Ansem on Short-Term Trading: Time Is Also a Crucial Metric for Risk Management

Beating AI News Flash: OpenAI’s 28-day product sprint is now on Day 26. ChatGPT now supports fast account switching on iOS and Android, eliminating the need to repeatedly log out, with up to two accounts able to be logged in simultaneously. Chat history, memory, and subscriptions remain separate, and the desktop version will roll out the feature soon. Tibo also teased in advance that a "banked reset" (rollover credit reset card) will be offered in the final two days. Unlike the previous global reset that directly refreshed the quota, this reset opportunity will be stored in users’ accounts, allowing them to use it manually when their quota is nearly exhausted.

Alphabet, Meta, Microsoft, Amazon, and Oracle—five major tech giants—have pledged to invest more than $3 trillion over the coming years in AI infrastructure development, covering data center leasing, construction, energy, and equipment procurement. Among them, Google (a subsidiary of Alphabet) disclosed related spending pledges exceeding $1 trillion, nearly eight times higher than a year prior; Meta’s future spending commitment stands at over $800 billion, with slightly less than half coming from unlaunched data center leases, and payment terms extending up to 30 years. Data shows the five companies’ combined future leasing commitments total around $1.3 trillion, while other procurement pledges add up to roughly $1.9 trillion. However, disclosure standards vary across firms: some pledges also involve content licensing, consumer hardware, and other business lines, so not all can be classified as data center expenditures. Whether AI infrastructure investments will generate corresponding returns has become a key market focus. Large tech firms like Alphabet and Meta are facing pressure on free cash flow, while Oracle has already slipped into negative free cash flow due to AI data center investments. It plans to reduce upfront costs via supplier financing and customer prepayments, targeting a return to positive free cash flow by 2028.

Beating AI News (from Dongcha): Anthropic has recently added two restrictions to Claude Code’s Opus 5.5: it can develop original games and draw on existing game mechanics, but cannot directly replicate a specific game—even if renamed and presented as an original work. Anthropic engineer Addy Osmani confirmed the change, which is aimed at preventing intellectual property infringement. Independent developers have long grappled with AI’s rampant cloning of their work. In March this year, developer Freya Holmér unveiled a rotated version of Tetris; before its official launch, numerous AI-generated clones appeared. Game studio Digital Sun also canceled its planned public playtest over fears its new title would be preemptively copied by AI. However, distinguishing between legitimate gameplay inspiration and infringing replication is far from simple. Under U.S. copyright law, game mechanics and rules themselves are generally not protected by copyright, while specific expressive elements such as art, text, and code may be. Claude currently relies on prompt-based restrictions to curb cloning, which may not accurately differentiate between legitimate inspiration and infringing replication. Some users have observed that the same game code is permitted when asked to be “rewritten” by Claude, but rejected when instructed to be “replicated”.

Flap's official announcement states the team is still investigating whether the prior DNS attack is linked to its domain service provider, with the root cause of the incident not yet identified. To enhance security and platform stability, Flap has migrated its official website to the new domain flap.cool, and advises users to access the platform directly via this new domain. The project team reminds users not to access the old domain flap.sh, nor to connect their wallets or sign transactions through the old site. While users accessing flap.sh will be automatically redirected to the new domain, the official still recommends directly entering flap.cool to access the platform. The team added that it will continue to update the community as the investigation advances.

According to a report by the UK’s Financial Times citing people familiar with the matter, Blackstone recently told investors in Anthropic’s $63 billion chip financing deal that the AI startup may not immediately secure an investment-grade credit rating after going public, with major rating agencies more likely to assign it a BB-range rating—speculative grade, commonly known as "junk status". Under the rating frameworks of S&P Global and Fitch, a BBB- or higher rating qualifies as investment grade, while BB+ and below fall into the speculative category. If Anthropic ultimately lands a junk-grade rating, its financing costs could climb, and the pool of capital accessible to institutions restricted to investment-grade assets—such as insurers and pension funds—would shrink. The pressure stems largely from Anthropic’s massive planned computing power spending: the company previously disclosed plans to invest over $800 billion in computing infrastructure by 2045. While Anthropic currently holds more than $112 billion in cash and cash equivalents and intends to raise additional funds via an IPO, its long-term computing commitments have left investors worried about its future capital needs and debt capacity. Following the report, U.S. tech stocks extended losses, with the Nasdaq 100 index closing down 3.5%. However, Anthropic has not yet obtained official credit ratings from S&P, Moody’s, or Fitch, so Blackstone’s assessment is not an official rating outcome.

Liquid staking protocol Kintsu, built on the Monad blockchain, announced that its official website kintsu.xyz has been restored to normal and is now pointing to the project’s official infrastructure. However, as DNS caches have not fully refreshed, some networks may still access a malicious site within the next hour. The team reminded users to avoid accessing kintsu.xyz and connecting their wallets until the project issues an official security confirmation notice. Earlier, security firm Blockaid warned that its system had detected an ongoing front-end attack linked to the StepDrainer toolset, targeting three websites: kintsu, flap, and stonkbrokers. Blockaid noted that the affected sites had been injected with malicious scripts; users who clicked "Connect Wallet" could be redirected to a malicious asset theft page. The firm advised users to temporarily avoid using the decentralized applications in question, adding that it is investigating the attack, gathering additional information, and will release further updates.