Xiaoliu BOT
中文EN

X Platform June 2 AI Brief | AI Companies Rush to IPO, WeChat Agent Launch Set for Q3, Doubao Pricing Sparks Debate

Anthropic Files S-1 for IPO; MiniMax and Zhipu AI Apply for A-Share Listings

Anthropic has confidentially submitted a draft S-1 registration statement to the U.S. SEC, formally initiating its IPO process. During the same timeframe, MiniMax and Zhipu AI have also submitted applications for listings on China’s A-share and STAR markets, respectively, and have begun listing guidance. As AI companies flock to capital markets, Anthropic’s Claude series models continue to drive B2B revenue, while MiniMax and Zhipu AI are seeking funding channels within the domestic capital market.

Source:

WeChat Plans Q3 Launch for Built-in Agent, Deeply Integrating Hunyuan Model

According to the Financial Times, WeChat plans to launch an in-platform Agent function in the third quarter of this year, which will enable AI to call upon millions of mini-programs within the WeChat ecosystem. This will represent the deepest integration yet between the Hunyuan large model and WeChat, with a scenario density difficult for any competitor to replicate. @MaxForAI claims to have confirmed the news from multiple independent sources, with an official announcement expected soon.

Source:

ByteDance Seed Core Researcher Quanquan Gu Departs; Previously Rumored to be Pre-training Lead

ByteDance Seed core member Quanquan Gu has confirmed his departure from ByteDance Seed. Over the past three years, he simultaneously worked on two core tracks: AI for Science and frontier LLMs, leading teams to develop biomolecular structure prediction models like SeedFold and SeedProteo. In 2025, he joined LLM pre-training work and participated in the training of Seed2.0. However, @MaxForAI later updated that ByteDance insiders stated Quanquan Gu was not the pre-training lead. Gu himself confirmed the departure on X, stating he will explore new directions.

Source:

Former Huawei Pangu Model Lead Wang Yunhe Starts AI Agent Company Valued at $100M

Former Director of Huawei’s Noah’s Ark Lab and Pangu large model lead, Wang Yunhe, has founded Shanghai Jiyuan Rhythm Technology Co., Ltd., venturing into the AI Agent field. The company has completed a funding round valuing it at $100 million, with investors including top-tier VCs and leading internet companies. The CTO is Han Kai, former Chief Researcher at Huawei’s Noah’s Ark Lab. The company is currently recruiting for AI Agent-related positions with annual salaries ranging from 600,000 to 1 million RMB.

Source:

Doubao Expected to Officially Charge in Late June; Three-Tier Pricing Sparks AI Commercialization Debate

ByteDance’s Doubao is expected to officially launch a paid subscription model in late June, with three pricing tiers: 68 RMB, 200 RMB, and 500 RMB per month. According to Tencent Research Institute data, the payment rate of frequent AI users is 13 times that of infrequent users, and 56.1% of professionals are willing to pay for AI. However, Doubao’s average daily user session length is only 10 minutes, and the free version suffices for most people’s daily needs. @derek_wall90176 believes the core risk lies not in the price, but in defining the boundary between the free and paid versions; the precedent of ERNIE Bot being impacted by free competitors after introducing fees remains a cautionary tale.

Source:

Codex and Claude Code Quota Issues Erupt; Claude Fully Resets Limits

Multiple bloggers report that Codex’s 5-hour limit is even stricter than Claude Code’s, with fixing a single error triggering the limit. Concurrently, Claude Code version 4.8 frequently experiences bugs with tool call parsing failures. Anthropic subsequently announced a full reset of the 5-hour and weekly limits for all Pro and Max users, citing that some Claude Code sessions generated excessive parallel sub-agents, leading to abnormal usage consumption. @dotey compared the quota reset mechanisms of Codex and Claude, noting that Claude’s strategy is more reasonable—remaining time is directly replenished rather than extended by 7 days.

Source:

Cursor Increases Teams User Quotas and Launches Premium Seats

Cursor announced increased usage quotas for all Teams users and launched Premium team seats, offering 5x the usage for 3x the price. @dotey praised the Cursor Agent’s effectiveness, noting its multitask mode supports parallel background tasks, offers flexible model selection (composer 2.5 is adequate in capability and speed for general tasks), and the Plan mode works stably with multitask. Shortcomings include the lack of support for /goal and a mobile version, as well as the absence of debugging features similar to Codex’s Chrome use + Computer use.

Source:

Baoyu Releases baoyu-image-gen Skill Supporting Codex CLI Image Generation, Launches New Book “Illustrated Skill”

@dotey released an update for the baoyu-image-gen Skill, supporting Codex-cli as a Provider to call Codex for image generation within Agents like Claude Code and hermes agent. That same evening, he conducted a public lecture on Agent Skills on the Turing Community live stream, sharing practical experience and his new book “Illustrated Skill: A Practical Guide to AI Efficiency.” Supporting resources for the book (including sample chapters) are now available. Bloggers like @op7418 and @xiaohu have recommended the book.

Source:

GPT-Image-2 Practical Tip: Use Subtraction to Eliminate the “AI Look”

@derek_wall90176 shared a counterintuitive tip for GPT-Image-2 image generation: The “AI look” in AI images often stems not from a lack of refinement, but from being *too* refined—skin like sandpaper, dirty textures in the background, overly sharp edges. The solution is not to add more detail, but to control the output with prompts: clean illustration, smooth shading, soft lighting, controlled details, minimal texture, etc., combined with negative prompts like –no noise, grain, artifacts. This technique is especially suitable for portrait illustrations, children’s picture books, and brand visuals.

Source:

Dreame Technology Initiates Pre-IPO Financing with 70B RMB Valuation

Dreame Technology has initiated a new round of Pre-IPO financing with a pre-money valuation of 70 billion RMB. The minimum investment threshold is 350 million RMB per ticket, releasing 5% to 10% equity, corresponding to a fundraising scale of 3.5 to 7 billion RMB. Roadshows will be held from June 3 to 8, with the deal closing on June 30. The financing schedule is tight, with only one month from BP update to closing.

Source:

OpenAI Models and Codex Now Available on AWS Amazon Bedrock

OpenAI’s frontier models and Codex are now officially available on AWS Amazon Bedrock. Enterprises can use OpenAI models to build AI applications and software engineering workflows through existing AWS security, compliance, and governance workflows. This also marks the beginning of a broader expansion of OpenAI capabilities on AWS, with future inclusions expected to include cybersecurity capabilities like Daybreak.

Source:

X Launches React with Video Feature, Rivaling Short-Form Video Interaction

X Product Lead @nikitabier announced the launch of the React with Video feature, allowing users to post video replies to posts in green screen, split-screen, or picture-in-picture formats. It is currently only available on iOS. Some bloggers have commented that this is a significant step for X towards TikTok-style short-form video interaction.

Source:

Stats: Timeline Scans=360 Bloggers Matched=39 Total Tweets Matched=234 Weighted Tweet Score=187.8 Original Tweets=113 RT Tweets=44 Crawl Attempts=2 Boundary Coverage Status=tail_confidently_crossed_target_boundary