AI Industry Development & Policy Dynamics

Brockman cites GPT-5.4 Pro solving a new Erdős problem as proof of sudden model leaps, and OpenAI hints at personalized writing advances while noting existing gaps in 'soul' and a forthcoming model. Abstract: The piece reports two OpenAI disclosures from the Core Memory podcast: a GPT-5.4 Pro milestone solving an Erdős problem, signaling rapid capability gains with broad implications; and OpenAI's plan for a new model to deliver more personalized, soulful writing, addressing critiques of LLM subjectivity.

Gate News message, April 22 — Sam Altman, CEO of OpenAI, has publicly responded to two attacks on his residence for the first time during a Core Memory podcast appearance. Altman stated that similar incidents are likely to occur in the future, though he did not elaborate on the details of the attack

Shanghai reports one new generative AI service registered by April 21, 2026, bringing Shanghai's total to 158 under interim measures requiring regulatory registration. Abstract: This brief notes that Shanghai's cyberspace administration announced that as of April 21, 2026, one new generative AI service completed registration, bringing the total to 158. The move follows the Generative Artificial Intelligence Service Management Interim Measures, which require providers to register with regulatory authorities.

PicWe announces public beta of PicWe Wallet, an AI-agent-enabled, on-device key wallet with no recovery phrases. It supports multi-chain assets, swaps, AI-accessible automation, and aims to unify RWA infrastructure. PicWe has launched the public beta of PicWe Wallet, an AI Agent-enabled wallet that stores keys on-device, eliminates recovery phrases, and keeps critical operations local. The beta supports multi-chain asset management, swaps, and stablecoin-based fees while enabling programmable AI interactions. Broader PicWe initiatives position the platform as unified infrastructure for real-world assets, enabling issuance, circulation, settlement, cross-border payments, tokenization, and supply-chain coordination for enterprise use cases.

Open-sourced ml-intern, Hugging Face's autonomous ML research agent that reads papers, curates data, trains, evaluates, and iterates across science, medicine, and math. Abstract: Hugging Face's ml-intern is an open-source autonomous ML research agent that reads papers, curates datasets, trains on local or cloud GPUs, evaluates results, and iterates improvements. Built on smolagents with CLI and web interfaces, it navigates arXiv/HF Papers, HF Hub, and HF Jobs. Demonstrations span science, medicine, and mathematics, showing end-to-end automation and performance gains.

Summary: JPMorgan lifts 2026 S&P 500 to 7,600, 2027 to 385 EPS, driven by AI enthusiasm and easing Middle East tensions; forward P/E stays 22x, potential to 23x and ~8,000 if tensions fall; warns of short-term consolidation amid oil and geopolitical risks. Abstract: JPMorgan’s strategy team, led by Dubravko Lakos-Bujas, raised the year-end S&P 500 forecast to 7,600 from 7,200, citing renewed AI enthusiasm and easing Middle East tensions. 2026 EPS is boosted to $330 and 2027 to $385, with the forward multiple held at 22x; a quicker geopolitical resolution could lift the multiple to 23x and push 2026 toward 8,000. The AI theme gained momentum after Anthropic unveiled Mythos, with about two-thirds of AI-related S&P 500 stocks outperforming. Risks include higher oil prices and lingering geopolitical tensions, suggesting a potential short-term consolidation before further gains.

Gate News message, April 22 — OpenAI has briefed the U.S. federal government and the Five Eyes alliance on a new cybersecurity product, according to Axios.

Gate News message, April 22 — OpenAI is forming a joint venture called DeployCo with multiple private equity firms including TPG, Bain Capital, Advent International, Brookfield, and Goanna Capital. OpenAI will initially invest $500 million in equity with majority voting rights and the option to

Gate News message, April 22 — Google Research released ReasoningBank, an agent memory framework that enables large language model-driven agents to continuously learn after deployment. The framework extracts universal reasoning strategies from both successful and failed task experiences, storing

SK Telecom and Nvidia push A.X K2 under Korea’s Proprietary AI Foundation Model initiative, expanding from A.X K1 to a full-stack, open-source AI platform via the Krafton-Rebellions consortium for academic and commercial use. Abstract: The article reports SK Telecom’s partnership with Nvidia to develop A.X K2 as a successor to A.X K1 under Korea’s government-backed program. The effort aims to create a full-stack, open-source AI platform through a Krafton-Rebellions-led consortium, with research focused on multimodal and vision-language models and open access to A.X models for academia and industry.

PwC Singapore to invest S$4m over 3 years to create a Trade Advisory Hub with EDB support, offering AI-driven trade, supply chain, and tax advisory; expands APAC team of 100+ specialists amid rising global trade complexity. Abstract: PwC Singapore will establish a Trade Advisory Hub backed by S$4 million over three years with Singapore EDB support to guide local and multinational firms on evolving trade rules, supply chain management, and international expansion. The initiative includes hiring specialists and developing AI-powered tools for trade, supply chain, and tax services, leveraging the firm’s existing APAC customs and trade advisory practice of over 100 consultants. The development responds to growing demand for advanced trade-planning solutions amid changing regulations, e-commerce growth, and increasingly complex global supply chains.

Gate News message, April 22 — Dai Jifeng, an associate professor at Tsinghua University's Department of Electronic Engineering, has founded Naive.ai, a company focused on open-source model post-training and AI agents. The startup has raised approximately $300 million at an estimated valuation of $80

Gate News message, April 22 — Amazon Web Services announced expansion of its agentic AI initiatives through multi-agent workflows, supporting Anthropic's Claude Opus 4.7 on Amazon Bedrock to help customers move beyond generative AI pilots. The company is expanding partner relationships as customers

Gate News message, April 22 — Zhipu AI announced that it will discontinue automatic renewal of the GLM Coding Plan unlimited weekly quota subscription starting at 10:00 AM Beijing time on April 30, 2026. The discontinuation affects users currently subscribed to the legacy plan with auto-renewal

Anthropic drops Claude Code from Pro plan, drawing criticism as developers migrate to OpenAI; Codex remains free/basic, GPT-5.4 and Image 2.0 boost performance, driving large user migration. Abstract: The article examines Anthropic's removal of Claude Code from the $20 Pro plan, which triggers backlash from developers who call it a hidden price increase and a reliability risk. It contrasts this move with OpenAI's policy of keeping Codex in free and basic tiers, while highlighting strong model performance from GPT-5.4 and ChatGPT Images 2.0, and notes a rapid migration of users to OpenAI, with Codex reportedly exceeding 4 million weekly active users.

Meta plans to deploy employee monitoring software that captures mouse movements, clicks, keystrokes, and periodic screenshots to train AI models; data not used for performance reviews, with safeguards. Abstract: Meta intends to install monitoring software on U.S. employees’ computers to capture mouse movements, clicks, keystrokes, and occasional screenshots to train its AI models. The company says this data won’t be used for performance evaluations and that security measures protect sensitive content, aiming to improve models for tasks like dropdown menus and keyboard shortcuts.

Gate News message, April 22 — Tesla has filed its generative AI-powered voice assistant with China's Cyberspace Administration, joining 158 AI tools that have completed the country's official registration process. The move represents a routine compliance step in China's strict regulatory environment

Gate News message, April 22 — Anthropic has not relaxed its policy on Claude CLI usage for third-party tools like OpenClaw, according to a clarification by the platform. The apparent policy shift was actually a misinterpretation by OpenClaw founder Peter Steinberger of a statement made by Claude Cod

Gate News message, April 22 — 0G Labs has announced a partnership with Alibaba Cloud to enable AI agents to directly invoke Alibaba's Qwen large language model on-chain through 0G's blockchain infrastructure. The collaboration marks a shift in AI access paradigms from API-based approaches to

Sandoll unveils Sandoll Square, a new AI and Web3 division to build a decentralized platform ecosystem, merging AI-driven content processing with blockchain for seamless data management and interoperability; CEO and CTO roles are being recruited.