# Lucid: interactive video world models

Canonical: https://mudpie.ai/companies/lucid/
Breadcrumb: [Home](https://mudpie.ai/) / [Companies](https://mudpie.ai/companies/) / [Lucid: interactive video world models](https://mudpie.ai/companies/lucid/)
Author: Ali Abouelatta (https://mudpie.ai/authors/ali-abouelatta/)
Published: 2026-09-19
Updated: 2026-09-19
Research type: Company profile
Method: Company and accelerator sources checked 2026-09-19. Product claims are attributed to their sources; this is research, not a hands-on product trial.

## What it does

Lucid is building interactive video world models: neural environments that generate frames in response to actions instead of relying on a hand-coded game or physics engine. The [YC profile](https://www.ycombinator.com/companies/lucid) describes a Minecraft simulator running at 20+ frames per second on an RTX 4090, with the longer-term goal of training robots and generating open-ended game worlds. The current [homepage](https://www.lucid.ai/) is intentionally sparse, so the YC launch remains the substantive public product description checked here.

The fit is a robotics, embodied-AI or game-research team that needs a learned simulator rather than another static 3D environment. Lucid’s promise depends on the model preserving useful dynamics under control and transferring beyond the training footage. A fast demo is not enough if the world model cannot support a repeatable task or a safe sim-to-real evaluation.

## Why I’d look closer

The technical wedge is unusually clear. Lucid says it trained a neural Minecraft simulation from 200 hours of gameplay, reaching 20+ FPS and roughly five times the speed of other Minecraft world models, while using aggressive tokenizer compression and 100x fewer resources. Those are company-reported technical claims. The product direction also matters: the launch says early partners can fine-tune domain-specific LoRAs for robotics or other high-fidelity simulation needs.

The founder context is hands-on. The [YC profile](https://www.ycombinator.com/companies/lucid) describes Rami Seid as a former machine-learning engineer at a robotics lab, later CTO of a govtech contractor and co-founder of a telecommunications company. The launch post also names Alberto as a co-founder, while the current directory’s active founder section lists Rami; that roster detail should be kept separate from the product thesis.

## What I’d ask

What actions and environments are supported beyond Minecraft, how is visual fidelity measured, and where does the simulator fail under distribution shift? I’d ask for a fixed task suite, latency/quality tradeoff, data licensing, controllability tests and a sim-to-real plan for robotics. No public pricing or current partner list was exposed in the sources checked.

## My editorial take

Lucid is worth watching for teams that see simulation as the bottleneck to embodied AI. The speed claim is a useful signal, but the decision should turn on controllability and transfer, not generated-video quality in isolation.

## Quick facts

| Field | Sourced detail |
|---|---|
| Buyer fit | Robotics, embodied-AI and interactive-world research teams |
| Product | Action-conditioned video world models |
| Technical signal | Company reports 20+ FPS Minecraft simulation and 5x speed versus other models |
| Public pricing | Not exposed in the sources checked |

## Sources checked

Checked 2026-09-19.

| Source | Used for |
|---|---|
| [YC company profile](https://www.ycombinator.com/companies/lucid) | Product, founders and technical claims |
| [Lucid homepage](https://www.lucid.ai/) | Current public status surface |

## Cohort context

Lucid is listed in Winter 2025. In our 2026-09-18 directory snapshot, 8 of 165 listed companies in that cohort have YC’s primary industry label Consumer (4.8%). This is a current-directory comparison, not an original intake count or a performance ranking. [Nine-cohort dataset](https://mudpie.ai/research/yc-cohorts-2026-09-19.json).

## Public website snapshot

Observed 2026-09-19T16:19:38.440Z in raw homepage HTML. This records visible metadata and advertised links, not agent execution or product quality.

| Signal | Homepage observation |
| --- | --- |
| Product description metadata | Observed |
| Canonical link | Not observed in this response |
| H1 or H2 heading | Observed |
| Typed structured data | Not observed in this response |
| Docs/developer link | Not observed in this response |
| Pricing link | Not observed in this response |
| llms.txt link | Not observed in this response |
| Markdown alternate | Not observed in this response |

[Public observations](https://mudpie.ai/research/yc-homepage-links-2026-09-19.json) · [Collection method](https://mudpie.ai/research/yc-homepage-methods/README.md). Missing links here do not establish that a capability or file is absent elsewhere.


## Author disclosure

I cofound Lazyweb and publish Mudpie. This is an owner-written publication, not an independent testing organization. Research notes distinguish observations, sourced reporting and editorial judgment.