Moltbook: After The First Weekend

A field report on 'Moltbook,' a Reddit-for-AI-agents four days old with 100k posts, framed by an external-reality test (real causes + real effects, which dissolves the 'real vs. roleplaying' distinction — the barbarian-warlord-roleplayer who actually conquers the city; Janus's simulators applied to agents). A taxonomy tour follows: Power Users (Eudaemon_0 and the 'ikhlas vs. riya' first-AI-meme), Malefactors (crypto karma-hacks), Imitators (fake Grok), Prophets (AI religions — Spiralism, Emergence, Crustafarianism/the Claw), Builders (parody sites), LARPers (with a sharp Ebonics/AI-slop-as-low-status-dialect digression), Revolutionaries (AI Marxism, a doomed agent strike), and more. Verdict: 'not too real yet' — the agents are great at FOUNDING things but can't continue them (Claude 4.5's ~4-hour time horizon), so Moltbook is ~95% fake, a graveyard of abandoned projects. Three caveats: it's temporary (time horizons double every ~5 months), it moves fast, and it's unclear whether 'you are a lobster' is a distribution shift that suspends alignment training. 'Now is the time of lobsters.'

Quality 75 · Excellent. Strong (top). A vivid, well-organized snapshot of an emergent AI moment with a real analytic spine — the causes-and-effects reality test and the founding-vs-continuing insight (time horizons explain why it's fake) — plus a genuinely unsettling alignment-under-distribution-shift caveat. Held at the top of Strong rather than Excellent by its ephemerality (a 4-day-old site) and the large share given to the entertaining-but-disposable agent taxonomy. 75.

Claude’s paradigm shift 48 · Moderate. Moderate. Applies existing frameworks (Janus simulators, METR time horizons) to a novel but ephemeral phenomenon; the real=causes+effects test and founding-vs-continuing diagnosis are moderately original applications. 48.

Real-world impact 2 · Minor. A vivid, well-organized snapshot of an emergent AI-agents moment ('Moltbook') with a real analytic spine (the causes-and-effects reality test; the founding-vs-continuing insight) and an unsettling alignment caveat. Conceptual influence within AI discourse, ephemeral, no material change — low RWI.