According to CoinWorld, Cognition has released Devin Fusion, a hybrid model architecture for agent programming. This system uses a "sidekick" mechanism to allow small and large model agents to run in parallel. The large model handles planning and review, while the small model handles code exploration and testing, avoiding cache invalidation overhead. Secondly, it features dynamic routing, adjusting the model according to task evolution to achieve "zero-cost" upgrades. Tests show that in the FrontierCode benchmark, Devin Fusion maintains leading-edge model performance while reducing the development cost of GPT-5.5 and Opus 4.8 level models by an average of 35%. When used with Fable 5, the cost can be reduced by 41%. In internal development, 88% of merged PRs were driven by Fusion's automatic routing. However, in complex tasks, over-delegation leads to performance degradation, with the score dropping from 54 to 27.

According to Bijie.com, Vibe Coding platform Base44 has announced the launch of its first self-developed large-scale model, Base 1, which is already live in the production environment. Unlike platforms that only call general model interfaces, Base44 attempts to achieve full-stack vertical integration from database and backend infrastructure to the intelligent layer through its self-developed model. The core advantage of Base 1 lies in its unique data training source: tens of millions of real user development records accumulated by the platform in the past, including users' original requirements, code generated by the intelligent agent, errors during operation, user modifications, and real closed-loop signals of whether the application ultimately runs successfully. Base44 founder Maor Shlomo pointed out that general-purpose large-scale models must take into account various programming languages and a wide range of tasks, while Base 1 focuses on specific application development environments. They not only trained the model to write code, but also trained it to have "product decision-making ability," enabling it to participate in product discussions, predict users' next needs, and even avoid unreasonable development paths. Currently, Base 1 has been added to the platform's model selector, alongside other mainstream models, and the system will automatically schedule it according to the task type.

According to Bijie.com, Kimi's valuation has risen to $31.5 billion, with annual recurring revenue (ARR) exceeding $300 million. Kimi recently completed its previous funding round with a valuation of $20 billion, and a new round of financing has begun, with a pre-money valuation of $31.5 billion. Sources close to Kimi revealed that the company disclosed its latest revenue data during this round of financing discussions: as of mid-June, ARR had exceeded $300 million. This revenue growth is mainly driven by increased developer usage and API revenue resulting from model iterations. Currently, API revenue accounts for over 70% of Kimi's total revenue and continues to rise. Kimi's revenue curve is beginning to show characteristics of Anthropic's early commercialization stage, including increased developer usage, a higher API share, and growth in overseas paying users, with the pricing structure shifting upwards due to model capability iterations.

According to Lookonchain monitoring, two whales are shorting BTC, with a total position exceeding $100 million. One whale is shorting 900 BTC with 40x leverage, amounting to $53.69 million; the other is shorting 800.75 BTC with 20x leverage, amounting to $47.76 million.