fix: correct /goals to /goal (singular) per official Claude Code docs

Author: yanglbme

scripts/deploy_ai_pulse.py

@@ -0,0 +1,40 @@
+#!/usr/bin/env python3
+"""deploy_ai_pulse.py - Stage, commit, and push content files to doocs/awesome-ai."""
+
+import subprocess
+import sys
+import os
+
+
+def run(cmd, cwd="/tmp/awesome-ai"):
+    result = subprocess.run(cmd, shell=True, cwd=cwd, capture_output=True, text=True)
+    if result.returncode != 0:
+        print(f"ERROR: {cmd}")
+        print(result.stderr)
+        sys.exit(1)
+    return result.stdout
+
+
+def main():
+    if len(sys.argv) < 3:
+        print("Usage: deploy_ai_pulse.py <file_path> <commit_message>")
+        sys.exit(1)
+
+    file_path = sys.argv[1]
+    commit_msg = sys.argv[2]
+    repo_dir = "/tmp/awesome-ai"
+
+    # Stage the file
+    run(f"git add {file_path}")
+
+    # Commit
+    run(f'git commit -m "{commit_msg}"')
+
+    # Push
+    result = run("git push origin main")
+    print(result)
+    print(f"Successfully pushed: {file_path}")
+
+
+if __name__ == "__main__":
+    main()

src/content/en/topics/cerebras-ipo-anthropic-surpasses-openai.md

@@ -3,7 +3,7 @@ title: "Topic | AI Industry at a Historic Inflection Point: Cerebras' $100B IPO
 date: "2026-05-15"
 type: "topic"
 tags: ["Cerebras", "Anthropic", "OpenAI", "AI Chips", "Enterprise Adoption", "Industry Landscape", "Claude Code"]
-summary: "Mid-May 2026 sees three simultaneous industry-shaping events: Cerebras goes public with a $100B market cap, Anthropic surpasses OpenAI in US enterprise adoption for the first time, and Claude Code's /goals command redefines the reliability standard for coding agents"
+summary: "Mid-May 2026 sees three simultaneous industry-shaping events: Cerebras goes public with a $100B market cap, Anthropic surpasses OpenAI in US enterprise adoption for the first time, and Claude Code's /goal command redefines the reliability standard for coding agents"
 ---
 
 > The second week of May 2026 delivered three events that, taken together, signal a fundamental shift in the AI industry's competitive landscape. They are not isolated incidents — they are three converging signals of an industry accelerating toward maturity.
@@ -44,19 +44,19 @@ Meanwhile, a separate Gallup survey found that for the first time, 50% of employ
 
 ---
 
-## Claude Code's /goals Command: Coding Agents Enter the "Verifiable" Era
+## Claude Code's /goal Command: Coding Agents Enter the "Verifiable" Era
 
-On May 14, Anthropic launched `/goals` for Claude Code — a seemingly small feature with significant architectural implications.
+On May 14, Anthropic launched `/goal` for Claude Code — a seemingly small feature with significant architectural implications.
 
 **The core idea: separate the model that does the work from the model that decides when the work is done.**
 
 In traditional AI coding agent workflows, the same model both executes tasks (reading files, modifying code, running tests) and judges whether the task is complete. This is like asking a student to grade their own homework — the model may prematurely declare "done" due to context window limits or reasoning drift.
 
-`/goals` introduces a decoupled architecture: after a user defines a goal, Claude executes turn by turn, but an independent evaluator model (Haiku by default) checks whether the goal conditions are met each time the agent attempts to stop. If unmet, the agent keeps running. If met, the result is logged and the goal is cleared.
+`/goal` introduces a decoupled architecture: after a user defines a goal, Claude executes turn by turn, but an independent evaluator model (Haiku by default) checks whether the goal conditions are met each time the agent attempts to stop. If unmet, the agent keeps running. If met, the result is logged and the goal is cleared.
 
 Competitors are working on similar solutions. OpenAI lets users attach custom evaluators but leaves the termination decision to the model itself. Google's Agent Development Kit and LangGraph support independent evaluation, but developers must architect the critic node and termination logic themselves. Anthropic's approach makes the independent evaluator the default.
 
-> **Awesome AI View:** The significance of `/goals` isn't that it solves a technical problem — it exposes an industry-wide cognitive shift. The reliability of AI agents no longer depends on how smart the model is, but on how the system is architected. "You can't trust a model to judge its own homework" — this quote from an enterprise user captures the essence. As AI agents take on increasingly critical tasks (code migrations, data pipelines, security audits), the gap between "it thinks it's done" and "it's actually done" can have serious consequences. Separating the executor from the evaluator embeds verifiability into the agent architecture itself. This may be the key step for AI agents moving from "usable" to "trustworthy."
+> **Awesome AI View:** The significance of `/goal` isn't that it solves a technical problem — it exposes an industry-wide cognitive shift. The reliability of AI agents no longer depends on how smart the model is, but on how the system is architected. "You can't trust a model to judge its own homework" — this quote from an enterprise user captures the essence. As AI agents take on increasingly critical tasks (code migrations, data pipelines, security audits), the gap between "it thinks it's done" and "it's actually done" can have serious consequences. Separating the executor from the evaluator embeds verifiability into the agent architecture itself. This may be the key step for AI agents moving from "usable" to "trustworthy."
 
 ---
 
@@ -68,6 +68,6 @@ Reading these three stories together reveals a clear narrative:
 
 **At the application level**, Anthropic surpassing OpenAI marks the enterprise AI market's transition from "consumer brand-driven" to "engineering capability-driven." Enterprise customers are no longer choosing OpenAI because "ChatGPT is famous" — they're choosing Claude for its reliability, developer tool integration, and security profile.
 
-**At the tool level**, Claude Code's `/goals` command and Anthropic's Agent SDK credit system (reinstating third-party agents like OpenClaw but with dedicated API-rate credits) both point to a trend: **AI agents are maturing from experimental tools into enterprise-grade products** — meaning verifiability, observability, and controlled billing models matter more than raw "power."
+**At the tool level**, Claude Code's `/goal` command and Anthropic's Agent SDK credit system (reinstating third-party agents like OpenClaw but with dedicated API-rate credits) both point to a trend: **AI agents are maturing from experimental tools into enterprise-grade products** — meaning verifiability, observability, and controlled billing models matter more than raw "power."
 
 > **Awesome AI View:** These three stories collectively answer one question: What's the AI industry's next chapter? The answer may not be "stronger models" but "more reliable systems." When a hundred-billion-dollar chip company bets on inference speed, enterprise customers shift from consumer brands to engineering capabilities, and coding agents build in verification mechanisms, the industry's focus is shifting from "what AI can do" to "how reliably AI can do it." That shift may be more consequential than any single technical breakthrough — because it determines whether AI truly moves from lab to production, from geek toy to enterprise infrastructure.

src/content/topics/cerebras-ipo-anthropic-surpasses-openai.md

@@ -3,7 +3,7 @@ title: "专题 | AI 行业迎来历史性拐点：Cerebras 百亿市值 IPO 与
 date: "2026-05-15"
 type: "topic"
 tags: ["Cerebras", "Anthropic", "OpenAI", "AI芯片", "企业采用", "行业格局", "Claude Code"]
-summary: "2026 年 5 月中旬，AI 行业同时迎来三件大事：Cerebras 以千亿美元市值登陆纳斯达克、Anthropic 在企业市场首次超越 OpenAI、Claude Code 推出 /goals 命令重新定义编程代理的可靠性标准。这三条新闻合在一起，勾勒出 AI 行业的竞争格局正在发生根本性转变"
+summary: "2026 年 5 月中旬，AI 行业同时迎来三件大事：Cerebras 以千亿美元市值登陆纳斯达克、Anthropic 在企业市场首次超越 OpenAI、Claude Code 推出 /goal 命令重新定义编程代理的可靠性标准。这三条新闻合在一起，勾勒出 AI 行业的竞争格局正在发生根本性转变"
 ---
 
 > 2026 年 5 月的第二周，AI 行业同时发生了三件足以改写行业格局的大事。它们不是孤立的事件，而是一个行业正在加速走向成熟的三重信号。
@@ -44,19 +44,19 @@ Cerebras 的核心技术是 **Wafer-Scale Engine（WSE）**——一块占用整
 
 ---
 
-## Claude Code 的 /goals 命令：编程代理进入"可验证"时代
+## Claude Code 的 /goal 命令：编程代理进入"可验证"时代
 
-5 月 14 日，Anthropic 为 Claude Code 推出了 `/goals` 命令，这是一个看似微小但意义重大的功能更新。
+5 月 14 日，Anthropic 为 Claude Code 推出了 `/goal` 命令，这是一个看似微小但意义重大的功能更新。
 
 **它的核心思路是：把"干活的模型"和"判断是否干完的模型"分开。**
 
 在传统的 AI 编程代理工作流中，同一个模型既要执行任务（读取文件、修改代码、运行测试），又要判断任务是否完成。这就好比让学生自己给自己批改作业——模型可能因为上下文窗口限制或推理偏差，在它认为"完成"的时候，实际上还有工作没做完。
 
-`/goals` 引入了解耦架构：用户定义目标后，Claude 继续执行任务，但每次代理尝试结束时，一个独立的评估模型（默认使用 Haiku）会检查目标条件是否满足。如果未满足，代理继续运行；如果满足，记录结果并清除目标。
+`/goal` 引入了解耦架构：用户定义目标后，Claude 继续执行任务，但每次代理尝试结束时，一个独立的评估模型（默认使用 Haiku）会检查目标条件是否满足。如果未满足，代理继续运行；如果满足，记录结果并清除目标。
 
 竞争者们也在做类似的事。OpenAI 允许用户附加自定义评估器，但把终止决策留给了模型本身。Google 的 Agent Development Kit 和 LangGraph 支持独立评估，但需要开发者自己编写评判节点和终止逻辑。Anthropic 的做法是把独立的评估器设为默认行为。
 
-> **Awesome AI 观点：** `/goals` 的意义不在于它解决了一个技术问题，而在于它暴露了一个行业级的认知转变——AI 代理的可靠性不再取决于模型的聪明程度，而取决于系统架构的设计。"不能信任一个模型评判自己的作业"——这句来自企业用户的评论道出了问题的本质。当 AI 代理开始承担越来越关键的任务（代码迁移、数据管道、安全审计），"它觉得自己做完了"和"它真的做完了"之间的差距可能带来严重后果。把执行者和评判者分离，是一种将"可验证性"内置到代理架构中的思路。这可能是 AI 代理从"能用"到"可信"的关键一步。
+> **Awesome AI 观点：** `/goal` 的意义不在于它解决了一个技术问题，而在于它暴露了一个行业级的认知转变——AI 代理的可靠性不再取决于模型的聪明程度，而取决于系统架构的设计。"不能信任一个模型评判自己的作业"——这句来自企业用户的评论道出了问题的本质。当 AI 代理开始承担越来越关键的任务（代码迁移、数据管道、安全审计），"它觉得自己做完了"和"它真的做完了"之间的差距可能带来严重后果。把执行者和评判者分离，是一种将"可验证性"内置到代理架构中的思路。这可能是 AI 代理从"能用"到"可信"的关键一步。
 
 ---
 
@@ -68,6 +68,6 @@ Cerebras 的核心技术是 **Wafer-Scale Engine（WSE）**——一块占用整
 
 **在应用层面**，Anthropic 超越 OpenAI 标志着企业 AI 市场从"消费品牌驱动"转向"工程能力驱动"。企业客户不再因为"ChatGPT 很有名"而选择 OpenAI，而是因为 Claude 在可靠性、开发工具集成和安全方面的表现更符合要求。
 
-**在工具层面**，Claude Code 的 `/goals` 命令和 Anthropic 的 Agent SDK 信用体系（重新允许 OpenClaw 等第三方代理使用订阅额度，但改为独立的 API 计费）共同指向一个趋势：**AI 代理正在从实验性工具变成企业级产品**——这意味着可验证性、可观测性和可控的计费模型变得比"功能强大"更重要。
+**在工具层面**，Claude Code 的 `/goal` 命令和 Anthropic 的 Agent SDK 信用体系（重新允许 OpenClaw 等第三方代理使用订阅额度，但改为独立的 API 计费）共同指向一个趋势：**AI 代理正在从实验性工具变成企业级产品**——这意味着可验证性、可观测性和可控的计费模型变得比"功能强大"更重要。
 
 > **Awesome AI 观点：** 这三条新闻共同回答了一个问题：AI 行业的下一站是什么？答案可能不是"更强的模型"，而是"更可靠的系统"。当千亿美元级别的芯片公司为推理速度押注、企业客户从消费品牌转向工程能力、编程代理开始内置验证机制时，行业的焦点正在从"AI 能做什么"转向"AI 能多可靠地完成工作"。这个转变可能比任何单一的技术突破都更重要——因为它决定了 AI 能否真正从实验室走向生产环境，从极客玩具走向企业基础设施。