1. OpenClaw is hyped but largely ineffective
Many commenters point out that the product’s claims are over‑blown and that the real‑world results are disappointing.
“It’s a shitshow.” – ricardobayes
“I deleted it and set up something much simpler.” – mikenew
“There is no evidence this is the case.” – enraged_camel
2. Security and prompt‑injection are major red‑flags
The discussion repeatedly highlights how OpenClaw’s unrestricted access can be abused, and that current mitigations are weak.
“The only real solution is to never give it untrusted data or access to anything you care about.” – habinero
“It can combine prompt injection with access to sensitive systems and write access to the internet.” – madeofpalk
3. The “manager‑in‑a‑box” narrative is misleading
Users argue that the idea of an AI that lets you “be a CEO” is a fantasy; real work still requires human oversight and the role of a manager is more complex than a chatbot.
“I’m not going to waste my time reading this AI‑generating post.” – phito
“You still have to jump into the project, set up the environment, open my editor and Claude Code terminal.” – yellow_lead
4. Lack of concrete examples or measurable results
Critics demand real code, projects, or metrics to back up the claims, and most posts fail to provide them.
“Show the code, the projects, or at least a tiny snippet of code.” – fullstackchris
“If you’re going to claim you built something, link to the repo or the product.” – charles_f
These four themes capture the core of the conversation: hype vs reality, security concerns, the unrealistic “AI manager” trope, and the absence of tangible evidence.