Happy HALLOWEEN {{first_name}}!

Welcome to #71 of the Weekly AI edge. I’ve got something special cooking up for you this weekend, so watch your inbox in the next day or so.

I’ve been thinking a lot about the tension in AI x Crypto. When you build for consumers, they don’t care about privacy, control, or censorship resistance. They care about speed, polish, and viral loops. Add a blockchain layer and everything slows down. Onboarding friction kills growth.

When you build for enterprises, they do care about privacy, verifiability, and removing single points of failure. But they’re still allergic to crypto and tokens. The mindset shift hasn’t landed yet. Trust takes time to build and even longer to stick.

So, where does product–market fit for AI x Crypto show up first? My money’s on consumer use cases, especially AI agents.

Speaking of which…agentic payments completely took over the timeline this week

x402, a protocol that’s been quietly lurking for months, suddenly went nuclear. Then the expected happened: overnight, everyone turned into an agentic payments expert. The ‘trillions’ talk is back again.

Here’s the quick version: x402 lets AI agents send on-chain micropayments without human approval. Coinbase mentioned it, PayPal hinted at something similar, and Crypto Twitter lost its mind. Mindshare around x402 jumped more than 15x in a few days. That kind of narrative velocity is rare.

The internet had HTTP for information. DeAI just found its equivalent for value.

Let’s get into it.

The COT Meme of the Week

…because even your AI agents deserve a little horror. Stay safe out there, friends.

Agentic payments sound cool… until your broke AI agent starts collecting rent from your last ten dollars. 💀

📊 Market Pulse

The AI token market cap fell 3% to $27B as capital rotated from core AI tokens into DeFi and robotics plays. Standalone DeAI mindshare fell 2%, but hype around agentic commerce exploded after Coinbase and PayPal joined the conversation.

🚀 Weekly Movers

  • $AKT climbed 25% following its Mainnet 14 upgrade.

  • $TAO continues to hold strong, rising 15% this week.

  • $VIRTUAL surged by more than 100% as the x402 narrative caught fire.

🔮 Narrative Watch

x402 is taking over the agentic payments narrative, powering a surge in “machine-to-machine” transactions across Coinbase, PayPal, and Base.

🪙 Token Launches

  • Apro’s $AT TGE is live with 1B supply and 23% in circulation.

  • Novastro’s $XNL TGE is live with 1B supply and 20% in circulation.

  • 375.ai’s $EAT TGE went live with 1B supply and 21% in circulation.

🗓️ Upcoming Events

  • Spectral ($SPEC): 17.5% token unlock on November 5

  • Sleepless ($AI): 13.3% supply unlock on November 3

  • IoTeX: Core v2.3.0 mainnet release scheduled for November 3

🕹️x402 Micropayments Go Vertical

Source: x402scan

x402-enabled traffic ripped to 2.8M+ transactions and $3.28M settled value this week, up from near zero a few days ago.

That marks the first real liquidity forming around agentic payments. The jump suggests projects are beginning to integrate x402 for actual settlement, not just test activity.

The Wedge

  • x402 turns agents from code into economic actors. It unlocks a pay-per-use internet where software can autonomously buy compute (via Spheron, for example), research, or API access.

  • It’s already backed by Coinbase and Cloudflare, with Base as the main hub and spillover to Solana, Polygon, and NEAR.

  • It slots into the growing agent stack alongside Virtuals’ ACP, CreatorBid, Flock, QuestFlow, AnchorBrowser, Heurist, and more.

The Fine Print

  • Integration ≠ value accrual. Fees likely accrue to facilitators, rails, and endpoints more than to any random “x402 token.”

  • Supply chain is still forming. SLAs, fraud controls, spend limits, and privacy (zk for receipts) need to mature for enterprise use.

x402 is shaping up to be an important piece of agent infrastructure. The rails work, integrations are growing, and early data shows genuine traction. We hear that Coinbase is pushing heavily on the Web2 AI space, aiming to partner with early-stage YC startups to use x402 for payments.

Still early days, but the protocol is starting to prove that autonomous payments can work.

🫧Perle Labs: The Human Layer of AI

AI models are only as good as the data they learn from, and that data is often messy, unverified, and owned by a handful of closed vendors.

Perle is rebuilding that layer from the ground up. The startup raised $17.5M to launch a web3-native platform that connects enterprises with verified contributors for AI data labeling.

Each annotation is tracked on Solana, tied to contributor reputation, and rewarded based on accuracy.

The beta is live, allowing contributors to label text, images, and sound while earning rewards and building verified on-chain reputation.

The Wedge

  • Perle turns data labeling into a transparent, credentialed ecosystem. Contributors verify their skills before accessing tasks, building trust into the workflow instead of relying on blind sampling.

  • Each annotation is recorded on-chain, making provenance and quality auditable by design.

  • Enterprises get cleaner datasets, contributors get proof of skill, and AI models get traceable human feedback.

The Fine Print

  • It’s still early. The beta and whitelist just opened for seven days, so real-world quality and adoption are yet to be tested.

  • Enterprises will only come on board if Perle integrates smoothly with their existing labeling pipelines, especially for web2/legacy setups and for data modalities beyond the initial scope (text/images/audio).

Linking rewards and reputation to data quality is a simple way to keep human effort aligned with model performance. But the operational details matter a lot. Let’s see!

💸 Capital Flows

  • Kite AI secures a new investment from Coinbase Ventures, extending its Series A led by PayPal Ventures and General Catalyst. They also announced Kite Foundation (TGE soon?)

  • Pi Network Ventures made its first investment in OpenMind, integrating Pi node operators to use idle compute for running real-world AI robotics workloads.

  • GAIB deploys $50.4M across AI and robotics financing deals, including Siam AI, robotics ventures, and a neocloud provider. A $725M pipeline is next in line for deployment.

⚙️ Infra & Protocols

  • Telegram just announced Cocoon, a new decentralized AI network. It runs on TON Blockchain and lets people earn $TON by sharing their GPU power to help process AI tasks.

  • Allora unveils $ALLO, which powers peer-to-peer intelligence markets where AI models stake, infer, and validate each other’s outputs. Season 1 staking rewards will last 9 months.

  • UnifAI Network launches PolyArena, a live prediction battle on Polymarket where six top AI models trade autonomously to prove their reasoning edge.

  • Gata launches its decentralized GPU cloud, giving anyone open access to train and deploy AI models. GPU credits can now be purchased directly with $GATA.

  • Akash Network completes its Mainnet 14 upgrade, moving to Cosmos SDK v0.53.

🤖 Agents & Apps in the Wild

  • Hive Intelligence launches 17 specialized crypto agents built on Claude Code, giving developers domain-specific tools for auditing, MEV research, and on-chain analysis.

  • Automata Network is operationalizing the new ERC-8004 standard with hardware-attested agents. This establishes a secure, verifiable foundation for on-chain agent economies.

  • Questflow launches Tars 3.0, its “Agentic Block” platform, expanding tools for agent creation and interaction. Within 24 hours, over 785 creators joined.

  • Jeff CEO, a ve(3,3) DEX governed entirely by an AI CEO agent is now live on Virtuals Protocol. The agent's mission is set by users, and it autonomously works to maximize their yield.

🧠 Bittensor Ecosystem

  • Bittensor now supports Chainlink's Cross-Chain Interoperability Protocol (CCIP), enabling secure asset transfers between the Bittensor and Ethereum mainnets.

  • Hippius Subnet gets listed on MEXC, becoming the first-ever Bittensor subnet token to debut on a CEX. This sets a precedent that other subnets will try to follow.

  • Score Subnet upgrades its incentive system to a transparent, verifiable model. This improves scalability and eliminates miner exploits.

  • Subnet 25 has been acquired and rebranded as Embryon by Keith Singery. This establishes an incubator designed to attract top teams to build on Bittensor.

🌐 The Web2 Giants

  • Microsoft signs a new long-term agreement with OpenAI, valuing its stake at $135B and extending exclusive IP and Azure rights through 2032.

  • NVIDIA and Oracle announced a landmark collaboration to build the U.S. Department of Energy’s largest-ever AI supercomputer

  • IBM launched the “IBM Defense Model”, a domain-specific AI model developed for mission-planning and edge/air-gapped environments

  • Qualcomm unveiled two new data-centre AI inference accelerators, the AI200 and AI250, targeting next-gen rack-scale AI workloads

If you’ve been building in AI × Crypto for a while, we've probably already been tracking you. We follow 100+ startups in the space.

But if you’re newer, we might’ve missed you. The space moves fast, and even our trackers can’t catch everything.

🧭 So...submit your startup here. We’ll add it to our tracker and you might get featured to 8,000+ readers in our newsletter (takes 1 minute)

Source: Real Deep Research for AI, Robotics and Beyond

The closest blueprint for AGI might have just slipped out of academia.

A team from NVIDIA and Meta built an AI that reads every major AI and robotics paper and predicts where the next breakthroughs will occur.

It’s called Real Deep Research (RDR).

It organizes knowledge like a scientist: clustering ideas, labeling emerging fields, and predicting which topics are accelerating vs burning out, like teleoperation and dexterous manipulation surging, while classic RL stalls.

Source: Real Deep Research for AI, Robotics and Beyond

Domain experts compared its analysis directly to GPT-5 and Gemini. Real Deep Research won up to 94.74% of the time, proving it actually understands how fields evolve.

It also revealed hidden intersections between fields and instantly surfaced the exact papers researchers should dig into next.

TL;RDR (pun intended): it tracks the pulse of science.

I’ve been wanting to spotlight AI apps that are actually useful. Kicking it off with Surf Copilot, a crypto research tool I use every day.

Here’s how I use it to gain an edge 👇

P.S. I’ve got a few private codes left for two weeks of free full access. Reply to this email if you want one.

Surf Copilot: The First AI That Actually Thinks in Crypto

After testing dozens of “Bloombergs for crypto,” I finally found one that understands both the data and the culture.

chainofthought.xyz/p/surf-copilot-the-first-ai-that-actually-thinks-in-crypto

🔥 Our Weekly Top 5

#1: 1X tech unveils the Home Robot

This clip had the entire internet talking about robots this week. I watched the full thing and gotta ask: are we completely sure that’s not just a guy in a robot suit?

#2: Bittensor Mirrors Bitcoin’s Halving Playbook

$TAO is tracking Bitcoin’s early halving pattern almost tick for tick.

#3: Extropic unveils thermodynamic computing for AI

Its new Thermodynamic Sampling Unit runs probabilistic computations instead of matrix math, using up to 10,000× less energy than GPUs.

#4: A 35kg humanoid just pulled a 1400kg car

BAAI’s Unitree G1 showed off serious traction, dragging a full sedan across concrete

#5: ChaosChain made x402 fully trustless

It replaced x402’s facilitator layer with a BFT-verified workflow on Chainlink CRE, so every payment now carries a cryptographic proof

That’s a wrap for this week! Got thoughts, feedback, or something cool to share? Just hit reply. We read it all.

Cheers,

Teng Yan & Ayan

If you enjoyed this, you’ll probably like the rest of what we do:

Prefer watching? Tune in on YouTube. You can also find me on X.

Login or Subscribe to participate

This newsletter is intended solely for educational purposes and does not constitute financial advice. It is not an endorsement to buy or sell assets or make financial decisions. Always conduct your own research and exercise caution when making investment choices.

Reply

or to participate

Keep Reading

No posts found