Cardano will hand over control of its core software to an external team starting in August.

Recently, Kalshi’s crypto perpetual contract trading data has faced widespread skepticism from the community. The controversy began with crypto perpetual contracts: a trader noted that ETH perpetual open interest is only around millions of U.S. dollars, yet daily trading volume reaches hundreds of millions of dollars, with unusually high turnover. Kalshi’s head of crypto business, IcoBeast, responded by addressing the perpetual contracts and its crypto prediction markets separately, stating that prediction markets have no equivalent market maker rebates and their statistical methodology is comparable to industry peers. However, community discussions did not stop at perpetual contracts; suspicions of fraud spread to Kalshi’s core prediction market business. Prominent prediction market trader @retardmode posted that roughly 61% of Kalshi’s disclosed trading volume comes from "multi-event parlay bets" (called "串关" in Chinese), which bundle the outcomes of multiple events into a single ticket that requires all selections to be correct to win. The platform does not calculate volume based on users’ actual wagers, but instead uses a $1 par value per contract. For example, a user spending $1 on a parlay ticket that pays out $14.10 if all selections are correct will be recorded as $14.10 in trading volume. Nearly half of these parlay bets involve 11 or more events, resulting in extremely low win probabilities. Using this methodology, Kalshi’s actual trading volume yesterday was around $136 million, while its publicly disclosed figure was $1.91 billion. Top Polymarket trader @CarOnPolymarket further estimated that the actual trading volume of Kalshi’s parlay bets may only be roughly 7.1% of its disclosed parlay bet amount. Extrapolated monthly, this could create an extreme gap of approximately $4 billion in actual volume versus $57 billion in reported figures.

BlockBeats and Hyperliquid Chinese Channel officially launch the HYPE Community Rant Session. The 7-day event runs from September 21 to September 28, where users can win HYPE rewards by sharing their trading stories. Eligible topics include: - Who missed out on Robinhood’s $PONS "wealth train"? - Who got liquidated while holding positions when crude oil was more volatile than BTC and ETH in the middle of the night? - Back this time last year, $ZEC was only $40... Did you catch the privacy narrative? - $HYPE and $LIT hit new highs—are you still holding $ASTER? The event has three reward tiers, with the top prize being 9.99 HYPE. To participate, users must quote the event’s promotional content, post original topic tweets, and follow the Hyperliquid Chinese Channel. The event is exclusively sponsored by ecosystem project Aqua, and supported by partners TradingBeats, CoinAnk, HyperEVM Chinese Station, and Hyperliquid Madhouse.

Insight Beating AI Flash News: Anonymous math community account Captain Sude has released a new proof discovered by GPT-6 Astra, solving the Liouville version of the Goldbach problem. This problem relaxes the "two primes" condition in the classic Goldbach conjecture to "two integers with an odd total number of prime factors". Previously, Durham University mathematician Alexander P. Mangerel could only prove this under the assumptions that the Generalized Riemann Hypothesis holds and the even number is sufficiently large. Astra has now removed both constraints, proving that all even numbers greater than 2 satisfy the condition. The core approach is to first assume that a given even number cannot be split in this way, then derive contradictory results step by step. The full proof has been written in Lean 4. The project compiles successfully, and independent audit repositories have successfully reproduced the result, with no `sorry` statements or additional mathematical axioms found. The classic Goldbach conjecture itself remains unsolved, as the two addends here can still be composite numbers.

According to Bitget market data, Samsung Electronics’ stock price surged 5%, marking its largest single-day gain in two weeks and leading the Korea Composite Index. Note: Samsung Electronics’ dividend equity registration date is September 30.

Kalshi’s Head of Crypto Business, IcoBeast.eth, published a post addressing external doubts about two key issues: alleged fraudulent trading volume in Kalshi’s crypto prediction markets, and claims regarding its perpetual contracts’ Self-Clearing Members (SCMs). He clarified that Artemis charts display prediction market trading volume shares, not perpetual contract trading volume. Kalshi does not offer rebates for its crypto prediction markets, and the number of contracts and notional amount calculations in its prediction markets align with those of other platforms, making them comparable. Regarding the claim that “SCMs are market makers selected by Kalshi,” he stated this is untrue. Any institution meeting regulatory requirements can become a Self-Clearing Member of a CFTC-regulated exchange, and “fair access” is a core regulatory mandate. He also noted that all exchanges use rebates and incentive programs to boost liquidity, including CME, Hyperliquid, and Binance. On Kalshi’s perpetual contract business, he admitted it is still in its early stages, but emphasized that unlike offshore perpetual contract trading platforms, Kalshi is required to publicly disclose its incentive programs, so all related incentive arrangements are transparent and accessible.

Flash News from Beating AI: TypeSafe AI announced that Jev is now fully open to all users, with the waitlist eliminated. Users can register and use it directly via the official Console, no longer needing to apply for Early Access. Jev is not designed for chatting or code writing; it specializes in classification, scoring, and judgment. Developers provide candidate options in advance, and Jev directly returns the selection, probability, and confidence level, enabling programs to proceed using the results. The official pricing remains $0.042 per million input tokens, while outputs are free of charge. Before its full launch, Jev had already spread rapidly across third-party platforms. Vercel reported that within 24 hours of integrating Jev into its AI Gateway, nearly 13% of paid teams adopted it, making it the fastest-adopted new model in AI Gateway history—its adoption rate is twice that of the GPT-5.6 series over the same period.