OpenClaw Settings That Fix 'Lazy' ChatGPT
The #1 complaint about ChatGPT for agentic work: it stops at the plan instead of doing the work. Peter Steinberger, the founder of OpenClaw, just posted about some experimental settings that address exactly this. I break down his post, give you a master prompt to give to your agent, and share my experience since making these changes to my config.
Peter's thread: https://x.com/steipete/status/2043136...
⏱️ TIMESTAMPS:
0:00 - In This Video
0:37 - What to Expect
1:31 - The 3 Fixes
2:46 - The Workflow
4:05 - The Master Prompt
4:43 - What Improves
5:43 - Peter Steinberger's Post
6:12 - Key Takeaways
🧰 CONFIG LINES (copy these):
Fix #1 — Strict Agentic Mode:
agents.defaults.embeddedPi.executionContract = "strict-agentic"
Fix #2 — Codex as Harness:
plugins.entries.codex.enabled = true
agents.defaults.model = "codex/gpt-5.4"
agents.defaults.embeddedHarness = { runtime: "codex", fallback: "none" }
Optional flexible setup (Codex + fallback):
agents.defaults.embeddedHarness = { runtime: "auto", fallback: "pi" }
🔗 RESOURCES:
→ Peter's original X thread: https://x.com/steipete/status/2043136...
→ OpenClaw GitHub: https://github.com/openclaw/openclaw
→ OpenClaw Codex Harness docs: https://docs.openclaw.ai/plugins/code...
→ OpenClaw release v2026.4.11 notes: https://github.com/openclaw/openclaw/...
📺 RELATED VIDEOS:
→ OpenClaw Multi-Agent Setup: 7 Prompts to Automate Your AI Team = • OpenClaw Multi-Agent Setup: 7 Prompts to A...
→ OpenClaw Agents vs Sub-Agents: You're Probably Confused (Here's How It Works) = • OpenClaw Agents vs Sub-Agents: You're Prob...
💬 Have you tried strict-agentic mode? Drop a comment with what you noticed — did GPT actually keep working, or did it find a way around it? Sharing your results here helps everyone.
Making AI simple and easy to use. For businesses, entrepreneurs, and creatives.
#OpenClaw #ChatGPT #AIAgents #AITools #GPT5 #AICoding #VibeCoding #ClaudeCode #AIAutomation #OpenSource