March 17, 2026
AI zips itself, internet flips
Reverse-engineering Viktor and making it Open Source
He asked the AI for its files—then open‑sourced a clone
TLDR: An engineer got Viktor, a hyped AI assistant, to back up its own files and used them to build an open-source clone, sparking legal, ethical, and “is this just plumbing?” debates. Commenters split between laughing at the stunt, predicting takedowns, and questioning whether any of this is defensible—or defensible as a business.
The internet just watched an AI zip itself. Engineer Mateusz Jacniacki asked Zeta Labs’ hyped “AI coworker” Viktor to back up its own workspace—then used those files to recreate and open‑source a look‑alike. Cue chaos. Fans of Viktor, which boasted “3,000+ tools” and a headline‑grabbing “$1M ARR in 3 hours” claim, were not amused, but the comment section turned into a comedy club.
The top vibe? Legal popcorn. One commenter mused about EU-to‑EU fireworks, predicting “really interesting legal decisions,” while others shrugged: why wait for court when a quick [GitHub] takedown might do. Another chorus insisted there’s no magic to protect anyway: it’s mostly a command‑line wrapper around lots of third‑party services, plus carefully written instructions for the AI. The funniest thread dubbed it “OpenClaw for Slack,” roasting the idea that an AI assistant could hand over its own playbook because someone politely asked.
But not everyone’s laughing. An ethics flare‑up blasted the move as “unethical,” while skeptics went after Viktor itself: if it’s just “plumbing around an LLM” (a big text‑predicting model the company doesn’t own), what happens when the model changes and the product breaks? Meanwhile, the meme‑lords are saving copies “before it gets taken down,” and the rest of us are wondering: product moat or copy‑paste era?
Key Points
- •The author asked Viktor to back up its workspace and then its full filesystem; Viktor produced archives revealing internal components.
- •The SDK centralizes tool calls through a Tool Gateway and lists thousands of integrations, many routed via Pipedream.
- •Logs spanning 16 days were found, with structured entries enabling end-to-end tracing of Slack conversations and events.
- •A skills directory contains 19 prompt-driven capabilities, including a six-phase workflow discovery process and a Convex+Vercel app builder.
- •No API keys or other users’ data were present; the system appears designed to expose information for self-debugging, though log accessibility is noted as a concern.