According to BlockBeats, on September 5th, Ember Monitoring reported that a whale who had held ZEC at an average price of $48.44 for over two years transferred 22,840 ZEC (approximately $23.11 million) to Binance 30 minutes ago, resulting in a floating profit of $21.96 million per coin and a 20-fold return on investment.

According to BlockBeats, on September 5th, Onchain Lens monitoring showed that in the past 24 hours, a whale bought 3.23 million PONS with 913 ETH (approximately $2.26 million); 1.69 million CASHCAT with 160 ETH (approximately $392,000); 7,270 AAVE with 1 million USDC; and 156,450 UNI with 850,000 USDC.

According to Beating AI, OpenAI has added a configuration update to GPT-6 Astra. Within the same long dialogue in the Responses API, the reasoning effort can now be increased or decreased at any time without rewriting the already cached Prompt prefix. Difficult tasks can be set to High or xHigh, while simple tasks can revert to Low, allowing the previously used long context to be reused.

Previously, this was a very hidden money pit. Reasoning effort would be the instruction at the beginning of the context, and switching to a new thinking mode midway would cause a large amount of cache to become unusable. A Codex user tested this and found that the cache hit rate dropped from 98.5% to 65% after switching modes, and only recovered to over 97% after rebuilding the cache in the next round.

This change is very useful for agents that run for a long time. When encountering difficult problems, the processing intensity can be temporarily increased, and then reduced back to normal for simple tasks. It prevents the cache of hundreds of thousands of tokens from being reprocessed as normal input just because of a shift in processing intensity.