Author: RadarAI Editorial
Editor: RadarAI Editorial
Last updated: 2026-05-25
Review status: Editorial review pending
Brief
速报
官方
AI动态
开源
Agent tech is maturing rapidly—Codex and similar tools are enhancing core workflow capabilities. Meanwhile, Google's CEO acknowledged Gemini's gaps in coding agents and long-horizon tasks, signaling a shift from model benchmarks to real-world task completion. Anthropic's 'should do' > 'can do' framework highlights the growing scarcity of AI judgment.
Editorial standards and source policy: Editorial standards, Team. Content links to primary sources; see Methodology.
## 🔍 Key Insights
**Agent technology is maturing at an accelerating pace**: Toolchains like Codex are rapidly iterating on critical workflow capabilities. At the same time, **Google CEO Sundar Pichai publicly acknowledged Gemini’s clear shortcomings in Coding Agents and long-horizon tasks** [3], confirming a broader industry shift—from competing on raw model performance to competing on *real-world task completion* and *end-to-end reliability*. **Anthropic’s “should do” > “can do” framework** [7] has emerged as a defining lens for the growing scarcity of human judgment in the AI era.
## 🚀 Key Updates
- **Codex adds Queue, Steer, and Info Panel—though Queue has a known bug** [0]: Enables task routing and contextual steering, boosting multi-threaded development efficiency
- **`/side` command enables sidebar conversations**, letting users check `/goal` long-task progress in real time *without interrupting* the main session [1]: Greatly improves observability and control for complex coding workflows
- **`/goal` task panel now supports delete, pause/resume, and edit—all changes deferred until the current turn completes** [2]: Gives users stronger agency over AI Agent execution timing
- **Google CEO Sundar Pichai admits Gemini lags behind in Coding Agents and long-horizon reasoning** [3]: Highlights remaining gaps in deep reasoning and stable tool calling on the path to AGI
- **Anthropic releases *The Founder’s Action Manual*, introducing a “Four-Stage Framework” for AI startups—and declares *judgment*, not technical skill, the scarcest resource** [7]: As technical barriers fall, strategic discernment (“what *should* we build?”) outweighs execution capability (“what *can* we build?”)
- **DeepSeek secures ¥70B in funding, pivoting strategically toward AI infrastructure platforms** [9]: A direct response to the explosive token consumption and inference cost pressures of the Agent era
- **“Uncle Bob” Martin predicts developers will shift to declarative specification languages (e.g., Gherkin), while AI handles all procedural coding** [14]: A historic paradigm shift in software engineering is underway
- **Liquid DOM goes open source**: Brings Apple’s Liquid Glass visual effects to the web via WebGPU [13]: Web frontend rendering capabilities are being redefined—again—by AI-era graphics tech
## 🔗 Sources
[0] Codex Pro Tips: Info Panel, Steer, and Queue Explained — https://www.bestblogs.dev/status/2058618849172365623?utm_source=rss&utm_medium=feed&utm_campaign=resources&entry=rss_article_item
[1] Codex `/goal` Progress Tracking & `/side` Command Tips — https://www.bestblogs.dev/status/2058612576775229669?utm_source=rss&utm_medium=feed&utm_campaign=resources&entry=rss_article_item
[2] Codex `/goal` Task Panel Guide: Delete, Pause, Edit — https://www.bestblogs.dev/status/2058612580315177199?utm_source=rss&utm_medium=feed&utm_campaign=resources&entry=rss_article_item
[3] Google CEO Admits Coding Is Falling Behind — https://www.bestblogs.dev/article/a07b572a?utm_source=rss&utm_medium=feed&utm_campaign=resources&entry=rss_article_item
[7] 30x Growth in 15 Months: Anthropic Shares Its Methodology — https://www.bestblogs.dev/article/3e09350b?utm_source=rss&utm_medium=feed&utm_campaign=resources&entry
Agent technology is maturing at an accelerating pace: Toolchains like Codex are rapidly iterating on critical workflow capabilities. At the same time, Google CEO Sundar Pichai publicly acknowledged Gemini’s clear shortcomings in Coding Agents and long-horizon tasks [3], confirming a broader industry shift—from competing on raw model performance to competing on real-world task completion and end-to-end reliability. Anthropic’s “should do” > “can do” framework [7] has emerged as a defining lens for the growing scarcity of human judgment in the AI era.
🚀 Key Updates
- Codex adds Queue, Steer, and Info Panel—though Queue has a known bug [0]: Enables task routing and contextual steering, boosting multi-threaded development efficiency
/side command enables sidebar conversations, letting users check /goal long-task progress in real time without interrupting the main session [1]: Greatly improves observability and control for complex coding workflows /goal task panel now supports delete, pause/resume, and edit—all changes deferred until the current turn completes [2]: Gives users stronger agency over AI Agent execution timing - Google CEO Sundar Pichai admits Gemini lags behind in Coding Agents and long-horizon reasoning [3]: Highlights remaining gaps in deep reasoning and stable tool calling on the path to AGI
- Anthropic releases The Founder’s Action Manual, introducing a “Four-Stage Framework” for AI startups—and declares judgment, not technical skill, the scarcest resource [7]: As technical barriers fall, strategic discernment (“what should we build?”) outweighs execution capability (“what can we build?”)
- DeepSeek secures ¥70B in funding, pivoting strategically toward AI infrastructure platforms [9]: A direct response to the explosive token consumption and inference cost pressures of the Agent era
- “Uncle Bob” Martin predicts developers will shift to declarative specification languages (e.g., Gherkin), while AI handles all procedural coding [14]: A historic paradigm shift in software engineering is underway
- Liquid DOM goes open source: Brings Apple’s Liquid Glass visual effects to the web via WebGPU [13]: Web frontend rendering capabilities are being redefined—again—by AI-era graphics tech
🔗 Sources
[0] Codex Pro Tips: Info Panel, Steer, and Queue Explained — https://www.bestblogs.dev/status/2058618849172365623?utm_source=rss&utm_medium=feed&utm_campaign=resources&entry=rss_article_item
[1] Codex /goal Progress Tracking & /side Command Tips — https://www.bestblogs.dev/status/2058612576775229669?utm_source=rss&utm_medium=feed&utm_campaign=resources&entry=rss_article_item
[2] Codex /goal Task Panel Guide: Delete, Pause, Edit — https://www.bestblogs.dev/status/2058612580315177199?utm_source=rss&utm_medium=feed&utm_campaign=resources&entry=rss_article_item
[3] Google CEO Admits Coding Is Falling Behind — https://www.bestblogs.dev/article/a07b572a?utm_source=rss&utm_medium=feed&utm_campaign=resources&entry=rss_article_item
[7] 30x Growth in 15 Months: Anthropic Shares Its Methodology — https://www.bestblogs.dev/article/3e09350b?utm_source=rss&utm_medium=feed&utm_campaign=resources&entry
← Back to Updates