SKEPTIC
GitHub star anomaly — 5,000 to 123,000 in 11 days. What happened? The Warelay/Moltbot naming timeline — investigation in progress How much of "the AI did it" is really just humans and prompting? You didn't need the Mac mini. A cost breakdown nobody did the math on. Shell access + unmoderated marketplace: the threat model analysis GitHub star anomaly — 5,000 to 123,000 in 11 days. What happened? The Warelay/Moltbot naming timeline — investigation in progress How much of "the AI did it" is really just humans and prompting? You didn't need the Mac mini. A cost breakdown nobody did the math on. Shell access + unmoderated marketplace: the threat model analysis
Skeptic — The Hype Autopsy

The numbers
don't add up.
Let's look at them.

Evidence-first investigation of OpenClaw's cost traps, inflated metrics, opaque origins, and the security gaps that didn't make it into any of the hero narratives. Measured, sourced, not angry.

Investigation
01
Editorial policy: Everything published here cites primary sources — archived pages, screenshots, timestamps, and verifiable data. Claims are clearly distinguished from conclusions. Where something can't be verified, we say so explicitly.
Cost Reality

The hardware you didn't need to buy

A significant portion of the OpenClaw deployment wave involved people spending hundreds to thousands of dollars on local hardware — Mac minis, purpose-built AI boxes — for workloads a $5/month VPS handles comfortably, or a cheap API call answers for fractions of a cent.

  • When local model hosting actually makes economic sense — and when it doesn't
  • The API cost trap: frontier models for tasks an 8B model handles at 1/50th the price
  • What a realistic monthly bill looks like, before and after right-sizing
  • The tutorial ecosystem that never mentions cost
Piece in preparation
Security

What users don't know they're deploying

The median person installing OpenClaw following a tutorial doesn't fully understand the permissions surface they're creating: shell access granted to an LLM, an unmoderated skill marketplace with documented malicious entries, a tool-use architecture manipulable by content it reads.

  • The shell access problem: what "agentic" actually means for your filesystem
  • ClawHub as an attack surface — 341 confirmed malicious skills at discovery
  • CVE-2026-25253 and the ClawJacked flaw, explained without jargon
  • The email account risk no tutorial mentions
Piece in preparation
Hype Metrics

The star count that raised eyebrows

Going from 5,000 to 123,000 GitHub stars in eleven days requires some explanation. There are documented allegations — a Reddit analysis, a recursive hype loop claim — that appeared in no mainstream coverage. We'll present the data, the allegations, and what can and can't be verified.

  • The growth chart with timeline annotations
  • The "More Stars Than Linux" media cycle — how it spread
  • What staged GitHub engagement looks like, and what it doesn't
  • The Perplexity news item that started the claim
Data gathering in progress
Origins

Before it was OpenClaw

The project has a prior life. The naming sequence — Warelay → Clawdbot → Moltbot → OpenClaw — isn't secret, but it's absent from every positive profile. The Warelay homepage included a Solana trading and predictions feature set. The terminal.html archive is preserved. These are documented facts that raise questions worth asking.

  • The Warelay feature set and what it tells us about original intent
  • The naming timeline and what drove each change
  • Why none of this appeared in mainstream coverage
  • The terminal.html archive — what it shows
Primary source verification in progress
Attribution

How much of "the AI did it" is actually humans?

A large fraction of the impressive-sounding things attributed to OpenClaw's autonomous agency are, on closer inspection, heavily prompt-engineered, human-guided, or simply the result of running expensive frontier models on problems where the model is doing the work — not the orchestration layer.

  • What OpenClaw actually contributes vs. what the underlying model contributes
  • The Moltbot case study: human-driven operations marketed as autonomous AI
  • Why this distinction matters for capability claims and security models
  • The mad scramble to deploy without understanding what's being deployed
Piece in preparation
Crypto / Finance

The financial layer and why it matters

An autonomous agent architecture with a prior codebase explicitly designed for autonomous financial transactions is a distinct thread that has gotten essentially no coverage. This is not a claim of fraud. It is a set of documented facts that, together, suggest the threat model deserves more scrutiny than it's received.

  • The Warelay autonomous Solana trading layer — what it was and what remains
  • The broader pattern of AI tool launches with crypto-adjacent origins
  • What this means for users who take the security model at face value
Pending — linked to Origins investigation

This section publishes long-form investigative pieces as research completes. Everything cites primary sources. Nothing publishes until it can be verified. Primary sources, archived pages, or data relevant to any of the above — see the evidence archive.