This AI Search Engine Drew 12,000 Paying Users and Raised $11 Million
Users and investors are both buying in.
🎨 Google Opens Access to Imagen 3 AI Image Generation Model
🌏 Ola Plans to Launch India's First AI Chip
🧬 USC Team Develops DeepPBS
🌐 Linux Open-Source OS — deepin V23 Officially Released
🤖 Xinghuo Ultra-Fast Hyper-Humanoid Interaction
📚 Baidu Wenku Launches New AI App "Chengpian"
💰 Consensus Closes $11 Million Series A
💹 CodeRabbit Closes $16 Million Series A
🖼️ Half_illustration: Turn Photos Into Artistic Masterpieces
Google Opens Access to Imagen 3 AI Image Generation Model
Google's latest text-to-image AI model — Imagen 3 — is now open for users in the United States to try out. The feature is accessible through AI Test Kitchen, where users can experience its advanced image generation capabilities.
Imagen 3 was announced at the I/O developer conference in May 2024, with select Vertex AI users invited to test it in June. It is now officially available to U.S. users. DeepMind CEO Demis Hassabis stated that Imagen 3 is more accurate at understanding text prompts and converting them into images, with greater creativity and richer detail. In Google's evaluations, Imagen 3 outperformed top models including Imagen 2, DALL-E 3, Midjourney v6, Stable Diffusion 3, and Stable Diffusion XL 1.0.
Ola Plans to Launch India's First AI Chip
Indian EV company Ola has announced it will launch India's first self-developed AI chip in 2026, built on the ARM architecture. The company unveiled products including the Bodhi series of AI chips, the Sarv-1 cloud-native CPU, and the Ojas edge AI chip.
Ola CEO Bhavish Aggarwal emphasized that India should develop its own AI capabilities independently and reduce reliance on third-party alternatives. Bodhi-1, the first chip, is expected to hit the market in 2026. Bodhi-2, the next-generation product offering higher-end AI workload performance, is slated for 2028. Sarv 1 is a cloud-native CPU based on the ARM instruction set, designed for AI computing demands in the data center industry. Ojas is an edge AI chip that may be integrated into Ola's next-generation electric vehicles, though specific details have not been disclosed.
USC Team Develops DeepPBS
Predicting the binding specificity of proteins with DNA is critical to understanding gene regulation. Researchers at the University of Southern California and the University of Washington have proposed a geometric deep learning model called DeepPBS [1], which can predict binding specificity from protein-DNA structures.
The current version of DeepPBS is tailored for double-stranded DNA and does not apply to single-stranded DNA, RNA, or chemically modified bases. The model has potential to be extended to different scenarios and other polymer-polymer interactions. DeepPBS could accelerate the design of new drugs and treatments targeting specific mutations in cancer cells, and may lead to new discoveries in synthetic biology and applications in RNA research.
Linux Open-Source OS — deepin V23 Officially Released
deepin V23 [2], China's first Linux open-source operating system with deeply integrated AI capabilities, runs on the Linux 6.6 LTS kernel with an all-new DDE (Deepin Desktop Environment). It features deep AI integration and supports mainstream CPU architectures including X86, ARM, LoongArch64, and RISC-V, providing extensive hardware support and an optimized OS experience.
deepin community founder Wenhuan Liu emphasized that the release of deepin V23 is not simply a version update, but rather an effort to push the boundaries of what Linux desktop distributions can do through substantial, genuinely user-needed, and innovative content — making the Linux desktop as powerful as commercial operating systems like Windows and macOS.
Xinghuo Ultra-Fast Hyper-Humanoid Interaction
iFlytek's Xinghuo voice large model has received an update, introducing the "Xinghuo Ultra-Fast Hyper-Humanoid Interaction" feature, delivering a more natural and fluid voice interaction experience for users.
The "Xinghuo Ultra-Fast Hyper-Humanoid Interaction" feature employs end-to-end modeling technology to achieve ultra-fast response and free interruption, making conversations more fluid and natural. Based on user commands, it flexibly controls emotion, style, dialect, and even speaking speed to meet personalized expression needs. It keenly perceives user emotions, resonates emotionally, recognizes moods, and responds with appropriate tone to enhance the warmth of interaction. It can imitate various character voices and tones, such as "Sun Wukong" and "Crayon Shin-chan," enabling arbitrary persona switching to add fun to interactions.
Baidu Wenku Launches New AI App "Chengpian"
Baidu Wenku's AI-native application "Chengpian" App has launched on major Android platforms, offering rich AI-assisted writing and editing features.
"Chengpian" supports intelligent web-wide search, AI daily hot topics, and other news-app-like functions. It also supports multi-image one-click video generation, ultra-long article writing, document comprehension and summarization, and other generative large model capabilities. It is the first in the industry to offer 100,000-word long-form generation and multimodal editing capabilities, becoming the industry's first "search-read-create-edit" one-stop AI free creation platform.
Consensus Closes $11 Million Series A
AI academic search engine Consensus [3] recently announced it has closed an $11 million Series A round, led by Union Square Ventures, with other investors including Nat Friedman, Daniel Gross, and Draper Associates.
Co-founder & CEO Eric Olson and Northwestern classmate Christian Salem founded Consensus together in 2021. Consensus currently has 400,000 monthly active users, including students, doctors, and health-conscious consumers. It has 12,000 paying users, primarily students and professionals, who use Consensus for research to reduce their dependence on Google. Consensus expects to reach $1.5 million in annual recurring revenue (ARR) this year.
CodeRabbit Closes $16 Million Series A
AI code review platform CodeRabbit [4] recently announced it has closed a $16 million Series A round, with investors including Engineering Capital, Flex Capital, and CRV (First Round Capital).
CodeRabbit is dedicated to automating the code review process, using AI technology to provide high-quality feedback that helps developers save time and improve code quality. With the Series A funding, CodeRabbit expects to expand its R&D team, further optimize its AI code review technology, and extend its market reach.
Half_illustration: Turn Photos Into Artistic Masterpieces
Half_illustration is an innovative AI image generation tool based on a Flux.1 LoRA model that transforms ordinary photos into illustration-style works full of artistic sensibility.
Users simply describe their desired scene in words, and Half_illustration generates corresponding images that seamlessly blend photography and illustration elements, creating visual effects that are both realistic and dreamlike. Users can unleash their imagination, adding elements like smoke, flowers, and flames to craft dramatic scenes. Including "in the style of TOK" in the description helps the model better capture the desired style.
The most glorious thing about being human is that one forever dwells within certain limitations, yet still assumes a posture of striving forward.
— Eros and Compassion
Editorial Team
Editor: Yuki
Design: Ivan
For business cooperation, add WeChat: Rwkfbcianvd
References
[1] DeepPBS: https://www.nature.com/articles/s41592-024-02372-w
[2] deepin V23: https://www.deepin.org/zh/download/
[3] Consensus: https://consensus.app/
[4] CodeRabbit: https://coderabbit.ai/