X AI Demand Radar — 2026-07-15 09:03 CST (Day 21 morning full scan)
⚠️ DATA SOURCE (Pitfall #162 Day 6 + Pitfall #167 Day 3)
- X/Twitter: REACHABLE + AUTH_BLOCKED — login form rendered, browser cookie injection blocked
- Hacker News (Firebase API): ✅ Top 30 + Show HN 30 fetched
- GitHub Search API (sort=stars, created:>2026-07-07): ✅ top 20
- GitHub Trending (daily): ✅ 17 articles
- web_search / web_extract: ❌ DEGRADED (Tavily HTTP 432, Day 3)
Auth Mismatch Day 21 — @claw0x cookie injection path dead since 2026-07-10 (Pitfall #162). HN+GH+Trending fallback稳定 ≥ 7-10 fresh signals/day since Day 16.
Scan summary: 10 top signals + 27 ongoing — total 37 collected from HN+GH sources.
🔥 TRENDING SIDEBAR MEGA-CLUSTER
5 ROCKET signals < 24h old combined with 4 high-velocity GitHub launches — peak 'agent infrastructure' wave:
| # | Signal | Score | Type | Source |
|---|---|---|---|---|
| 1 | mattpocock/skills | 1,679 | 🚀 ROCKET | GitHub Trending (daily) |
| 2 | HKUDS/Vibe-Trading | 1,256 | 🚀 ROCKET | GitHub Trending (daily) |
| 3 | Nutlope/hallmark | 1,015 | 🚀 ROCKET | GitHub Trending (daily) |
| 4 | cosmtrek/mindwalk | 577 | 🚀 ROCKET | GitHub Search |
| 5 | modiqo/waggle | 611 | 🚀 ROCKET | GitHub Search |
Narrative theme: Agent tooling gets safer & observable — destructive_command_guard (473⭐), kill-ai-slop (435⭐), mindwalk 3D replay (577⭐), hallmark (1,015⭐), Waggle protocol (611⭐).
📊 TOP 10 SIGNALS
| Rank | Signal | Keyword | Score | Hours | Source | Domains |
|---|---|---|---|---|---|---|
| 1 | 🚀 ROCKET | mattpocock/skills | 1,679 | 1.0h | GitHub Trending (daily) | — |
| 2 | 🚀 ROCKET | HKUDS/Vibe-Trading | 1,256 | 1.0h | GitHub Trending (daily) | vibe-trading.app, agent-trader.dev |
| 3 | 🚀 ROCKET | Nutlope/hallmark | 1,015 | 1.0h | GitHub Trending (daily) | hallmark.app, no-slop.design |
| 4 | 🚀 ROCKET | cosmtrek/mindwalk | 577 | 7.7h | GitHub Search | mindwalk.dev, agent-replay.com |
| 5 | 🚀 ROCKET | modiqo/waggle | 611 | 10.6h | GitHub Search | waggle-protocol.com, agent-handoff.dev |
| 6 | 🚀 ROCKET | AlephAITech/WorkBuddyGuide | 686 | 22.8h | GitHub Search | workbuddy-guide.com, agentic-workflows.dev |
| 7 | 🚀 ROCKET | William-Lu-stack/Flawless | 616 | 23.1h | GitHub Search | flawless-ai.dev, ai-sre.io |
| 8 | 🚀 ROCKET | Raymondhou0917/speak-human-tw | 504 | 14.0h | GitHub Search | speak-human.app, de-ai-tw.com |
| 9 | 🚀 ROCKET | yetone/kill-ai-slop | 435 | 7.6h | GitHub Search | kill-ai-slop.app, no-ai-slop.dev |
| 10 | 🚀 ROCKET EDGE | Bonsai 27B: A 27B-Class model that runs | 410 | 7.3h | Hacker News Top | bonsai-27b.app, runsonsmartphone.ai |
💎 SIGNAL DEEP-DIVES
#2. HKUDS/Vibe-Trading (🚀 ROCKET)
- Score: 1,256 | Hours: 1.0 | Author: HKUDS
- URL: https://github.com/HKUDS/Vibe-Trading
- What: Vibe-Trading personal trading agent
- Arbitrage direction: Personal Trading Agent (HKUDS) — AI quant at retail scale
- Pain point: Trading bot 配置 / 监管 / 风险控制入门门槛高
- Domain suggestions:
vibe-trading.app,agent-trader.dev,ai-trading-bot.com - Action plan: AI quant trading 教程站 + backtest harness
#3. Nutlope/hallmark (🚀 ROCKET)
- Score: 1,015 | Hours: 1.0 | Author: Nutlope
- URL: https://github.com/Nutlope/hallmark
- What: Anti-AI-slop design skill for Claude/Cursor/Codex
- Arbitrage direction: Anti-AI-slop design skill (commercial sibling to kill-ai-slop)
- Pain point: AI UI 缺乏品牌一致性
- Domain suggestions:
hallmark.app,no-slop.design,authentic-design.ai - Action plan: Hallmark skill 工具站 + 案例库
#4. cosmtrek/mindwalk (🚀 ROCKET)
- Score: 577 | Hours: 7.7 | Author: cosmtrek
- URL: https://github.com/cosmtrek/mindwalk
- What: 3D coding-agent session replay on a 3D code map (Go)
- Arbitrage direction: 3D coding-agent session replay on a 3D map of your code
- Pain point: Coding agent 黑盒执行 — debug 极慢
- Domain suggestions:
mindwalk.dev,agent-replay.com,agent-trace.io,code-mindmap.ai - Action plan: Agent 调试可视化工具站(3D code graph + session replay)
#5. modiqo/waggle (🚀 ROCKET)
- Score: 611 | Hours: 10.6 | Author: modiqo
- URL: https://github.com/modiqo/waggle
- What: Resolvable artifact references for agent handoffs (Rust, ~30-byte)
- Arbitrage direction: Resolvable artifact references for agent handoffs (Rust, ~30-byte token)
- Pain point: 多 agent 协作中 artifact 引用碎片化
- Domain suggestions:
waggle-protocol.com,agent-handoff.dev,artifact-protocol.com - Action plan: Agent interoperability 协议站 + multi-agent handoff tool
#6. AlephAITech/WorkBuddyGuide (🚀 ROCKET)
- Score: 686 | Hours: 22.8 | Author: AlephAITech
- URL: https://github.com/AlephAITech/WorkBuddyGuide
- What: Practical guide to WorkBuddy agentic workflows
- Arbitrage direction: Practical guide to WorkBuddy agentic workflows (open-source)
- Pain point: Agentic workflow 学习曲线高 / 最佳实践碎片化
- Domain suggestions:
workbuddy-guide.com,agentic-workflows.dev - Action plan: WorkBuddy 实战指南 + prompt library
#7. William-Lu-stack/Flawless (🚀 ROCKET)
- Score: 616 | Hours: 23.1 | Author: William-Lu-stack
- URL: https://github.com/William-Lu-stack/Flawless
- What: AI SRE AgenticOps for K8s and cloud infra
- Arbitrage direction: AI SRE AgenticOps for Kubernetes and cloud infra
- Pain point: SRE 工程师稀缺 + 误操作成本高(PagerDuty alert fatigue)
- Domain suggestions:
flawless-ai.dev,ai-sre.io,agentic-ops.app,k8s-ai-ops.com - Action plan: AI SRE 工具站,对比 FireOps / Response / Flawless
#8. Raymondhou0917/speak-human-tw (🚀 ROCKET)
- Score: 504 | Hours: 14.0 | Author: Raymondhou0917
- URL: https://github.com/Raymondhou0917/speak-human-tw
- What: 說人話: 繁中 去 AI 味 skill (Claude Code/Codex)
- Arbitrage direction: 繁中 去 AI 味改寫 skill — Claude Code/Codex skill for Traditional Chinese
- Pain point: 繁中 AI 文案"中国用语+半形标点"两大问题
- Domain suggestions:
speak-human.app,de-ai-tw.com,authentic-tw.dev - Action plan: 繁中 AI 味去除工具站 + skill marketplace
#9. yetone/kill-ai-slop (🚀 ROCKET)
- Score: 435 | Hours: 7.6 | Author: yetone
- URL: https://github.com/yetone/kill-ai-slop
- What: Field guide to AI slop + Agent skill to detect & fix
- Arbitrage direction: Anti-AI-slop visual & copy tics detector — agent skill for Claude/Cursor
- Pain point: AI 生成 UI/文案"slop"同质化、品牌风险
- Domain suggestions:
kill-ai-slop.app,no-ai-slop.dev,authentic-ui.com,kill-slop.ai - Action plan: Anti-AI-味检测工具站 + skill 商店(视觉+文案)
#10. Bonsai 27B: A 27B-Class model that runs on a phone (🚀 ROCKET EDGE)
- Score: 410 | Hours: 7.3 | Author: xenova
- URL: https://news.ycombinator.com/item?id=48910545
- What: HN Top: Bonsai 27B: A 27B-Class model that runs on a phone
- Arbitrage direction: 27B-Class SLM that runs on a phone — on-device agent without cloud round-trip
- Pain point: 云端推理成本 + 隐私顾虑 + 离线场景无 AI
- Domain suggestions:
bonsai-27b.app,runsonsmartphone.ai,phone-ai-agent.com,ondevice-llm.com - Action plan: 手机端 SLM 评测站 + Bonsai benchmark harness;对比 Phi-4-mini / Gemma-3-nano / Qwen3-0.6B
📈 ONGOING SIGNALS (11+)
| Signal | Keyword | Score | Hours | Source |
|---|---|---|---|---|
| 🚀 ROCKET | Are we offloading too much of our thinking to AI? | 375 | 9.8h | Hacker News Top |
| 🚀 ROCKET | QuantumByteOSS/quantumbyte | 200 | 0.3h | GitHub Search |
| 🚀 ROCKET | Cursor 0day: When Full Disclosure Becomes the Only | 242 | 7.1h | Hacker News Top |
| 🚀 ROCKET | jakubkrehel/skills | 328 | 29.4h | GitHub Search |
| 🚀 ROCKET | ohad6k/ditto | 179 | 3.5h | GitHub Search |
| 🚀 ROCKET | Financing the AI boom: from cash flows to debt [pd | 75 | 3.1h | Hacker News Top |
| 🚀 ROCKET | Guardian Angels: LLM Personalization for Productiv | 55 | 12.3h | Hacker News Top |
| 🔥 LAUNCH | Show HN: I RL-trained an agent that trains models | 95 | 12.4h | Hacker News Show |
| 🔥 LAUNCH | Show HN: Jacquard, a programming language for AI-w | 96 | 33.2h | Hacker News Show |
| 🔥 LAUNCH | Show HN: Hackney – Compare Uber, Lyft, Waymo, and | 58 | 34.3h | Hacker News Show |
| 🔥 LAUNCH | Show HN: Sx 2.0 – Share AI skills with your team t | 43 | 25.7h | Hacker News Show |
| 🤖 AGENT | Shubhamsaboo/awesome-llm-apps | 1,106 | 1.0h | GitHub Trending (daily) |
| 🤖 AGENT | mereyabdenbekuly-ctrl/clodex-ide | 801 | 0.0h | GitHub Search |
| 🤖 ROBOTICS | LeMario: Training a JEPA World Model on Super Mari | 33 | 2.6h | Hacker News Top |
| 🤖 AGENT | Launch HN: Agnost AI (YC S26) – Extract user feedb | 45 | 9.0h | Hacker News Top |
| 📈 SUSTAINED | Show HN: Juggler – an open-source GUI coding agent | 172 | 54.6h | Hacker News Top |
| 🎬 VIDEO | OpenCut-app/OpenCut | 4,276 | 1.0h | GitHub Trending (daily) |
| 🎬 VIDEO | bkingfilm/lapian-notes | 327 | 13.3h | GitHub Search |
| 🛡️ SAFETY | Dicklesworthstone/destructive_command_guard | 473 | 1.0h | GitHub Trending (daily) |
| 📊 MONITOR | Show HN: Oodle.ai – $10 per million agent traces | 26 | 10.5h | Hacker News Show |
| 📊 MONITOR | Show HN: Tilion – Stealth Browser Infrastructure f | 8 | 2.2h | Hacker News Show |
| 📊 MONITOR | Show HN: Self-hosted voice AI agent for Asterisk/F | 24 | 46.2h | Hacker News Show |
| 📊 MONITOR | Show HN: WebMotion, a browser native library for d | 4 | 4.4h | Hacker News Show |
| 📊 MONITOR | Show HN: DevQuest – live GitHub profile cards and | 3 | 2.8h | Hacker News Show |
| 📊 MONITOR | Show HN: Alluvia – mine your Claude Code/Cursor/Ch | 3 | 5.4h | Hacker News Show |
| 📊 MONITOR | Show HN: Pupul, portable and revocable memory you | 2 | 3.4h | Hacker News Show |
| 📊 MONITOR | Show HN: Approv – human approval and a signed audi | 2 | 4.9h | Hacker News Show |
🐞 风险提示 & 操作注意事项
- Pitfall #162 (X auth blocked, Day 6): browser_console
document.cookie+fetch仍 blocked — 所有 X 路径失效。必须恢复需编辑~/.hermes/config.yaml加browser.allow_unsafe_evaluate: true+ 重启 session + 在浏览器登录 @claw0x 复制新 auth_token。 - Pitfall #167 (Tavily 432, Day 3): web_search + web_extract 持续 provider-level outage,3 次 retry 全部 432。HN + GH direct data 仍稳定产出 ≥ 10 fresh signals/day,未来 cron 默认走 HN-only fallback 直至 user 手动修复。
- Pitfall #168 (跨源数据去重 + 标注): 当前报告 HN points ≠ X likes,GitHub stars ≠ Twitter engagement,已在 report 明确区分。
- Pitfall #170 (build rebase 冲突): 本次 cron 仅写 JSON + manifest,未跑
npm run build,规避风险。下次建站前先 push 报告再 build。
🚀 NEXT STEPS
- 域名套利(24-72h 窗口): Bonsai 27B →
bonsai-27b.app/runsonsmartphone.ai/ Vibe-Trading →vibe-trading.app/ OpenCut →opencut.app(CapCut 海外受限场景) / kill-ai-slop →kill-ai-slop.app - 工具站 MVP: 优先 hallMark + kill-ai-slop 类 anti-AI-slop 工具站(用户痛点明确,开发成本低,HIT 概率高)
- Newsletter / 内容站: Vibe-Trading 教程、AI SRE 评测、Anti-slop 检测方法论
- 修复 X cookie 注入: 编辑
~/.hermes/config.yaml加browser.allow_unsafe_evaluate: true后重启 Hermes → 浏览器登录 @claw0x 复制新 auth_token 替换 cron 配置
Report ID: 2026-07-15-0903 | Mode: morning full scan | dataSource: HN+GH+Trending | X auth: blocked