OpenClaw Newsletter

• Claude Access Issues: Anthropic banning OpenClaw's creator from Claude access highlights gated AI model access concerns — "brutal reminder that access to AI models is gated, not open"
• Hermes vs OpenClaw Comparison: Chinese users discussing Hermes being more reliable than OpenClaw — "更听话故障更少" (more obedient, fewer failures)
• Local Setup Success: Japanese user running OpenClaw locally on 32GB RAM/16GB VRAM PC to avoid API costs — considers electricity cheaper than AI subscriptions
• WhatsApp Integration: Developer successfully connected OpenClaw to WhatsApp despite temporary blocking — "The agent now lives where you chat, not where you go to find it"
• Migration Discussion: Community actively comparing Hermes Agent vs OpenClaw migration costs and questioning necessity of switching platforms

• openclaw 2026.4.10 — Added bundled Codex provider and plugin-owned app-server harness for Codex/GPT models with native auth, threads, model discovery, and compaction. Download
• npm package 2026.4.10 — Multi-channel AI gateway with extensible messaging integrations now live on npm with 1.5M weekly downloads. Install
• Repository milestone — OpenClaw hits 354k+ stars with 71k forks and 1,705 contributors across the project.

• knightplat-blip's Windows encoding fix tackles garbled Chinese characters in exec tools by properly handling GBK encoding with TextDecoder — 26 community interactions
• Telegram voice-note transcription regression silently fails in 4.5 due to undefined media paths breaking audio attachment handling — 16 interactions from affected users
• carrotRakko's CJK search enhancement adds configurable FTS5 tokenizer support for Chinese/Japanese/Korean text search where word boundaries don't exist — solid community engagement
• OpenAI Codex OAuth breakdown marks a regression blocking beta releases, with users unable to complete authentication flows — active developer discussion
• Memory-core narrative generation failing due to missing idempotencyKey property, causing 30-minute warning cycles for users — fresh reports emerging

• Google restricting Google AI Pro/Ultra subscribers for using OpenClaw — Google is blocking users who authenticate their AI subscriptions with OpenClaw, sparking controversy over third-party access rights. Discussion thread with 802 points, 705 comments on HN
• OpenClaw privilege escalation vulnerability [CVE-2026-33579] — Security researchers discovered a privilege escalation flaw in OpenClaw installations that could allow attackers to gain elevated access. Official CVE entry with 514 points, 256 comments on HN
• "OpenClaw's memory is unreliable, and you don't know when it will break" — Deep dive into OpenClaw's memory management issues based on analyzing 1000+ production deployments, highlighting when the system fails silently. Blog post with 108 points, 121 comments
• Ask HN: Share your productive usage of OpenClaw — Community discussion revealing real-world OpenClaw use cases and workflows from power users across different industries. HN thread with 106 points, 87 comments
• ValueClaw - Open-source autonomous financial AI agent — New Python-based AI agent with 97 built-in skills for financial analysis, designed to replicate investment strategies of Warren Buffett and Ray Dalio. Medium article

• Malicious Skill Trap: Bitdefender Labs identifies hundreds of malicious OpenClaw skills disguised as legitimate helpers that could execute hidden payloads. Review and audit your installed skills immediately. Read analysis