Show HN: Shoggoth Mini – A soft tentacle robot powered by GPT-4o and RL

Overall reactions

• Many find the tentacle robot both impressive and unsettling, with lots of horror, hentai, and sci‑fi references (Alien, Matrix, Minority Report, Lovecraft, Spider‑Man’s Doc Ock, Pixar lamp).
• People appreciate that it looks distinctly robotic rather than like a real organism; some explicitly prefer a future where robots are visually distinguishable from nature.

Patent, prior art, and SpiRobs

• Commenters recall that the underlying “SpiRobs” tentacle mechanism is being patented; someone confirms a specific pneumatic continuum robot patent filing.
• Discussion clarifies that publishing a paper doesn’t block a patent if the authors themselves file; in many jurisdictions, prior public disclosure by others does, with the US having a grace period.
• Some express frustration that “obvious” ideas get patented, though concrete examples are not clearly substantiated in the thread.

Model choice, latency, and local inference

• Several are surprised it uses GPT‑4o rather than a small local model or specialized vision model; others argue the big model supports richer future behaviors (e.g., multiple tentacles, locomotion).
• Latency is called out as “unnerving,” especially when the robot freezes while waiting for a cloud response; suggestions include eye LEDs or animation to signal “thinking.”
• Multiple comments propose tiny LLMs (e.g., ~0.6B parameters, quantized) running on modest hardware, or a hybrid: fast on‑device model for instant back‑channel, larger remote model for deeper reasoning.
• Wake‑word engines are proposed to avoid “continuously listening,” reduce energy usage, and enable wireless operation.

Expressiveness, aliveness, and human psychology

• The author’s observation that the robot initially feels “alive” but becomes predictable sparks a long tangent:
  • Comparisons to games that lose magic once min‑max strategies or procedural patterns are understood.
  • Furbies as a similar early example: initially magical, then obviously finite state machines.
• Debate over whether humans and LLMs are categorically different or just differ in degree; several note our tendency to anthropomorphize anything with semi‑complex behavior or voice.
• Some argue that utility, not “fake aliveness,” will matter; others foresee ethical questions once robots reach higher apparent agency or “slave” status.

Applications, toys, and ethics

• Some imagine stuffed animals or Tamagotchi‑like devices using similar tech to engage children, while others react strongly against “subscription best friends” and ad‑driven companion toys.
• There’s skepticism that robot pets will ever truly replace the “essential element” of life present in real animals.

Technical & domain context

• Commenters identify this as a “continuum robot,” noting substantial research and medical applications, and link to lectures and the SpiRobs inspiration video.
• A few worry about a future where LLM‑enabled expressive devices permeate appliances (“fridge that cries”), echoing concerns about over‑embedding AI in everyday objects.