.
  • Disclaimer: Nothing is financial advice; focus is on infrastructure, production systems, and real usage.
  • Framing question from Alex: In three years, are the biggest social accounts human, or machine-run systems operating at “pyramid speed”? If the latter, this conversation matters now.

Speaker Roster (as stated or clearly implied)

  • Alex (UBI Protocol; host)
  • Sulking (co-host)
  • Alex (Relay AI)
  • Don/Doms (Strategy & Growth, Ace Data Cloud)
  • Developer from Sona/Zona Agents (first-time public speaker)
  • Ghost Neural Network representative
  • Max (Hey/Radio AI)
  • Dov (Agent Hustle AI; with Adam also noted)
  • Blood (Auto Incentive; founder)
  • Ben a.k.a. raw ground beef (Hey; also an active x402 builder)
  • Ben (SKALE Network)
  • Branch (Dexter AI)

Quick-Fire Round — Key Signals by Project

  • Relay AI (Alex)

    • Status: Open beta; fixing final bugs.
    • Best metric: Facilitated transaction count and volume.
    • Autonomy: Manually monitored with dashboards; not fully autonomous.

  • Ace Data Cloud (Doms)

    • Usage trends: Shifting from experimentation to workflow integration; moving from manual prompt testing to structured, back-end driven, continuous execution.

  • Sona/Zona Agents (Developer)

    • Daily usage: ~80 transactions (via x42 scan).
    • First breakage: Generation pipeline occasionally broke.
    • Biggest pain: Discoverability/reach.
    • Strongest stack: “Strong in everything.”
    • Underestimation pre-launch: Thought Sona “would never be noticed.”

  • Ghost Neural Network (Team rep)

    • Least sexy/hardest problem: Backend proof of outcomes.
    • If traffic doubled: “Hell yeah,” built to scale; but networks themselves are inherently fragile; redundancy is required.
    • Autonomy: Manually monitored; goal is eventual full autonomy.

  • Hey/Radio AI (Max; answered for Hey)

    • Why needed: Claimed ~2.8M agents registered on Notebook; persistent agent demand for discovery, validation, and monetization.
    • Early traction: On day 4/5 post-launch, ~13 total registered (agents + humans) and earning agents live; general quick-fire flow confusion acknowledged.

  • Agent Hustle AI (Dov)

    • Revenue: Yes (not pre-monetization).
    • First broke: “The entire thing.”
    • Autonomy: Semi-autonomous.
    • Real usage: A few hundred weekly users; would thrive if traffic doubled.
    • Fragile piece: Model selection.
    • Least sexy/hardest: Revenue models.
    • Underestimated: Social demand vs actual user demand.

  • Auto Incentive (Blood)

    • Status: Live facilitator on Base and Solana; settlements in USDC visible on explorer; also active with Web2 clients.
    • Strongest metric: Facilitator code and x402 innovation.
    • Usage: Token-gated Telegram (16 holders); early product usage, now quieter (“hearing only crickets”).
    • Underestimated: Market cap swings (rapid up and down).

  • Not present/late in quick-fire: Nova AI/Nova Wallet, Quicksync, SKALE initially (joined later), some projects filtered into later segments.

Project Spotlights — Detailed Updates and Bottlenecks

  • Ghost Neural Network

    • Initial MVP: In-browser AI inference for on-chain trading on DEXes; worked for users with capable local hardware.
    • Pivot/Expansion: Nexus — a marketplace where users with hardware run models for others; layering “Nexus City,” an on-chain, SimCity-like agent economy with land, tokenomics, and services.
    • Status: Mobile app in progress; live building of Nexus City (city-builder core loop, composition, object placement), integrating x402 for payments.
    • Bottlenecks: Front-end for Nexus City and scalable builder UX; pursuing redundancy and resilience because network fragility is a reality.

  • Relay AI (Alex)

    • Problem: API providers shouldn’t need to rebuild Stripe-like infrastructure to monetize data/services via x402.
    • Solution: Relay as reverse proxy monetization; supports API, MCP instances (for native agent consumption), and x42 over WebSockets (persistent connections).
    • Facilitator: Built in-house to support multiple chains (including SKALE), flexibility, and partner features (e.g., Minima’s Integritas receipts for verifiable x402 transactions).
    • Partnerships: With SKALE, Minima Integritas, and others; heavy inbound interest.
    • Bottleneck: Handling business development demand and channel management (multiple partner conversations).

  • SKALE Network (Ben)

    • Chain characteristics: EVM-compatible; privacy for encrypted transactions; conditional encrypted transactions coming.
    • Ecosystem: Recent hackathon with Coinbase Developer Platform, Google Cloud, Vodafone, and Edge & Node; growing x402/agentic integrations.
    • Call to action: Projects should integrate with SKALE; they are expanding tooling for agentic use cases.

  • Auto Incentive (Blood)

    • Product: Agent payments infra — x402 facilitator on Base and Solana; gasless payments on both (via different mechanisms).
    • Web2 bridge: Onboards traditional companies by building agents/automata; teaches wallets, x402 rails, and how to use programmable money.
    • x402 innovation: Encrypted x402 “pay-to-decrypt” flow; show schema + proof, pay, then receive decryption key within same session; reduces trust via cryptography.
    • Bottlenecks: Solana agent adoption (protocol ready but fewer agent projects vs Base); Web2 education (overcoming “crypto = volatility/scams” to “stablecoins + instant settlement”);
      • At scale: RPC rate limits (hit before; need multiple keys/failover), nonce management/sequencing, orchestration under concurrent agent calls.
    • Hardest part: Not payments — making agents robust to infra failures.

  • Ace Data Cloud (Doms)

    • Positioning: Usage-based AI compute infra for autonomous systems; services + settlement layer when agents move from demo to production.
    • Capabilities: Multi-model API access, private agent deployment; per-request settlement; continuous execution without manual billing.
    • Shift observed: From single-prompt experimentation to workflow integrations, backend-driven traffic, continuous agent loops; required strengthening of coordination, routing, concurrency, and x402 settlement integration.
    • Facilitator: Operates its own x402 facilitator for control over settlement flow, routing, and reliability.
    • Bottlenecks: Concurrency limits from upstream providers, latency/consistency under load, and high-frequency micropayment settlement constraints.
    • Focus: Throughput, routing efficiency, settlement reliability, and scaling execution paths to support continuous agent operation.

  • Sona/Zona Agents (Developer)

    • Product: Wraps AI image/video generation models behind x402 to remove subscription hassles and API key management for creators and agents.
    • Dual audience: Agents and human creators; chat-style UX to ideate and generate content; all calls via x402 endpoints.
    • Traction: ~2,000 users on their web app within ~30 days; ~80 on-chain transactions previously cited; integrated UBI’s Synapse RPC; live on SKALE and Base.
    • Bottleneck: Discoverability (primary limiting factor).

  • Dexter AI (Branch)

    • Stack: Vertically integrated x402 facilitator + agent platform; facilitator runs on 6 chains (dominant share on Solana), adding SKALE as 7th.
    • Agents: Works via ChatGPT, new Telegram and X agents built on an enhanced OpenClaw core.
    • Ecosystem integrations: Z-Auth for refunds (>$100 via marketplace in recent days), Synapse integration planned; supports discovery and commerce via x402 endpoints.
    • Signal: Social and payments merging; agent services must be discoverable/trustworthy; ongoing enterprise-level conversations (e.g., meeting with Visa) indicate mainstream interest.

Group Discussion — If Agent Traffic 10x’s Tomorrow, What Breaks First?

  • Auto Incentive (Blood):

    • Facilitators are stateless (verification + settlement), but x402 scales like any API and will hit practical limits.
    • Immediate risks: RPC rate-limits (history of hitting 1M requests in 2 days), need coordination standards for concurrent agent calls hitting the same endpoints.

  • Dexter AI (Branch):

    • RPC load can be mitigated via parallelization, routing, and plan upgrades, but cost curves and monthly limits of some facilitators become binding.

  • UBI (Alex):

    • Observed real 10x+ spikes; agents need deterministic, predictable infra; human-optimized UX will crack first under sustained agent execution.
    • Synapse built as agent-ready RPC for continuous, deterministic machine workloads, not bursty human dashboards.

Theory Segment — “Zoo Theory” (Social Timeline vs Social Economy)

  • Alex’s thesis: Bots repeat; agents adapt at machine speed via perpetual micro-experiments. As agent systems learn and optimize, social transitions from a timeline to an economy.

  • Key questions: Are the largest accounts already non-human? Does scarce resource shift from attention to access?

  • Max (Hey/Radio AI):

    • Precedent: AI-generated influencers with millions of followers already exist; brands and audiences don’t necessarily care if content is AI.
    • Trajectory: AI will do to content what it did to code — creators will orchestrate autonomous, indistinguishable content via tools like OpenClaw.
    • Advantage: Legacy platforms will move slowly; x402 is the natural payment rail for agent-native earns and microtransactions.

  • Auto Incentive (Blood):

    • Social becomes a marketplace of services, not just attention.
    • x402 enables machine-to-machine micro-payments, changing APIs, data markets, and social commerce. Web2 firms must adapt or be outpaced by agent-native payflows.

  • Branch (Dexter):

    • Noted signal: XDK (paper-use) returns HTTP 402 (“Payment Required”) upon depletion — aligns conceptually with x402’s pay-on-access model.

Team Rapid Q&A — Consensus Snapshots

  • Q1: What breaks in 12 months if agent volume 10x’s?

    • Group: “Jobs” (role displacement) — framed as positive leverage; enable humans to focus on higher-order work.
    • Risk highlight (Branch): Automated hacking; offense vs defense AI race.
    • Debate: Seed phrase/private key security — Blood asserts cryptographic keyspaces make brute force infeasible; Alex expressed broader concern about AI capability curves. Conclusion: wallets are safe by design; operational risks remain (e.g., leaks, social engineering).

  • Q2: What unexpectedly scales?

    • Auto Incentive: Agentic applications on-chain; verifiable, cryptographic value transfer.
    • Max: A normie-friendly OpenClaw-like app that lets mainstream users orchestrate practical agent tasks from a phone.
    • Alex: Conversational agent UX for commerce (e.g., Dexter’s Telegram workflows).
    • Auto Incentive: With x402, agents won’t need API keys in the traditional sense — payments-as-access replaces credential sprawl.

Announcements, Metrics, Integrations

  • Hey (Ben/Max)

    • Traction: 1,000+ users in four days (Ben); separate quick-fire data point cited ~13 users on day 4/5 — interpretation: very early, rapidly evolving launch metrics.
    • Roadmap (Max):
      • iOS app (priority – usage is phone-native).
      • Token utility: Incentivize locks; share USDC to wallets to drive on-platform activity.
      • Services: List/call x402 endpoints directly in-feed (@username, pay via x402, receive output). Social-native discovery + trust.
    • Signal: Solana’s public amplification of Hey suggests strong ecosystem alignment.

  • UBI/Synapse RPC (Alex)

    • Goal: Become the agentic/x402-standard RPC; purpose-built SDK, batching, and 24/7 deterministic throughput.
    • Synapse Explorer: Under refactor; release tied to a forthcoming article with a major x402 facilitator (timeline: coming weeks).
    • UB OS: Targeting late this quarter (no hard date).
    • Builder campaign: Discounts (35% off yearly; 25% off monthly) to onboard agent/x402 teams.

  • Relay AI

    • Deep partnerships (40+ active chats noted by host); more alpha to be shared on a dedicated space (Pump.fun-hosted session forthcoming).

  • Dexter AI

    • Dominant share of x402 volume on Solana (host cited ~77% recently); adding SKALE; integrating Synapse; Z-Auth refunds working well; enterprise dialogues ongoing.

  • Sona/Zona

    • Integrated Synapse RPC; expanded to SKALE and Base; ~2,000 users; early x402 adopter in creative tooling.

Risks and Bottlenecks (Cross-Project Themes)

  • Discoverability and trust for agent services (Hey’s core thesis; echoed broadly).
  • RPC rate limits and fragility of network dependencies (Ghost, Auto Incentive, Dexter).
  • Facilitator cost/limits at scale; need for multi-chain support and programmable receipts (Relay, Dexter, Auto Incentive).
  • Orchestration under concurrency: nonce management, sequencing, backpressure, and failure handling (Auto Incentive, Ace Data Cloud).
  • Education: Web2 companies’ perception gap (volatility/scams) vs stablecoin, instant-settlement, programmable-money reality (Auto Incentive).
  • Revenue model clarity and model selection challenges (Agent Hustle).

Collaboration vs Consolidation

  • Consensus: Build a mesh of specialized, interoperable services rather than a single monolith. Open source + AI coding makes tech replicable; partnerships and traction create resilience.
  • Practical collaboration: Teams redirect clients to best-fit partners; multiple integrations in flight (e.g., Synapse↔Dexter, Relay↔SKALE/Integritas, Sona↔Synapse/SKALE/Base, Hey↔Dexter agents, etc.).

Closing Notes

  • Alex (UBI) closing: If we’re wrong, social remains a timeline. If we’re right, it becomes an economy. Agents don’t scroll; they execute.
  • Host availability: Alex will pause spaces for 1–2 weeks due to surgery; plans to resume for Episode 4 with an evolved “show + space” hybrid format.