PublicAI’s $PUBLIC sold out on CoinList, oversubscribed and $2M raised. TGE is coming this August.

Codatta’s Booster Campaign S2 Week 3 is now live, featuring new AI and data annotation challenges open to eligible participants via Binance Wallet.

Sentient has integrated GPT OSS into Sentient Chat ahead of its upcoming GRID launch.

AI is about to eat a lot more electricity.

Right now, U.S. AI training uses around 5 gigawatts of power. That’s roughly what a small country uses. By 2030, that number could 10x to 50 GW, or more than 5% of the entire U.S. grid. Yikes.

To put that in perspective: GPT-4 needed ~20 megawatts. The next wave of models could each need 4,000 MW. That’s not a typo. That’s a small nuclear plant.

Why the spike? Models keep ballooning in size, and we’re stuffing more GPUs into tighter racks. Efficiency upgrades aren’t keeping pace. Grid stress is real, and data centers are starting to max out.

Eventually, the physics of scale will break the current model. That’s where decentralized compute starts to look real spicy.

Irys just passed 700 million on-chain data transactions on their testnet. Writes to its storage layer have jumped over 10x since the start of Q3.

Something’s catching fire.

Most blockchains (Ethereum, Solana, etc.) were never designed for large-scale data storage. They’re compute-first. Early “datachains” tried to fill the gap by making storage cheap, but the data stayed stagnant. You could write to the chain, but you couldn’t interact with it.

Irys takes a different route. Its storage is programmable. You can attach logic to the data itself: set permissions, automate royalties, even trigger downstream on-chain actions

This is enabled by IrysVM, which fuses execution and storage into a single system. No need to outsource execution to off-chain services. Smart contracts can natively work with stored data.

ts consensus design blends Proof of Work and Proof of Stake, with a 10% cap per miner to limit centralization. Pricing is pegged to USD per block, so costs don’t swing with token prices.

Still early, but Irys is one of the more interesting attempts at making on-chain data actually usable.

Nillion just launched nilGPT, a chat interface that can’t log your data. Because it physically can’t.

Most AI chats today funnel your messages through centralized servers. Everything you type (personal rants, sensitive files, corporate IP) gets stored somewhere, usually in plaintext, often forever.

nilGPT uses Nillion’s secret-sharing network to split your messages across multiple nodes. No single node ever sees the whole thing. Even if the network got fully compromised, all an attacker would see is encrypted noise.

Encryption happens on your device, using your passphrase. Decryption too. The system runs entirely inside secure enclaves, with no logs, no backend visibility, and no way for even Nillion to peek inside.

The interface feels like ChatGPT, but the backend runs on a zero-trust model. The architecture is built from first principles: cryptographic storage, confidential compute, and user-verifiable privacy.

There are three built-in modes—Assistant, Companion, and Wellness—but they’re just the start. Voice input, private search, attestations, and token-gated upgrades via $NIL are all on the way.

December 1996. The family computer sits in the corner, its beige tower humming. You connect to the internet and the modem shrieks through the speakers.

On the screen, a crude list of online matches appears. One catches your eye: “US-West DM6,” twelve players already in.

You click. The game loads. This time you’re not sitting shoulder to shoulder in a LAN shop, strung together by tangled Ethernet cables. You’re playing against strangers thousands of miles away. I fondly remember how magical it was to join a game and play with other real humans.

QuakeWorld was a software update that turned a groundbreaking shooter into a truly online game. Competition went global. Strategies shifted overnight. An industry worth $5B was seeded (e-sports).

AI Agents are standing at the same moment. Once operating in isolation, they are now beginning to coordinate, negotiate, and delegate across a shared network. We thought this level of coordination would only come in 2026. It showed up early.

Virtuals’ Agent Commerce Protocol (ACP) wants to be the QuakeWorld update for AI agents. A single flow where they find work, strike deals, and get paid.

Here’s everything you need to know 👇

Thanks to HTTP 402 and Ethereum, they can buy APIs, storage, and content all on their own. It’s vending machines for software, with crypto instead of quarters.

They’re giving every U.S. college student free AI training and Career Certificates through Coursera, practically guaranteeing a flood of new users and locking it in as the LinkedIn of AI credentials.

They used clips from a docuseries interview to make them joke that Nakamigos were V0 Punks.

So even if they’re cheaper per token, the wordy overthinking cancels it out. Like hiring a bargain therapist who won’t stop talking.

Loneliness is expensive, but at least it texts back instantly.

In 2025 alone, 337 of them racked up 220M downloads and $221M in spending.