A manipulated pool fed the Nostra prediction machine a fake NSTR price—an amount that was 8,306 times the true value—unlocking a $3.5 million loan on Starknet.
PragmaOracle said it had warned in advance that the feed price was high risk, and Nostra did not respond. The lending market is still paused, and there has not yet been a post-event review.
For anyone borrowing and lending on Starknet, the direct cost of relying on a single feed price is this: the pool’s liquidity can be compromised, and the price can then be written into contracts. In short—unless Nostra publicly discloses the source and limits of the feed price, even if the lending market restarts, it will be difficult to restore deposit intentions to the same scale.
$FIGHT 24h rose 24.7%, yet the open-position amount over the last 1 hour fell by 6.55%.
The price dropped from 0.004527 to 0.004408. Trading volume is 2.74 times the 24h hourly average. The buy/sell ratio of takers is 0.945, slightly sell-leaning—more like longs taking profits than new money entering. The 15-minute buy/sell ratio has returned to 1.144 and the price has rebounded slightly; selling pressure doesn’t appear to be continuously increasing. The funding rate at 0.0383% is still slightly positive, meaning there is still some room left for longs to crowd. If it were to reclaim 0.004527, the “deleveraging/减仓” interpretation would not hold.
$SUPER also reduced positions after a rise, and the funding rate turned negative; $MUBARAK fell 16.22% while shrinking volume and reducing positions—this is longs exiting.
#North Korean Hackers Pretend to Interview to Steal Wallets
Seven organizations jointly announced that WaterPlum and the North Korean remote IT worker program have been grouped into the same operation: at least 30,000 devices in more than 100 countries were infected, funds or credentials from more than 7,000 wallets were taken, and about $11 million flowed to Pyongyang.
The most directly affected are developers and designers who are currently applying for Web3 jobs—this fake interview makes you download files to “fix video glitches,” and the malicious code is inside. The falsifiable test is simple: if the interviewer keeps insisting that you install software from unknown sources, that clue alone is worth treating as a risk signal.
#AI stocks keep rising—what other investment opportunities exist IonQ announced that its latest quantum computer, Superion 256, will become the first system deployed by the NVIDIA Quantum Research Center.
Its role is not a standalone lab device; instead, it will function as a dedicated coprocessor working alongside NVIDIA GPU clusters. Quantum tasks run on CUDA-Q, while classical computing and AI are handled by GB200.
The real impact of this is in the compute narrative: beyond AI hardware, quantum is being placed into the same underlying infrastructure. Deployment timing hasn’t been disclosed yet, so treat this as a directional signal—not a progress update.
#AI shares continue to rise—what other investment opportunities are there? AI’s “bill” is shifting from GPUs to storage.
TrendForce on September 23 said that AGI and Agent AI are driving North American cloud service providers to urgently place additional enterprise SSD orders. A structural shortage has emerged in high-end storage; based on the current procurement pace, the order volume in the fourth quarter may further exceed the already elevated third quarter.
The consumer side is affected: PC and smartphone demand remains weak, and capacity is being prioritized for higher-margin enterprise products. In our view, this round of shortages is more likely to continue beyond Q4; if the consumer side recovers at the same time, the divergence may then narrow.
#AI stocks keep rising—what other investment opportunities are there AI money is spilling over from the model layer into the data layer and the hardware layer.
Snorkel AI has completed a $350 million Series E funding round with a valuation of $3.5 billion; its annualized revenue Run Rate for this year reached $375 million, up from about $20 million a year earlier. Complex training data has evolved from bundled services into a business that can be charged for independently, giving data companies their own pricing power.
Assessment: this wave of AI spending looks more like a continuation than a pullback—so long as orders for “shovels” like data and storage continue to move higher, the cost pressures faced by those capital-intensive players like $META $MSFT $TSLA will be discussed again and again.
#AI stocks continue to rise—what other investment opportunities are there?
The credit spread on bonds issued by AI-related issuers is currently about 115 basis points, roughly 1 notch higher than the overall investment-grade market at about 78 basis points. Goldman Sachs expects that hyperscalers may issue a record $420 billion in debt by 2027.
Rising financing costs are squeezing visibility of returns, not defaults. This is a clue that is more favorable to top players who have cash flow in hand and do not need to rely on issuing bonds to support CapEx.
#ETF Money Flows At the beginning of the year, it was still seeing net outflows, but after August 19 it turned around to record inflows of about $4.6 billion.
According to Bloomberg, this year the cumulative net inflow into US-listed spot Bitcoin ETFs is about $320 million. Specifically, since August 19, when the US Treasury said it would increase long-term bond repurchases, inflows have amounted to about $4.6 billion, reversing the net outflows seen earlier in the year. Over the same period, Bitcoin rose by about 35% to above $86,000.
For those holding spot positions, the source of buy pressure is shifting from leverage to the creation/redemption channel. One point that can be falsified: if the ETF weekly data turns negative again afterward, the rebound logic supported by these liquidity flows will need to be reassessed.
Binance now supports ZIL deposits and withdrawals on the Zilliqa EVM network, and $ZIL on the legacy mainnet will be migrated 1:1, with the exchange handling the process. The reason is that a Ledger app flaw exposed 6,772 accounts, and at least 683.13 million ZIL were stolen, leading the project to retire the legacy transaction system. For those holding on Binance, this is more convenient; for self-custody users, it is another ZKP migration process. Verdict: once old-chain addresses have completed mapping, deposits and withdrawals on the legacy network will not resume.
#AI stocks continue to rise—what other investment opportunities are there?
Sam Altman is expected to push for the development of AI standards during a speech at the United Nations. OpenAI’s position is that it does not support establishing a specific international regulatory body; instead, it favors baseline safety standards with democratic oversight.
Who sets the standards, and what exactly the standards are for, matters more for determining pricing power across compute and data layers. This line is unlikely to directly hit coin prices in the short term—if you’re not sure, don’t rush.
#AI stocks continue to rise—what other investment opportunities are there?
Firmus, backed by Nvidia, plans to raise about $10 billion by combining debt and equity to procure Nvidia chips for its data centers in Indonesia; the same company also plans to conduct an IPO in Australia by late October, potentially raising up to $5 billion. The money flows first into compute power, then into chip orders—within this chain, the beneficiaries are the ones selling cards, not the ones issuing tokens.
Compute-power financing continues to ramp up; in the near term it looks more like a continuation than a pullback. But if these purchases fail to materialize for a long time, the narrative will cool before the orders do.
Ivan Chebeskov, Deputy Minister of Finance of Russia, said that tax residents must report to the Federal Tax Service any crypto transactions conducted outside the regulated framework; if a foreign issuer freezes $USDT or $USDC and the non-Russian depository institution can be controlled, the loss is borne by the investor.
This is a real cost transfer for Russian tax residents holding stablecoins: compliance obligations are pushed forward, and freeze risk remains with the holder. The point that can be disproven is this—if later the official regulatory text shifts the frozen-loss back to the depository party, then this stance would only be a transitional interpretation.
#AI stocks continue to rise—what other investment opportunities are there Binance Spot and Convert have listed three bStocks trading pairs: Axe Compute (AGPUB), AMC Entertainment (AMCB), and Cypherpunk Technologies (CYPHB). The announcement does not specify the exact time of availability.
Who is affected: For people who want to directly trade AI and U.S. stock themes within their Binance account, this provides an additional channel. However, the underlying assets are not technology blue chips; the scale of Axe Compute and Cypherpunk Technologies is far smaller than mainstream names like $PLTR , $META , $TSLA .
Assessment: For tokenized stock trading pairs like this, trading depth in the short term will most likely still be concentrated in just a few names, and order books for less popular listings may be very thin. If, within a week, the order books for these three pairs don’t pick up, it means demand hasn’t followed through.
#Federal Reserve Raises Rates After the rate hike is implemented and the CLARITY vote fails, Bitcoin doesn’t keep sliding lower—it instead returns above $86,000. BitGo’s view is that it has been “digested.”
For leveraged longs, this feels more like a stress test passed: with no fresh lows after bad news has played out, it suggests selling pressure may have been released earlier. But “absorption” doesn’t mean “a turn.” If BTC falls back below $86,000 again, this assessment would be falsified.