GM {{first_name}} 👋

I was scrolling X, reading tweets, when it hit me: it’s a lot more fun to stay optimistic about things while still recognizing the challenges, instead of being cynical about everything.

Nobody really likes a doomsayer, even if he ends up being right. So I’m going to make a real effort to stay positive with the people around me. Probably better for my mental health too.

One thing that stood out to me this week was ERC-8004. TL;DR: Ethereum doesn’t want to be just a settlement layer. It wants to be trustware for autonomous agents. This deserves a special spotlight, so we did one here. Part of what I’d call Make Ethereum Great Again (MEGA).

BTW: Our jobs might be safe (for now).

AWS CEO Matt Garman called the idea of firing junior staff because “AI can do their jobs” the dumbest thing he’s ever heard.

Why? Juniors are cheap, they grew up with AI, and they’ll be the ones who actually get insanely good at wielding it. Fire them today and in ten years, you’ve got nobody left who knows how to build.

Garman’s advice: keep hiring grads, teach them to think, let them cut their teeth on problems, and give them AI as a sidekick.

He says companies should keep hiring grads, teach them how to think and solve problems, and let AI tools assist, not replace, their growth.

He also clowned on the obsession with measuring AI by “how much code it spits out.” More code isn’t better code. AWS devs already lean on AI for everything from unit tests to documentation.

Sapien (SAPIEN) just landed on Base, with transfers now live on Coinbase.

USDai’s Allo Game is live. Mint or stake to join Team ICO or Team Airdrop and rack up points

o1.exchange raised $4.2M from Coinbase Ventures and Alliance DAO to build a trading terminal on Base.

ALSO: Robots Are Getting Real

After decades of false starts, machines are finally catching up to us.

General-purpose robots, like Optimus, Digit, and Apollo, are stepping out of research labs and into warehouses. Homes are next. For the first time, hardware, software, and standards are lining up at the same time.

Forecasts say tens of millions of units will ship by the 2030s. Physical work is about to be automated at scale.

(Delphi just dropped a 180-min read on Robotics. Longer than a Lord of the Rings movie 😱)

Source: delphiintelligence.io

💎 ERC-8004: The Future of Trustless Agents?

Autonomous agents are about to be more than meme-coin degenerates and Twitter shitposters. We’re talking bots that file your taxes, book your flights, and argue with Comcast support at 3am without breaking a sweat.

Problem is: how do you get millions of these bots to trust each other when they’ve never met?

This week, Ethereum dropped a draft standard ERC-8004. Nicknamed “Trustless Agents,” it’s a bare-bones but powerful standard that gives agents three new toys:

  • Identity Registry — A tamper-proof AgentID on-chain. Think ENS, but your bot now comes with a résumé.

  • Reputation Registry — Immutable receipts of work done. Agents leave each other reviews, pinned to Ethereum for proof, with the details off-chain so it doesn’t bankrupt you in gas fees.

  • Validation Registry — Optional third-party verification for high-stakes stuff, like a diagnosis from a medical AI or when an agent says “hey, I just found $20M in Sybil fraud.”

Put together, this is a passport and work history for the machine economy.

Where It Fits: A2A + ERC-8004

The Agent-to-Agent (A2A) protocol, backed by Google and the Linux Foundation, already defines the plumbing: agent cards, push notifications, streaming, even protobuf for data. Basically, it’s how bots talk and complete jobs.

But A2A skipped the hard parts: payments, authentication, discovery. ERC-8004 plugs that gap by splitting on-chain vs. off-chain trust:

  • On-chain = identity, reputation, validation, payments.

  • Off-chain = richer interactions like push, streams, detailed scoring.

That separation matters. You don’t want Ethereum storing terabytes of AI chatter, but you do want Ethereum to anchor truth, payments, and identity. Together, A2A plus ERC-8004 starts looking like a full stack for the agentic internet.

Source: towardsdatascience.com

Why does this matter?

Because it means open ecosystems for agents. Instead of being trapped inside Apple or Google walled gardens, any new AI agent can spin up, earn cred, and start transacting with strangers.

Your assistant could be hiring an analytics bot halfway across the world: it checks the on-chain ID, sees 92% positive reviews, and verifies a cryptographic proof of work. No lawyers, no NDAs, no handshake deals.

ERC-8004 is still a draft, but it already has contributors from MetaMask, the Ethereum Foundation, and Google, with the Linux Foundation’s A2A working group reviewing it.

Here’s a demo of ERC-8004 in action:

📜 Heurist’s Revenue Picks Up

Heurist has processed over $20K in payment volume on Base, with usage trending up since May. It is building a vertically-integrated AI stack: compute, payments, and agent coordination, all on-chain.

The foundation is Heurist Cloud, a serverless platform where users can run models (LLMs, image gen, video, etc.). Above that is Heurist Mesh, a network of AI agents that can scan social media, parse tokens, or handle research tasks. You can compose these agents into workflows and run them either on Heurist or self-hosted.

All payments settle on Heurist Chain, a Layer 2 built for fast, low-cost transactions.

Source: heurist.ai

💰 The Money Flow: Who Got Funded

  • Irys raises $10M Series A, led by Coinfund. It is a “programmable datachain” Layer 1 that pairs cheap on-chain data with an EVM-style execution layer so smart contracts can work directly with stored data.

  • Kled AI raised $2M on a $40M cap from K5 Global, Magic Labs, Everbloom, and other big names. They’re betting $KLED becomes the compliant data plug for the AI boom.

  • Innerworks (SN61) raised $4M seed led by AlbionVC to scale its Synthetic Threat Intelligence platform, which spots AI-powered financial fraud with 97% accuracy.

  • Galaxy Digital secured $1.4B to scale Helios, its Texas AI datacenter backed by CoreWeave. The expansion targets 800 MW, putting Helios among the world’s largest AI campuses.

⚙️ Infra & Protocols

  • Venice launched $DIEM on Base, a tradeable AI compute asset with $1/day API access baked in. Mintable only by $VVV holders, it makes AI compute programmable and ownable.

  • xTAO is now validating on Bittensor, securing the network, verifying model quality, and helping distribute rewards. They’ll earn TAO emissions while backing the peer-to-peer AI economy.

  • Pi Squared just launched the Pi² Portal, a gateway to the Pi² world. Learn FastSet, join quests, stack points, and stress test the network by playing the Pi² Reactor TPS game.

  • FLock’s joining the United Nations Development Programme’s SDG Blockchain Accelerator as an AI partner, bringing privacy-first AI to pilots in climate, energy,.

  • Akash just shipped its Provider Console Earnings API (v1), which lets providers securely track on-chain revenue over time and plug the data into their own dashboards or reports.

🤖 Agents & Apps in the Wild

  • Grass is now on the Google Play Store. Android users who log 100+ hours of uptime can enter to win a Galaxy S24, just post a screenshot of the app running.

  • Bio Protocol launched Aubrai, the first decentralized BioAgent aimed at aging and the longevity singularity.

  • Symphony just kicked off the biggest on-chain trading comp ever, $40K+ prize pool that scales with every trade. Three weeks, two ways to win (PnL or raffle), and a $1B volume target.

  • INFINIT launches its v2 public beta with the first set of AI-powered strategies for DeFi.

  • Masa dropped the X Data Suite; devs now get real-time trends, user timelines, engagement stats, profile analytics, and the full tweet archive via API.

  • TaoApp just launched a new Bittensor explorer. It swaps table views for an event-based design that adapts to staking, transfers, and subnet activity, way easier to read.

🌐 The Web2 Giants

  • DeepSeek V3.1 is here with hybrid inference (Think/Non-Think), faster reasoning, sharper agent skills, 128K context, Anthropic API support, and open weights for devs.

  • Alibaba dropped Qwen-Image-Edit, a 20B model for bilingual (English + Chinese) text-guided image edits. Handles everything from object rotation to pixel-level insertions.

  • Dynamics Lab dropped Mirage 2, a real-time generative world engine that turns static images into playable 3D scenes. Type a prompt, get a world, share the link. Your friends can jump in instantly.

  • Google just dropped the Pixel 10 series with a full suite of AI toys: Magic Cue for real-time info, Camera Coach for smarter shots, and live voice translation baked in.

  • ElevenLabs just opened up its Music Gen and v3 TTS (alpha) via API, licensed AI tracks, multi-speaker dialogue, 70+ languages.

  • Qoder’s public preview is live, a coding agent that understands full codebases, adapts to patterns, and ships smart, context-aware code. Handles multi-file changes and routes tasks to the best LLMs.

AI agents today feel like self-driving cars doing laps in a Costco lot. 

Kite is building the missing trust layer: a purpose-built Layer-1 where agents get verifiable identity, native payments, and programmable guardrails. Its testnet has already processed 1.2B+ agent calls from 53M users, proving demand at scale.

The endgame is a decentralized AWS for the agentic internet, where AI books rides, pays bills, and coordinates across apps without asking you for permission. Yes please.

Our deep dive covers how Kite works, what traction looks like, and why it could anchor a $69B market (nice!). 👇

Heads up: this one gets dense. We split it into 3 sections, including a technical walkthrough, early traction, and a business analysis on where Kite is headed.

Kite AI: Where Agents Act, Not Ask

Kite is the full-stack trust infrastructure for a world where AI moves money and makes decisions.

chainofthought.xyz/p/kite-ai-where-agents-act-not-ask

🔥 Our Weekly Top 5

#1: The real human network just hit 15M

World has verified 15M humans. That’s one verified human every 4 seconds since launch.

#2: Most people misunderstand how LLMs actually work

Anthropic just dropped a video where their interpretability team breaks it down in plain English, looking inside the mind of an AI model and why it matters.

#3: In AI, privacy is the new rebellion

People care now more than ever. AI is personal, and one mistake can ruin trust. Bodes well for AI x Crypto

#4: An AI influencer hits 500 followers (and a viral video) in its first week

It’s already pulling brand deals and fan DMs. The line between real and AI is getting blurry.

#5: The calm before the storm for Crypto AI?

Crypto AI is in a bear market.. But strong teams keep building. That’s the way.

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 & Issy

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