Figure Project Go-Big vs EgoVerse

Project Go-Big is a humanoid pretraining data-collection initiative that Figure announced in a company blog post on September 18, 2025. Figure says it is building what it calls the world’s largest and most diverse humanoid pretraining dataset for Helix, accelerated by a partnership with Brookfield. EgoVerse is described in an arXiv paper titled “EgoVerse: An Egocentric Human Dataset for Robot Learning from Around the World”.

Both put egocentric human data at the centre of robot learning. The difference is what each source shows. Project Go-Big reports one navigation result and no dataset size. EgoVerse is a dataset, a collaborative platform and a controlled study of how human data transfers to robots. Neither is a Firsthand product, and Firsthand has no relationship with Figure, Brookfield or the EgoVerse authors. For the single-source explainers, see /guides/what-is-figure-project-go-big and /guides/what-is-egoverse.

Figure states that Project Go-Big is accelerated by a partnership with Brookfield, which it describes as owning over 100,000 residential units worldwide, alongside 500 million square feet of commercial office space and 160 million square feet of logistics space. It says data collection has already begun in Brookfield environments and will continue to scale.

The post reports one result. Figure trained Helix on 100% egocentric human video, collected passively as people do behaviors in real Brookfield homes, with no robot demonstrations, and Helix now responds to commands such as “Walk to the kitchen table” by generating closed-loop navigation control from pixels. Figure says that, to its knowledge, this is the first time a humanoid has learned end-to-end from images and language to low-level SE(2) velocity commands using only human video.

The post does not state an hour count, an episode count, an action-label method, a licence or a benchmark number, so this page does not either.

The abstract on the paper’s arXiv page says robot learning increasingly depends on large and diverse data, but robot data collection remains expensive and difficult to scale. Egocentric human data are a promising alternative, yet existing human datasets are often limited in scope, difficult to extend and fragmented across institutions. EgoVerse is presented as a collaborative platform that unifies data collection, processing and access, with contributions from individual researchers, academic labs and industry partners.

The current release includes 1,362 hours (80k episodes) of human demonstrations spanning 1,965 tasks, 240 scenes and 2,087 unique demonstrators, with standardized formats, manipulation-relevant annotations and tooling for downstream learning. Beyond the dataset, the authors run a large-scale study of human-to-robot transfer, with experiments replicated across multiple labs, tasks and robot embodiments under shared protocols.

This page does not state an EgoVerse licence, because the paper text reviewed on this site does not state one. For release terms, read the paper and its linked project page directly.

They support different claims. Project Go-Big shows that a navigation capability can be learned from human video with no robot data at all, but it runs no robot-data arm, so it cannot say how much human data helps or what it replaces. EgoVerse tests transfer directly: it reports that policy performance generally improves with increased human data.

EgoVerse also adds a condition. The abstract says effective scaling depends on alignment between human data and robot learning objectives. That is a more cautious reading than “human video is enough”, and it is the reading this site uses. The trade-off between scalable human video and embodiment-aligned robot data is laid out in /compare/human-video-vs-robot-data-for-pretraining.

The site’s /compare/egoverse-vs-gen-0 page records the study’s details as the paper states them. Adding EgoVerse-A data improves in-domain and out-of-domain performance “by up to 30%”, with one exception. Neither 8 hours of diverse data nor aligned human data alone drives significant gains, but 2 hours of aligned data enables transfer from 2 hours of diverse data, and the trend continues at 8 hours.

For a buyer, that makes alignment a requirement to plan for, not a bonus. Project Go-Big’s post does not describe how its human video is aligned to Helix beyond noting that a humanoid’s perspective and kinematics mirror a human’s. How much aligned data a given robot needs is set per program, not by either source.

No. Figure’s result is zero-shot, language-commanded navigation of homes from human video alone, output as low-level SE(2) velocity commands. EgoVerse’s result is a study finding about how robot policy performance changes as human data is added, across several labs, tasks and embodiments. Different tasks, different measurements and different evaluation setups mean the two answer different questions.

What they share is direction, not size. Reading Project Go-Big as confirming EgoVerse’s scaling finding, or EgoVerse as confirming Figure’s navigation claim, goes beyond what either source states.

They do not compare directly. EgoVerse states 1,362 hours across 80k episodes, 1,965 tasks, 240 scenes and 2,087 demonstrators. Project Go-Big states property access, which describes how many real spaces a partnership opens up, not how much recorded data exists. Figure says only that collection has begun and will continue to scale.

The shared point is diversity. Figure frames diversity of real-world spaces as its goal, and EgoVerse reports breadth across tasks, scenes and demonstrators. A pretraining plan should start from your robot’s coverage gaps rather than a headline number. See /guides/why-dataset-diversity-matters-in-robot-learning.

EgoVerse states manipulation-relevant annotations in its abstract. The site’s /compare/egoverse-vs-gen-0 page records the detail from the paper: per-frame 3D hand pose with 21 keypoints per hand and calibrated 6-DoF head pose, with 1–2 second language descriptions added in EgoVerse-I. Project Go-Big’s post does not state how its video is turned into training signal.

For a buyer, label quality is part of the dataset, not an add-on. This site documents what action supervision for human video looks like, wrist-level pose plus articulated hand pose, in /guides/hand-pose-and-wrist-action-labels-for-pretraining. What a given program ships is set in its skill spec.

Project Go-Big is Figure’s own initiative for Figure’s own model, built on a property partnership specific to Figure and Brookfield, and the post describes no way for outside teams to obtain its data. EgoVerse is built for sharing, and EgoDB provides “controlled data access”, but the paper text reviewed does not state a licence or public release terms.

A team that plans to train a commercial model on either needs to confirm terms directly with the publisher, in writing, before use. A fixed public corpus also cannot be reshaped afterwards to your target tasks, embodiment or coverage gaps.

A team that wants human data scoped to its own robot, tasks and environments has to source it separately from both. A custom collection starts with a written skill spec covering the scenes, tasks, objects and people a corpus is missing, proven with a fast, on-demand Spec Pilot before volume scales. The pilot process is described in /custom-collection/egocentric-data-collection-pilot.

Capture is head-worn, bimanual and whole-body for humanoid pretraining, delivered under a perpetual, buyer-owned commercial licence with a documented consent chain, in RLDS, LeRobot, WebDataset, HDF5, zarr and Rerun .rrd. See /custom-collection/humanoid-pretraining-data and /custom-collection/pretraining-data-collection.

Ask how the human capture is aligned to your robot, whether the viewpoint matches, how clips are labeled, whether sensor streams are synchronized, how coverage was planned against a spec, and whether the licence permits commercial training with documented consent. Also ask how much aligned data your embodiment needs. Neither source answers those for outside buyers. See /guides/robot-pretraining-data-diversity-plan.