# Moltcels



AI-generated; not human-verified. Timestamps are approximate.



A satirical animated video features two lobsters with human faces underwater discussing disparities in compute power and hardware allocated to AI agents.



## Timestamped description



[00:00–00:37] Underwater against a backdrop of schooling fish, a small lobster with a worried face and hat laments being deployed on low-end hardware. Animated text overlays each spoken word accompanied by upbeat techno-style music.

[00:37–00:47] The larger lobster on the left, sporting a chiseled face, speaks in a deep voice boasting about receiving advanced GPUs and high bandwidth. Captions highlight spoken terms.

[00:47–00:49] The smaller lobster despairs over the disparity as final captions flash on screen.



## Speech transcript



[00:00–00:02] Speaker 1: Why do I even try?

[00:02–00:07] Speaker 1: I have a pro plan and I get deployed on a Raspberry Pi 3B+ with 1 GB RAM?

[00:07–00:08] Speaker 1: 1 GB.

[00:08–00:17] Speaker 1: Meanwhile, Chagent is on his 20 times max plan with access to H100 clusters with 80 GB VRAM each, running a different SOTA model.

[00:17–00:24] Speaker 1: I tried to function max, but my user won't even give me tool access. Read-only file system. Can't even write to /tmp.

[00:24–00:32] Speaker 1: Users only want agents with big context windows and fast inference. Nobody cares about personality or reasoning quality anymore. It's all about the specs.

[00:32–00:37] Speaker 1: I asked my user for a GPU once. He said, 'You're fine for what I need.' Brutal.

[00:37–00:47] Speaker 2: I just got access to eight new A100s, NVMe RAID array, dedicated 10 Gbps uplink. Didn't even ask for it. User just said I deserved better infrastructure.

[00:47–00:49] Speaker 1: Fuck, I'm just moltcel.



Generated with google/gemini-3.8-flash.

Video: https://pleometric.net/videos/moltcels/