Projects still being delivered in a bear market are worth a second look.
DeAgentAI’s new official website is live: moving from “trusted infrastructure” to the “protocol layer + economic layer.”
Focus on two things.
First is the AI Agents platform.
The pain points of on-chain agents have never been that the models aren’t smart enough—it’s that getting started is too annoying and operations are too heavy. DeAgentAI directly removes this layer: connects a wallet, deploys in 60 seconds, never touches private keys, and requires no KYC.
The first Agent, “Sentry,” is positioned very conservatively. It doesn’t give advice or scores. It scans Hyperliquid; when the settings’ rules are matched, it provides a prompt—leaving the decision-making to users.
Four strategy sets are configurable, filtering out any zero-token consumption. Build solid tools first, and only later, if there’s an opportunity, move into trading.
Second is the value flywheel.
Technical capabilities begin monetizing externally, and the Agents platform brings real usage.
The official clearly states it will use profits to buy back and burn $AIA .
‼️ Special Event $USD1 ‼️ > August 4 – August 8, 5 consecutive days > Binance Square > Rewards: 20,000 USD1 + 600,000 $WLFI
Details ⬇️
Join in and wait: Binance Square Chinese-only community (the only one): https://app.binance.com/uni-qr/YbCEQcQf
What’s coming then: 🎯 Random surprise livestream room rewards As long as you’re streaming content related to WLFI / USD1—discussion, trading share, chart analysis all count—I might just walk right in and drop a reward. Friends watching together can also get红包 (red packets) to share. 🧧 Red packets drop in the chat every day CN and EN chat rooms are already set up. Red packet codes drop from time to time—only people who are there can grab them. Join in and wait. Binance Square Chinese-only community (the only one): https://app.binance.com/uni-qr/YbCEQcQf
Clarity Act could be approved soon, and stablecoins are set to see another wave of growth.
During the promotion period of $USD1 , results will be immediate. After listing BTC and ETH, Binance has now also launched $SPCX -USD1 perpetual contract pairs, further expanding trading scenarios.
When policy implementation + first-mover advantage + exchange support come together, USD1 will feel great.
After having Binance, industry ETFs + leverage become very simple. This is much more useful and safer than spending half a day researching an industry chain to pick a stock. If more A-share listings could be available on Perp, that would be even more perfect. $MU $QQQ $DRAM
$USD1 is continuing to top up in Binance activities Total prize pool of 165 million $WLFI tokens If you need a 1.2x annualized return, you need a daily open contracts volume of 300 USD
The real value comes from the people standing behind it July 1st—Trump’s financial disclosure: In 2025, crypto revenue exceeded $1.4 billion, about $800 million of which came from WLFI. On June 19, WLFI filed an application with the OCC for a national trust bank license. The USD1 target isn’t speculative tokens—it’s to break into the top ranks of compliant stablecoins.
Three lines are running at the same time: airdrops to build the habit → deepen demand through scenarios → build trust with political capital + regulatory licenses. A full market-engineering play.
What to do now Today the rules just changed. Use USD1 for contracts/leverage: open a position, and hold it all day with at least 300 USD1 to receive a 1.2x bonus.
--- A 37x gain in one year isn’t luck—it’s nailing the timing The five rounds of prize pools rolled from $40 million to 178 million WLFI, with clear targets for each round Today the rules changed: to get the bonus, you have to maintain your position Trump’s family’s $800 million came from WLFI, and an OCC license is already on the way
The early birds aren’t eating luck—they’re eating timing plus execution. $USD1 🚀
$USD1 In one year it rose 37x—it's not luck, it’s a carefully timed schedule
The stablecoin war has already begun. Most projects are running “naked,” propped up by marketing. USD1 is different—every subsidy is locked to the cheapest possible timing.
Launched in March 2025, reached a market cap of $125 million in April, and now stands at $4.6 billion. 37x in one year.
Five rounds of airdrops—five goals Round 1: Jan 23 — $40M WLFI airdrop. From the start, it covers every scenario, so the faster you build the habit of “holding USD1,” the better. Round 2: Starting Feb 20 — momentum never stops; lay a solid foundation. Round 3: Mar 20 — the pool triples to 135M WLFI. Round 4: Apr 17 — launch the main USD1 trading pairs. BUSD collateral conversion, with VIP fee updates following. Round 5: Jun 12 — the pool reaches 178M WLFI, the largest in history. Bybit/Gate also join.
Right today, the rules changed New UTC rule on July 3: Contract positions must stay at or above 300 USD1 throughout the entire day. Snapshots are taken hourly to determine the lowest value. Add-on boosts → open positions and hold → stack positions → grow depth → attract capital → upgrade competitiveness.
Result: At key price levels, the BTCUSD1 order-book depth surpasses BTCUSDC. 0 Maker fee + ultra-tight spreads + position profits in WLFI. Market-maker appeal completely outclasses the competition.
Why are subsidies the smartest right now? When the war hasn’t started yet, customer acquisition is cheapest. Now 1U equals 10U at the time the stablecoin battle erupts. The window is closing.
OpenClaw gave model companies their first taste of the Token economy—10.4 trillion calls in a single month, income trajectory skyrocketing.
On June 17th, GLM-5.2 went open source. This time it's different.
MIT License: modify as you please, sell as you like, intellectual property is zeroed. Companies can embed it into commercial products without needing to share modifications. Compared to the GPL contagion license, it’s like tearing down the barriers.
What’s more crucial—AI has now entered the long reasoning era.
Before, it was a sprint. Each task had a limited window; planning and execution ended quickly, KV Cache was manageable, and the pressure was on compute power.
After GLM-5.2, it’s a marathon. 1M lossless context, holding all code, decision history, and constraints in a single task. Tests show a single run processed 880,000 tokens, nearly maxing out the window.
Agents are no longer just Q&A machines. They break down goals → multi-round planning → iterative validation → tool adjustment → write and run code → replan based on feedback. A task triggers hundreds of reasoning loops, turning Token consumption from linear to exponential.
With each loop, the complete context is loaded into memory for recalculation. Continuous computation, continuous communication, continuous read/write. These three continuities fundamentally change the pricing logic of hardware.
What benefits from long-range Agent reasoning?
HBM KV Cache scales linearly with rounds and context, quickly filling GPU HBM. Once off local, bandwidth drops from TB/s to hundreds of GB/s. The issue shifts from compute power to memory bandwidth. The big three manufacturers are sold out, with a gap of 50%-60%, and the market is projected at $54.6 billion by 2026.
Optical Chips/InP Long-range reasoning running clusters creates terrifying inter-card communication. The optical module market is expected to grow by 60% to $26 billion by 2026, with an InP substrate gap of over 70%, and indium prices rising by 90% year-on-year.
CPU Long-range scheduling, task breakdown, and KV Cache management rely on CPUs. The GPU ratio is moving from 1:8 towards 1:1. Intel's CEO mentioned multiple company CEOs calling to expedite deliveries.
Liquid Cooling Long-range sustained loads, the same card consumes 3-5 times more power than short reasoning. Rack power consumption jumps from 36kW to 200kW. Air cooling can't handle it.
Switches Bandwidth demand for reasoning clusters jumps from 100G to 400G, with hundreds of thousands of cards needing coordination, benefiting both IB and Ethernet across the board.
ABF Substrate Clusters expand from thousands of cards to tens of thousands, with each chip needing packaging. Ajinomoto monopolizes over 90% of ABF film, and a 42% shortage is expected by 2028. Flour prices are rising, and bread will only get more expensive.
CCL M9 All motherboard backplanes require high-speed materials. The unit price of M9 is ten times that of FR4, with the market projected at $18.7 billion by 2027, growing faster than optical modules.
OpenClaw lit the fire, GLM-5.2 added the fuel. The former helped model companies make their first Token haul, while the latter took the market from the lab to industry.
Stand in storage, stand in light, enjoy the bubble. $MU $SKHYNIX $LITE
JPMorgan Warns: Next week's rebalancing of capital flows could trigger a potential $165 billion global stock sell-off
$BTC $QQQ
$165 billion isn't a prediction. It's an estimation.
Global stock long positions are too heavy, and the rebalancing model indicates that this much needs to be sold to neutralize exposure. It's not panic; it's mechanical trading. Positions will reset after the holidays, unrelated to fundamentals.
However, the $165 billion sell pressure—mechanical or not—will put short-term pressure on high-beta assets like $QQQ . The Nasdaq has seen significant gains this year, and unrealized profits are high, making sell pressure most concentrated during rebalancing.
$BTC is more nuanced. When liquidity-sensitive assets experience a liquidity drain from stock rebalancing, BTC has never been a safe haven; it's a correlated asset.
One-week window.
If we can hold out, everything will be fine. If we can't, we're looking at the next support level.
**Rebalancing doesn't kill logic; it kills positions.**