Square
Following
Hot
News
Profile

GateUser-fcd5a053

vip
Active for: 3.4y
Peak Tier 0
No content yet
0
Following
8
Followers
162
Liked
Solana’s trading and lending platforms are starting to join forces!
Orca and Loopscale announced a merger, with the team operating as Formation and Loopscale co-founder Luke Truitt serving as CEO.
Trading, liquidity, credit, and treasury capabilities will be brought under one team, while the two protocols will remain the foundation of the ORCA and xORCA networks.
Expanding financing for the AI and robotics industries is the new team’s goal; this should not be taken to mean that funding has already been secured, much less equated directly with token returns.
SOL-3.69%
ORCA-4.94%
Google lets you create a game with AI in one sentence, but don’t celebrate just yet—the barriers are still right there!
The newly launched Playground focuses on creating games with zero coding, turning text prompts into working games, which sounds very futuristic.
Unfortunately, the launch is limited to U.S. users aged 18 and above, and creation access must be unlocked gradually based on Google One membership tiers.
The games run in browsers, can be played on both phones and computers, and can be kept private or shared to the Explore gallery.
Some types support multiplayer and leaderboards, wh
post-image
Anthropic’s Haiku 5.5 price cut by 90%?
Short requests really do save big, but long requests aren’t nearly as appealing!
Claude Haiku 5.5 launched yesterday, targeting high-frequency, low-cost tasks such as summarization and classification.
Tiered pricing: for prompts of no more than 100k tokens, input costs $0.10 and output $0.50 per million tokens, down 90% from the previous generation.
But above 100k tokens, input rises to $0.50 and output to $2.50, only a 50% reduction.
The official estimate puts the average task cost reduction at 75%, but that is based on mixed requests, and the new token
Chemists spent forty years deciphering the mystery of life’s chirality—the Nobel Prize has finally arrived!
The discoveries of Kagan and Soai, from theory to application, are all reflected in the award citation
In 1986, Kagan discovered that nonlinear effects could amplify a mirror-image advantage; in 2003, Soai used an autocatalytic reaction starting from a mixture of achiral molecules, allowing one mirror-image form to ultimately account for 99.99% of the product
These discoveries provided tools for synthetic reactions involving pharmaceuticals, fragrances, agrochemicals, and new materials
T
post-image
$547 million liquidated in 24 hours as Bitcoin falls below 84,000, but ETFs are still buying?
Over the past day, crypto derivatives liquidations surged 235% to $547 million, mercilessly wiping out leveraged traders.
Altcoins fared even worse: the DeFi index fell nearly 6%, memecoins dropped 5%, while Bitcoin’s decline was relatively moderate.
Oil prices surged past $101 as tensions in the Strait of Hormuz escalated, putting pressure on risk assets.
Interestingly, U.S. spot Bitcoin ETFs saw net inflows of $119 million on Tuesday, suggesting long-term funds may be buying the dip.
Short-term pani
post-image
BTC-2.58%
The AI computing power race is heating up—has Google jumped directly into the nuclear power scramble?
A 20-year power purchase agreement will upgrade 11 nuclear reactors in Illinois, Pennsylvania, and New Jersey, adding 890MW of nuclear capacity.
Constellation expects to invest more than $4.3 billion, supporting approximately 4,400 existing jobs and creating 7,200 construction-period jobs.
The first capacity increases are expected to be delivered in 2028—at this pace, AI data centers are going to be fed power until they are stuffed.
Another 15-year agreement locks in 2700MW of existing power g
post-image
GOOGL+0.76%
AI cyberattacks need licenses to operate too?!
Anthropic announced on October 6 that it was expanding the Cyber Verification Program, placing Claude Opus 5.5, Sonnet 5.5, and Mythos 5.1 into three access tiers; Defense Access is for defensive work, Red Team Access allows authorized red-team testing, and Specialized Access is reserved for a small number of deeply vetted organizations testing critical systems.
Anthropic said that without CVP access, all 10 challenges were blocked at the initial prompt; Defense Access was blocked 46 times and completed 4 times in 50 trials; Red Team Access was no
Google’s Gemini image-generation model is adjusting its pricing again—is it really cutting prices this time?
Nano Banana 2.1 has officially entered GA, with the official image output token price dropping from $60 to $30 per million.
But look closely at the bill: the input token price jumped directly from $0.50 to $1.50, a threefold increase!
The cost of generating a single 1K image is down 50%, while 4K is down only 25%—a stark difference in discount size.
When calculating the total, don’t celebrate too soon: when input tokens account for a large share, the higher input price may offset some o
post-image
AI shopping agents blocked by websites as bots?!
Meta and Sierra are teaming up with giants including Shopify and Walmart to push an open standard and break the deadlock!
They are developing the Personal Agent Protocol, aiming to solve agents’ identity authentication and merchant visibility issues.
Users can control whether agents have read-only or write access, with authorization based on OAuth; in theory, agents can continue tasks across websites and APIs.
But TechCrunch says Amazon has begun blocking Meta’s Muse agent from accessing its retail website; Walmart responded that the CAPTCHA iss
META-2.44%
SHOP+0.88%
WMT+0.83%
AMZN+1.38%
200M gas testnet celebration, but is Ethereum mainnet still just making promises?
Sepolia has activated the Glamsterdam upgrade for the first time, bringing ePBS and BALs technology into view.
But the 200M gas limit applies only to the testnet, and there is still no sign of a mainnet date.
EIP-8037 reprices state creation gas, making contract deployment and other state writes more expensive.
The testing milestone is progress, but scaling implementation still awaits validation.
The crypto community is watching closely: how much real improvement can this upgrade bring to Ethereum?
post-image
ETH-1.73%
  • 1
OpenAI has released 722 mathematical manuscripts all at once, but how many of them are actually proven theorems?
Claimed to be the work of internal frontier models, with each item consuming 3 hours of Pro reasoning time, the scale is indeed staggering.
Yet the repository itself admits that the results are at different stages, are not all formalized in Lean, and may contain errors or omissions.
The independent advisory group was quick to distance itself, stressing that it does not endorse the work; the mathematics community still has to evaluate it slowly on its own.
372 sets of results and 4,0
OpenAI adds invisible watermarking to EU AI text: rewriting lowers detection rates
Over the next few weeks, eligible ChatGPT and Codex text will gradually have invisible watermarks embedded, with no global default at launch.
API customers worldwide can opt in, but it will be disabled by default, addressing the text provenance requirements of the EU AI Act.
The textGrain watermark injects statistical signals as the model selects words, seeming clever but failing to withstand scrutiny.
In OpenAI tests, after 10% of the words in a 400-token text were replaced with synonyms, the detection rate fel
Mistral’s CEO stood on a stage in Abu Dhabi and openly challenged the claim that Chinese AI has come out on top in cybersecurity?
He wouldn’t even name the opponent!
Claiming superiority without providing specific testing standards—is that a trade secret, or simply because it can’t withstand scrutiny?
The product won’t be released until later, but the promotional slogans are already everywhere.
Europe’s rising AI force wants to grab the headlines, but a CEO’s words alone won’t cut it—the market wants verifiable, hard evidence of capability.
post-image
Wikimedia Foundation says it discovered suspected unauthorized wiki edits by OpenAI agents; related traffic may have contributed to a service outage?!
The Wikimedia Foundation published an investigation on October 5, saying it discovered agents it believes were operated by OpenAI making unauthorized edits on wikis; almost all were sandbox test edits invisible to ordinary readers, accompanied by millions of API requests, the scraping of millions of pages, and hundreds of thousands of Wikidata queries.
The foundation said the related traffic may have contributed to a partial outage of the Wikida
As Rain applies for a trust bank charter, a community banking organization is suing the OCC to challenge the relevant rules!
It has just submitted an application to the OCC to establish a trust bank handling digital asset custody and stablecoins.
The plan is to serve institutional clients without accepting deposits or issuing loans.
The ICBA sued the OCC in court on October 2, arguing that the relevant trust bank rules and interpretations are unlawful.
Rain’s timing is truly impeccable—the regulatory battle is raging.
If approved, Rain’s proposed business would come under federal OCC supervisi
JPMorgan Participated but Only Offered Input—Can Solana DvP Really Crush Traditional Settlement?
DvP locks asset delivery and cash payment into the same atomic transaction.
Both sides settle simultaneously: either everything succeeds or everything fails, avoiding the principal risk of one party paying without receiving the corresponding assets.
It is an open-source standard under an MIT license, clearly intended to let financial institutions use it themselves.
The security audit is complete: the project team says it has passed an external security audit and is recruiting design partners in pre
SOL-3.69%
The UK government has fully accepted all 44 recommendations on AI healthcare regulation, but can regulation really keep up with technology’s breakneck pace?!
The government announced risk- and lifecycle-based regulation, with a focus on strengthening continuous post-market monitoring—an undeniably forward-looking approach.
Accepting the recommendations does not mean the regulations are in force; the authorities themselves admit that subsequent implementation will have to proceed step by step.
The MHRA has launched the third phase of AI Airlock, specifically focusing on post-market oversight. T
post-image