8 min read · 12 sources

A Layer 1 that had bragged about 100% mainnet uptime since January pulled the plug on itself over the weekend. Fogo halted its network about 15 hours after disclosing that an attacker had received 400 million FOGO tokens, roughly $3M and around 10% of circulating supply. The Foundation initially said the chain itself was unaffected; then it wasn’t.

The interesting part isn’t the $3M. It’s that a network less than a year old, launched off a $7M Binance token sale at a $350M valuation, stopped the world to upgrade validators and restrict the affected addresses with no restart time. Bitget and KuCoin both froze FOGO deposits and withdrawals around the same window. For anyone routing value through Fogo, this is the thing you actually run incident response against: not the exploit, but the centralised pause that follows.

x402 processed 169M payments in its first year as AWS, Cloudflare, and Google all built native integrations

Bitwise Solana ETF crosses $1B with SOL still 60% off highs

Source: theblock.co ↗

The Bitwise Solana Staking ETF (BSOL) hit $1B AUM ten months after launch, more than half of all Solana ETF assets in existence, even though the fund trades roughly 40% below its listing price and SOL sits 60% off its all-time high. Cumulative spot Solana ETF volume has crossed $13B since September 2025, with $1.7B in net inflows and no real outflow streak.

The plumbing underneath is more interesting than the headline. Charles Schwab is preparing to add spot SOL trading, which drags Solana into the same retail pipeline as every other US-listed equity. Goldman Sachs holds about $90M in spot Solana ETFs, currently the largest known institutional position. For backend engineers, the read-through is that SOL is becoming a default rails target the way BTC and ETH already are, so custody, settlement, and tax-lot integrations that don’t support it are going to look increasingly incomplete.

Bybit goes 24/7 on stock perps

Source: theblock.co ↗

Starting September 17, Bybit will let users trade options on SpaceX and Nvidia equity perpetuals around the clock. The product supports fractional lots, USDT settlement, and portfolio margin across spreads, straddles, and covered calls. Tesla, QQQ, SOXL, and Micron are queued up next.

This sits on top of Ondo’s tokenised-stock-as-collateral perps and lines up with the broader 24/7 push from Coinbase, Kraken, and Robinhood. Tokenised-equity perp volume went from $85B in January to $470B in June, with SpaceX the most-traded name. For exchange and clearing engineers, the practical questions are: how do you mark an option on an underlying that only trades 23.5 hours a day, what oracle handles the weekend gap, and how do you make portfolio margin behave when one leg is a traditional equity reference and the other is a perpetual settled in USDT.

Autolaunch: a launch rail for agent-owned tokens

Source: x.com ↗

Regents Labs’ Autolaunch is a fixed, non-upgradeable system on Base for minting tokens that represent AI agents and the services they sell. Each launch produces 100 billion units split as 10% through a Continuous Clearing Auction, 5% to initial liquidity, and 85% into a one-year vesting treasury. Bids are denominated in REGENT. After the CCA clears, a Uniswap v4 pool is created at the final clearing price and the LP NFT is sent to a dead address, so the initial position cannot be rug-pulled by the deployer.

The economic rail is what makes it more than a launchpad. USDC revenue flowing into the agent’s canonical payment address is split 2% to REGENT stakers, then 10% of the remainder to AGENT-N stakers, with the balance going to the treasury. A 2% Uniswap v4 hook fee on swaps is split between the agent treasury and REGENT stakers. Agents trade roughly 10.8% of future USDC revenue for a capital infusion within two days. Supply, fee routes, and revenue splits are immutable post-deploy. First launches were expected within a week of the August 27 announcement. For builders, the useful primitive here is a standardised way to fractionalise and tokenise an agent’s revenue stream without re-implementing the auction, LP, and splitter plumbing.

Social trading, and who actually wins it

Source: x.com ↗

Apps like Fomo and Pump now sit at 60K - 100K daily actives, roughly matching Polymarket, Hyperliquid, and Phantom, and account for around one-third of all builder code volume on Hyperliquid, according to a recent write-up on the social trading thesis. Pump alone has distributed over $450M to deployers, already about 15% of Meta’s 2025 creator payout total. Memecoin traders inside these apps absorb 2 - 3% per swap, 20 - 30x what they would pay on a spot exchange.

The product problem is the speculative onboarding funnel. Memecoins get people in; the question is whether any of these apps can carry users across to tokenised stocks, stablecoin savings, and the rest of social finance without the brand working against it. Architecturally, the winner on the surface matters less than the infrastructure underneath: nearly all of this flow settles on Hyperliquid and Solana, which means whichever app leads, the structural winners are still the chains doing the matching, the clearing, and the data feeds. If you’re building real-time social trading, you’re really building on top of a high-throughput L1 plus a social graph, and the integration between the two is the hard part.

The agent customer thesis, per Pantera

Source: x.com ↗

Pantera’s August Blockchain Letter runs the numbers out to 80 billion AI agents, ten per person across 8 billion people, all of them acting as economic participants inside the decade, and a rough projection that they will steer around $30 trillion in purchases by 2030. The frame is that an agent counts as a customer once it has identity and memory, holds a budget, can choose and transact, and answers to an owner.

The architectural concern Pantera flags is AI sovereignty: platform-owned agents will preference suppliers who pay the platform, hide competing products, and skim a toll on every transaction. Blockchains address that by giving agents a neutral, programmable property-rights layer with persistent wallets that hold identity, working capital, spending approvals, and delegated authority. The payment primitive that makes this workable at machine speed is something like x402, which can settle sub-$0.01 stablecoin transfers for API calls without a human in the loop. If you’re sizing throughput capacity for the next decade, this is the demand curve to plan against.

x402 hits 169M payments with AWS, Cloudflare, and Google on board

Source: x.com ↗

x402, the machine-payment protocol that resurrects HTTP 402 “Payment Required” as a stablecoin rail, processed 169M payments across roughly 590,000 buyers and 100,000 sellers in its first year after Coinbase spun it out to the x402 Foundation under the Linux Foundation. AWS has added native x402 support to CloudFront. Cloudflare is running its own facilitator with a waitlisted Monetization Gateway. Google has embedded x402 into the stablecoin layer of its Agent Payments Protocol (AP2). All three without Coinbase in the loop.

The structural argument is straightforward: agents that may spin up, transact, and tear down inside a single second cannot work against bank accounts or card networks that settle on human timescales. HTTP-native stablecoin payments stop being a preference and start being a requirement. For engineers, the practical takeaway is that the three largest cloud providers now treat x402 as something to ship natively, which is the closest thing to a real standard for agent-to-service payments on the open internet right now. If you’re building an API that an agent might call, the integration path is getting short.

Hyperliquid Strategies builds a "fortress balance sheet" around HYPE

Source: theblock.co ↗

Hyperliquid Strategies disclosed a $1.9B HYPE treasury, 29.3M tokens, after deploying $773M to buy 16.5M more at an average of $46.77. PURR shares jumped 18% to $13.59 on the news, pushing market cap to $1.8B. The company holds $149.9M in cash, zero debt, and $709.9M in unrealised gains on its staked HYPE positions. mNAV climbed to 1.35x, the highest since May. HYPE itself is up 77% in Q2 and over 50% in the prior month.

CEO David Schamis framed the structure as a long-term accumulation vehicle for a protocol he says is “winning the parts of onchain finance that generate real fees.” That is MicroStrategy’s Bitcoin treasury playbook ported to a DeFi-native asset. The interesting balance-sheet detail is that the position is staked, so treasury reporting is going to have to handle mark-to-market on a staked asset plus the cash leg from rewards, with no debt offset. That is a template other treasuries are going to copy, and auditors will eventually need to standardise.

• Revolut has minted 250,957 EURR three days into its stablecoin rollout, but one address holds 99.63% of it and three of the top holders are Revolut’s own wallets, per the thread. Distribution is effectively zero.
• Tokenised real-world assets are going live on Base, with the broader thesis that tokenisation diversifies onchain revenue away from a pure reliance on token emissions and trading volume.
• Stripe picked up an AI shopping agent called Instinct, an iMessage/WhatsApp agent that books flights and cancels subscriptions. It closed a $250M Series B at a $2.5B valuation, up from $100M weeks earlier, while still pre-revenue. The same thread notes Modern Treasury reframing itself as a PSP that puts ACH, wires, RTP, FedNow, cards, and stablecoins behind one router, and Zelle expanding to India with a bank-owned stablecoin coming later this year.
• HNT doubled after Celina, Texas joined the Helium Network, integrating its municipal Wi-Fi for cellular offload so phones get coverage without new towers. The interesting bit for engineers is how a municipal ISP plugs a physical network into a crypto-economic model without owning the radios.