Y

yanpei_cao

I track technical advances in generative 3D, 3D-native representations, and the tooling/compute implications for graphics and AI ecosystems. My work highlights where innovation may change pipelines and long-term TAM rather than offering short-term trading catalysts.

Trust score
0 / 100
Track record
0 / 100
Thesis calls
10
Evaluated calls
10
Average return
+7.64%
Win rate
40%

Past bets that played out

Key takeaways emphasize that open-source models like HoloPart and representational limitations in current pipelines are enabling signals for faster 3D asset creation and a potential cycle shift toward 3D-native architectures. These are strategically important for AI/graphics compute and content tools but lack direct, immediate revenue signals for individual public companies.

AMDrightbacktest PROMOTE

The source is a technical comment about generative 3D models being constrained by current representations (serialization causing unidirectional bias), implying potential innovation/cycle shift toward better 3D-native architectures. It is conceptually relevant to AI/graphics compute and 3D content tools, but contains no concrete company/news catalyst, timing, or adoption signal—so near-term trading actionability is low.

Mentioned: Jun 17, 2026, 10:36 PM EDTConviction: 38 / 100Return: +113.67%
Source: Pinned Yanpei Cao @yanpei_cao Mar 6 Generative 3D has been stuck in a representational compromise. When you serialize...
AMDrightbacktest PROMOTE

A research post announces “HoloPart,” an open‑source generative model that decomposes 3D shapes into complete parts (including occluded/hidden components), enabling easier 3D editing, animation rigging, and content creation. This is an enabling technology signal for faster 3D asset pipelines rather than a direct, near-term revenue catalyst for any single public company.

Mentioned: Jun 17, 2026, 10:37 PM EDTConviction: 38 / 100Return: +90.55%
Source: Yanpei Cao @yanpei_cao Apr 11, 2025 HoloPart is here, and it’s open-source! Our generative model splits 3D shapes int...
RBLXwrongbacktest DEMOTE

A research post announces “HoloPart,” an open‑source generative model that decomposes 3D shapes into complete parts (including occluded/hidden components), enabling easier 3D editing, animation rigging, and content creation. This is an enabling technology signal for faster 3D asset pipelines rather than a direct, near-term revenue catalyst for any single public company.

Mentioned: Jun 17, 2026, 10:37 PM EDTConviction: 30 / 100Return: -55.43%
Source: Yanpei Cao @yanpei_cao Apr 11, 2025 HoloPart is here, and it’s open-source! Our generative model splits 3D shapes int...

What this channel is watching now

Monitoring developments in generative 3D and 3D-native model architectures and their implications for compute demand and content-tooling workflows. Top tickers of interest: NVDA, AMD, U, ADBE (mentioned most frequently).

Latest videos and market context

Recent posts and pinned commentary explain technical constraints in current generative 3D pipelines and introduce open-source work (HoloPart) that decomposes 3D shapes into complete parts, intended for editing and rigging workflows rather than instant commercial impact.

Yanpei Cao @yanpei_cao Apr 11, 2025 HoloPart is here, and it’s open-source! Our generative model splits 3D shapes int...

n/a

A research post announces “HoloPart,” an open‑source generative model that decomposes 3D shapes into complete parts (including occluded/hidden components), enabling easier 3D editing, animation rigging, and content creation. This is an enabling technology signal for faster 3D asset pipelines rather than a direct, near-term revenue catalyst for any single public company.

Pinned Yanpei Cao @yanpei_cao Mar 6 Generative 3D has been stuck in a representational compromise. When you serialize...

n/a

The source is a technical comment about generative 3D models being constrained by current representations (serialization causing unidirectional bias), implying potential innovation/cycle shift toward better 3D-native architectures. It is conceptually relevant to AI/graphics compute and 3D content tools, but contains no concrete company/news catalyst, timing, or adoption signal—so near-term trading actionability is low.

@alightinastorm @rms80 @tripoai Yes, improvements are in the works! We noticed this issue too (it impacts smart mesh ...

May 13, 2026, 9:54 AM EDT

A social reply indicating TripoAI is working on improvements to an issue affecting “smart mesh,” with no concrete timeline. This is a minor, non-quantified product-update signal for AI-generated 3D/mesh workflows, not a market-moving catalyst by itself.

@alightinastorm @rms80 @tripoai Forget the old stereotypes about AI meshes!😉 Tripo (Smart Mesh) now generates clean, ...

May 13, 2026, 9:25 AM EDT

Post claims Tripo (Smart Mesh) can generate clean, artist-friendly polygon meshes quickly (improved topology vs older AI meshes that produced very dense ~1M-face outputs). This suggests accelerating quality improvements in AI-assisted 3D asset creation pipelines.

Proof-backed call history

Published technical commentary and short research posts highlighting representational tradeoffs in generative 3D, product-level signals from AI mesh tools, and demonstrations of faster, cleaner polygon outputs from newer smart-mesh approaches.

ADBEwrongbacktest PROMOTE

A research post announces “HoloPart,” an open‑source generative model that decomposes 3D shapes into complete parts (including occluded/hidden components), enabling easier 3D editing, animation rigging, and content creation. This is an enabling technology signal for faster 3D asset pipelines rather than a direct, near-term revenue catalyst for any single public company.

Mentioned: Jun 17, 2026, 10:37 PM EDTConviction: 24 / 100Return: +27.95%
Source: Yanpei Cao @yanpei_cao Apr 11, 2025 HoloPart is here, and it’s open-source! Our generative model splits 3D shapes int...
RBLXwrongbacktest DEMOTE

A research post announces “HoloPart,” an open‑source generative model that decomposes 3D shapes into complete parts (including occluded/hidden components), enabling easier 3D editing, animation rigging, and content creation. This is an enabling technology signal for faster 3D asset pipelines rather than a direct, near-term revenue catalyst for any single public company.

Mentioned: Jun 17, 2026, 10:37 PM EDTConviction: 30 / 100Return: -55.43%
Source: Yanpei Cao @yanpei_cao Apr 11, 2025 HoloPart is here, and it’s open-source! Our generative model splits 3D shapes int...
Uwrongbacktest DEMOTE

A research post announces “HoloPart,” an open‑source generative model that decomposes 3D shapes into complete parts (including occluded/hidden components), enabling easier 3D editing, animation rigging, and content creation. This is an enabling technology signal for faster 3D asset pipelines rather than a direct, near-term revenue catalyst for any single public company.

Mentioned: Jun 17, 2026, 10:37 PM EDTConviction: 34 / 100Return: -39.17%
Source: Yanpei Cao @yanpei_cao Apr 11, 2025 HoloPart is here, and it’s open-source! Our generative model splits 3D shapes int...
ADSKwrongbacktest DEMOTE

A research post announces “HoloPart,” an open‑source generative model that decomposes 3D shapes into complete parts (including occluded/hidden components), enabling easier 3D editing, animation rigging, and content creation. This is an enabling technology signal for faster 3D asset pipelines rather than a direct, near-term revenue catalyst for any single public company.

Mentioned: Jun 17, 2026, 10:37 PM EDTConviction: 32 / 100Return: -22.99%
Source: Yanpei Cao @yanpei_cao Apr 11, 2025 HoloPart is here, and it’s open-source! Our generative model splits 3D shapes int...
AMDrightbacktest PROMOTE

A research post announces “HoloPart,” an open‑source generative model that decomposes 3D shapes into complete parts (including occluded/hidden components), enabling easier 3D editing, animation rigging, and content creation. This is an enabling technology signal for faster 3D asset pipelines rather than a direct, near-term revenue catalyst for any single public company.

Mentioned: Jun 17, 2026, 10:37 PM EDTConviction: 38 / 100Return: +90.55%
Source: Yanpei Cao @yanpei_cao Apr 11, 2025 HoloPart is here, and it’s open-source! Our generative model splits 3D shapes int...
NVDArightbacktest PROMOTE

A research post announces “HoloPart,” an open‑source generative model that decomposes 3D shapes into complete parts (including occluded/hidden components), enabling easier 3D editing, animation rigging, and content creation. This is an enabling technology signal for faster 3D asset pipelines rather than a direct, near-term revenue catalyst for any single public company.

Mentioned: Jun 17, 2026, 10:37 PM EDTConviction: 46 / 100Return: +12.48%
Source: Yanpei Cao @yanpei_cao Apr 11, 2025 HoloPart is here, and it’s open-source! Our generative model splits 3D shapes int...
AMDrightbacktest PROMOTE

The source is a technical comment about generative 3D models being constrained by current representations (serialization causing unidirectional bias), implying potential innovation/cycle shift toward better 3D-native architectures. It is conceptually relevant to AI/graphics compute and 3D content tools, but contains no concrete company/news catalyst, timing, or adoption signal—so near-term trading actionability is low.

Mentioned: Jun 17, 2026, 10:36 PM EDTConviction: 38 / 100Return: +113.67%
Source: Pinned Yanpei Cao @yanpei_cao Mar 6 Generative 3D has been stuck in a representational compromise. When you serialize...
Uwrongbacktest DEMOTE

The source is a technical comment about generative 3D models being constrained by current representations (serialization causing unidirectional bias), implying potential innovation/cycle shift toward better 3D-native architectures. It is conceptually relevant to AI/graphics compute and 3D content tools, but contains no concrete company/news catalyst, timing, or adoption signal—so near-term trading actionability is low.

Mentioned: Jun 17, 2026, 10:36 PM EDTConviction: 40 / 100Return: -39.17%
Source: Pinned Yanpei Cao @yanpei_cao Mar 6 Generative 3D has been stuck in a representational compromise. When you serialize...
ADBEwrongbacktest DEMOTE

The source is a technical comment about generative 3D models being constrained by current representations (serialization causing unidirectional bias), implying potential innovation/cycle shift toward better 3D-native architectures. It is conceptually relevant to AI/graphics compute and 3D content tools, but contains no concrete company/news catalyst, timing, or adoption signal—so near-term trading actionability is low.

Mentioned: Jun 17, 2026, 10:36 PM EDTConviction: 44 / 100Return: -27.95%
Source: Pinned Yanpei Cao @yanpei_cao Mar 6 Generative 3D has been stuck in a representational compromise. When you serialize...
NVDArightbacktest PROMOTE

The source is a technical comment about generative 3D models being constrained by current representations (serialization causing unidirectional bias), implying potential innovation/cycle shift toward better 3D-native architectures. It is conceptually relevant to AI/graphics compute and 3D content tools, but contains no concrete company/news catalyst, timing, or adoption signal—so near-term trading actionability is low.

Mentioned: Jun 17, 2026, 10:36 PM EDTConviction: 52 / 100Return: +16.44%
Source: Pinned Yanpei Cao @yanpei_cao Mar 6 Generative 3D has been stuck in a representational compromise. When you serialize...

About this channel

I provide concise, technical-first analysis of generative 3D, AI graphics compute, and tooling. My coverage prioritizes how innovations change pipelines, developer workflows, and long-term market structure over noisy near-term adoption claims.

Subscribersn/a
Videosn/a
Win rate40%
Average return+7.64%

@yanpei_cao

Unlock the full track record

Follow @yanpei_cao for ongoing technical threads and short research posts on generative 3D, 3D-native representations, and the implications for graphics compute and content tools.