ideo

AI startup Tavus released a new model called Griffin on October 1, referring to it as the first "human interaction model" that can understand and generate face-to-face conversations, paying attention not only to wording but also to expressions and pauses. Tavus stated that in tests, after a one-minute video call with Griffin-Lite, 26 out of 54 participants (48%) believed the other party was a real person; its previous generation system only convinced 1 out of 41 participants.Participants were told they would have a one-minute video call with another person to discuss things they were looking forward to this year, and only at the end of the call were they asked if they suspected the other party was not a real person. Tavus noted that those who became suspicious typically noticed within 20 seconds. This result comes from Tavus's own research page, where participants were recruited through a channel they call an independent research platform. Community comments on X pointed out that these results have not been independently verified and do not meet standard protocols.In the NVIDIA VideoFDB benchmark, Tavus claimed Griffin-Lite ranked first, with a generation dimension score of 3.83, the next best system scoring 2.8, and human reference scoring 3.92; the perception dimension score was 3.73, with the strongest baseline at 3.44 and human reference at 4.2. Tavus stated that NVIDIA independently conducted this evaluation. Griffin supports full duplex, allowing it to listen, see, and speak simultaneously, with an average audio-video delay of 0.43 seconds on the NVIDIA H100 chip. Tavus completed a $40 million Series B funding round led by CRV in November 2025.

As TOKEN2049 in Singapore approaches, Gate has released a promotional video for the event. In the video, Gate's founder and CEO Dr. Han shares his thoughts on continuous innovation and change, stating that the next one to two years will mark "a new starting point," and Gate will continue to advance its products, services, and market development.It is reported that Dr. Han will attend this TOKEN2049 in Singapore, delivering a keynote speech on the main stage and engaging in a fireside chat with four-time F1 World Champion Max Verstappen. During the event, Gate will set up a brand booth and exclusive lounge at the TOKEN2049 venue, providing participants with an opportunity to gain in-depth insights into Gate's products, business layout, and global ecological development.Additionally, Gate will host the Gate Gala awards dinner at the Raffles Hotel in Singapore on October 8, expecting nearly 300 global industry founders, institutional representatives, investors, content creators, and ecological partners to attend.Currently, Gate's global user base has surpassed 61 million, with the platform supporting over 5,300 cryptocurrencies and more than 12,800 stocks and ETFs, covering multiple areas including spot trading, contracts, stocks, wealth management, payments, and AI. As the business landscape continues to expand, Gate is building a global platform that encompasses diverse assets and financial services. During this TOKEN2049 trip to Singapore, Gate will focus on presenting the latest developments and future plans of the platform, opening the next phase with a more complete business map and broader global connections.

On September 23, Google announced that any user with a Google account can generate 1080p HD videos for free in its browser-based video editor Google Vids, accessible at vids.new. This feature is based on the Gemini Omni 1.1 Flash model and can transform text prompts, photos, audio, or existing clips into new videos. Previously, such generative capabilities mostly required a paid subscription to use.Google has also added several editing controls that previously required paid tools: scene extension can lengthen clips while maintaining consistency in lighting, characters, and scenes; users can set precise clip durations to align with scripts or voiceovers; and old AI clips with lower resolutions can be upgraded to 1080p without needing to regenerate. Each video segment generated by this tool embeds Google DeepMind's SynthID invisible watermark, which is not visible to the naked eye but can be read by machines. Users can check whether a segment was generated by AI using the SynthID Detector.The free quota varies by account type: personal Google accounts receive a standard quota, paid AI packages can unlock more, and Workspace Business and Enterprise subscribers receive a larger generation pool and team management controls. Gemini Omni will be launched as a paid feature for Google AI Plus, Pro, and Ultra subscribers in May 2026, with free access previously limited to YouTube Shorts and YouTube Create.

According to Bloomberg, ByteDance is preparing to launch an artificial intelligence model for real-time spatial video generation, with applications covering fields such as robotics and autonomous systems, and will compete with companies like Meta and Alphabet.Insiders say that ByteDance founder Zhang Yiming is personally overseeing the development of the model, coordinating multiple business departments within the company, and investing in artificial intelligence resources and computing power support. The model may be released as early as next month, but the specific timing is still subject to change.

Harmony has released two proposals to comprehensively shut down the mainnet launched in 2019, migrate the native token ONE to Ethereum, and shift towards an AI video "mashup economy" business. The team stated that the threats posed by national-level attackers and AI entities are the reasons for proposing the network shutdown plan.The migration plan proposes to take a snapshot of user wallets, staking delegations, validator rewards, smart contracts, and tokens within centralized exchanges at the last block of the network, airdropping new ONE to the same wallet addresses on Ethereum, with holders not needing to actively claim; delegated stakes and unclaimed rewards will be airdropped to their respective governance vaults. The total supply of ONE and the issuance rate will remain unchanged, with newly issued tokens intended for the new business and feedback from governors being considered.Multi-signature wallets, liquidity pools, and on-chain applications cannot be migrated. The team urges users to exit all smart contracts by September 10, 2026, and plans to publicly disclose token contracts, snapshot calculations, and airdrop scripts for auditing. Validators can stop running nodes starting from September 10 at 22:00 Beijing time. The team plans to compensate for the difference in issuance rewards between node shutdown and the final block of the network, establishing a one-time compensation pool of $1.372 million, to be paid in four quarters to validators and their delegators who timely shut down, sign agreements, retain stakes, and serve as governors of the new project.The new business will open up prompts and materials for users to create secondary content, with AI entities expanding video stories, and will recruit operators responsible for video generation, distribution, and content review. Harmony plans to subsidize GPU hardware in the first year and promote demand for video generation, with operators required to stake tokens to earn rewards based on service online time. The team plans to help operators generate up to $1 million in total revenue in the first year, provided they meet staking and online rate requirements; promoters can initially earn a 30% ongoing commission from each $10 monthly subscription they recommend. Both proposals are non-binding and the plans may still be adjusted.

This set of posts seems to be the first public exposure of James's formal relationship with Polymarket. Polymarket, as well as representatives from James and the 76ers, did not immediately respond to The Block's request for comment. Previously, James's free agency decision generated approximately $273 million in related contract trading volume across major prediction platforms, according to Legal Sports Report, surpassing any single basketball game; data from Front Office Sports indicated that the related trading volume on Kalshi was about $226 million, while Polymarket's international platform exceeded $43 million.Most traders bet on the wrong outcome: before the official announcement, the odds of the 76ers winning were only about 9%, while the Heat were around 45%. James subsequently signed a two-year, $8 million contract with the 76ers, which includes a player option. Additionally, James had signed a multi-year talent ambassador agreement with DraftKings in 2024. Some fans expressed criticism of James's collaboration with prediction platforms in social media comments.

OpenAI released a video on X, previewing the upcoming release of the Astra model (the video content switches between ASTRA and A6TRA), and the model name may be GPT-6-Astra.

Google DeepMind has released the video understanding capabilities of the Gemini intelligent agent, which are now available for the Gemini 3.7 Flash, 3.6 Flash, and 3.5 Flash-Lite models. This feature enhances the efficiency and accuracy of long video analysis by dynamically searching, scanning, and inspecting video clips.Compared to traditional fixed frame rate processing methods, it can reduce token consumption by up to 88%, lower analysis costs by 66%, and improve accuracy by up to 7%. This feature is now available through the Gemini API in Google AI Studio and the Gemini Enterprise Agent Platform, and will be applied in the Gemini app and the YouTube "Ask YouTube" feature in the future.

Alibaba's video large model Wan 3 is officially launched, and Nvidia is negotiating an investment in Perplexity, which is valued at over $30 billion.Xiaomi releases the Xuan Ring O3 processor using 3 nm technology, and smart ring manufacturer Oura plans to raise up to $3 billion in an IPO.

MiniMax officially launched the AI video creation tool MiniMax Design today, a Harness product built on the native multimodal video model H3. Users only need to present their creative requirements, and the system can understand the goals, break down tasks, call relevant models and Skills, and complete the entire video production process from material processing and generation editing to final delivery, without the need to switch repeatedly between multiple professional tools.MiniMax Design features a built-in Agent 3D director's platform, where users can describe scene and shot requirements in natural language, and the system can preview composition and character relationships in 3D space. After confirmation, it hands over to H3 to generate the video, moving trial and error to the storyboard phase. This product is aimed at commercial content scenarios such as brand creative testing, knowledge video production, PV/MV, etc., and can iteratively generate multiple versions in batches. Creators can also connect their local ComfyUI workflows to MiniMax Design, with the Agent assisting in adjusting nodes and parameters.

Elon Musk stated that the image and video understanding capabilities of Grok 4.6 have undergone significant upgrades.

The Volcano Engine officially launched the Seedance 2.5 API. Compared to Seedance 2, the new version has significantly improved in areas such as instruction adherence, long narratives, character consistency, realism, lighting and shadow quality, and sound effects. It natively supports 30-second video output, a maximum of 50 multimodal material references, and is compatible with more than a dozen languages.

According to Jinshi reports, BeamR has introduced NVIDIA (NVDA.O)'s video super-resolution technology into live sports events to enhance the viewing experience for audiences.

According to Jinshi reports, on the evening of July 30, Beijing time, the Chinese lead in Sino-U.S. economic and trade relations, Vice Premier He Lifeng, held a video call with the U.S. lead, Treasury Secretary Janet Yellen, and Trade Representative Katherine Tai.The two sides discussed the implementation of the important consensus reached during the meeting between the leaders of China and the U.S. in Beijing, focusing on stabilizing economic and trade relations, pragmatic cooperation, and addressing each other's concerns. The Chinese side expressed serious concerns regarding the recent economic and trade restrictions imposed by the U.S. side.

According to official news, the Solana ecosystem 3D clay dinosaur NFT project Claynosaurz has recently launched an animated short film on the mainstream streaming platform Amazon Prime Video.Its floor price surged to 24 SOL today, currently reported at 23.25 SOL, with a 24-hour increase of 43.59%.

Japan's 105th Prime Minister Sanae Takaichi will deliver a speech via video at the "WebX2026" conference.WebX2026 will be held in Tokyo from July 13 to 14, 2026.

According to a report by The Wall Street Journal, the prediction market platform Polymarket is suspected of paying dozens of social media creators to instruct them to record fake betting and high-profit videos on a simulated website that closely resembles the official site, using this as a promotional tactic to attract new users. The report indicates that the campaign required creators to conceal the official sponsorship relationship and specifically targeted the domestic audience in the United States.The investigation shows that the related marketing videos have accumulated over 140 million views on platforms such as TikTok, YouTube, and Instagram, involving up to $1.9 million in false bets. In response, Polymarket did not directly address the specific details of the investigation but stated in a statement that, as part of promoting industry transparency, the platform is conducting a comprehensive audit of active promotional content to ensure it strictly complies with internal standards and applicable regulatory and legal disclosure requirements.

According to Bitcoin News, Greek politician and former finance minister Yanis Varoufakis admitted that he once held 2,000 bitcoins.Yanis Varoufakis stated that he later donated all of his bitcoins due to his disagreement with the ideology represented by bitcoin.

Alibaba (09988.HK) mentioned in its financial report that it is continuously expanding the boundaries of AI capabilities through deep innovation. Recently, by launching a series of brand new large language models and multimodal models, we have made significant progress in model intelligence. In March, we released Qwen3.6-Plus, achieving a remarkable overall performance improvement, particularly excelling in programming and agent programming, becoming an industry benchmark in front-end web development and complex warehouse-level tasks.Qwen3.6-Plus also possesses stronger multimodal perception and reasoning capabilities, with a native context window length supporting up to 1 million tokens, further enhancing stability and reliability. In addition to the Qianwen model family, we are also continuously enriching our specialized model layout, including the world model HappyOyster that supports real-time creation and interaction, as well as the multimodal model HappyHorse for video generation, both of which are currently undergoing gray testing.