Bloomberg: Binance theft incident may not receive insurance payout, insiders say it did not take out cyber security insurance.

Beating AI Express (Dongcha) reports: Hume has launched the Audio Decision API, which enables direct queries of audio recordings. Supported questions include "How many people are speaking?", "Did the customer service agent interrupt the client?", and "Does the client sound agitated?" The tool can also pinpoint the exact timestamps of laughter and sighs. Users can submit up to 25 queries at a time; the model delivers judgment results and probabilities without prior transcription of audio content. Previously, analyzing metrics like number of speakers, emotional shifts, interruptions, and background noise in a call required combining multiple specialized models, with each new analysis task often needing adjustments to the entire workflow. Hume centralizes these judgments into a single model that processes audio directly to answer queries. Developers only need to modify their questions to analyze new audio features and conversational behaviors, eliminating the need to retrain models or design custom classifiers. A single API can cover a wide range of distinct audio analysis tasks. The model is trained on over 10 million human judgment data points, aiming to predict how average people would react to the same audio. For example, a 70% "agitated" output indicates approximately 70% of listeners would perceive the speaker’s tone as agitated. In official preliminary tests, the model’s agreement rate with multi-reviewer consensus reached 70.7%, outperforming the 56.5% average of individual human raters. Hume reported an average score of 82% across 10 audio benchmarks, higher than Gemini 3.8 Flash’s 78% and Qwen 3.8-Omni-Flash’s 76%. The API is currently available via permission application only, priced at $0.47 per million input tokens. Per official examples, analyzing a 5-minute call costs roughly $0.012. It supports audio clips from 5 seconds to 5 minutes per request; real-time analysis requires submitting short audio segments sequentially.

Binance founder CZ said in an interview, "Security breaches always seem impossible until they happen, right? But often, right before an incident, people overlook key details. Therefore, every security incident is actually preventable with sufficient effort and attention. I believe we can learn many lessons from these cases. Without sufficient technological application and market uptake, the tech sector may struggle to secure adequate investment. Thus, market demand must grow in tandem, and the two must reinforce each other. We cannot wait until technology reaches 100% maturity to act, as technology will never be absolutely mature—it is always in a state of continuous improvement. AI technology is advancing at a rapid pace, but it also brings many challenges, including security risks. In the past, we might see a single hack every month or even less frequently, but now they can happen weekly. Just yesterday, four cryptocurrency exchanges were hit by separate hacking incidents. I predict that the frequency and speed of hacks will continue to rise in the future, as bad actors increasingly leverage AI. At the same time, we must also make greater use of AI technology to strengthen defenses. To counter a powerful adversary, you often need to possess comparable capabilities. Therefore, we must build our own AI-powered capabilities to address these threats. There is no so-called 'perfect time' to wait until technology is fully mature and all security issues are resolved before acting. Technology and security defenses are always evolving."

Binance founder CZ said in an interview: "If we cannot contain such security attacks, if similar incidents keep happening without noticeable improvements to security protections, market adoption will slow down, which will in turn affect the entire crypto industry’s mainstream adoption process. Overall, I believe any new technology needs an adaptation period in its early development stage, and this process often comes with growing pains. But in the long run, I am confident positive factors will ultimately prevail, and the technology will continue to advance over time. I don’t know when the industry will hit bottom or reach a turning point. I think we may face more problems before the overall situation starts to improve. I hope that day comes soon. We hope the U.S. government can provide help in this regard. I believe that whether it is me personally, Binance as a company, or the entire crypto industry, we very much hope to receive support from the U.S. government. If anyone can communicate with the North Korean side, it is likely the U.S. government."

According to an official announcement, Clearpool will open the migration of CPOOL to CLEAR on November 7, with the deadline set for February 17, 2027, and a 1:1 migration ratio.

According to monitoring by Onchain Lens, a Hyperliquid trader holds two 10x leveraged long positions in crude oil: Brent crude oil worth approximately $91.77 million, and WTI crude oil worth approximately $56.62 million, with a total position size of around $148 million.

Beating AI News Flash: Revolut, Europe’s most valuable fintech unicorn, has made Zhipu GLM the default AI coding model for around 2,000 of its engineers. Tools like Claude Code and Codex have been demoted to secondary options, reserved primarily for complex tasks. David Abitbol, Revolut’s head of AI operations, revealed that the company previously allocated a $300 monthly per-user quota for Claude Code, but it delivered limited overall efficiency gains. In April this year, the company began self-deploying GLM and opened unlimited tokens to its engineers. Currently, Revolut uses around 56 NVIDIA B200 GPUs to run GLM 5.3, with a combined cost of approximately $0.30 per million tokens—just one-tenth of the bills the company previously paid to Anthropic and OpenAI. According to internal company statistics, engineers’ coding time metrics improved by around 60%, and end-to-end task delivery cycles improved by approximately 70%. In the past, limited tokens led engineers to reserve them only for complex tasks. Now, with no need to conserve tokens, they are using AI extensively for internal document searches, Jira ticket processing, and Slack requirement clarification. Only when tokens are cheap enough to avoid careful budgeting can AI transition from handling a small number of complex tasks to integrating into the entire R&D workflow.