Uploaded July 2026 | Updated September 2026, 15 minutes ago
GPU acceleration is no longer “just for graphics,” and WebAssembly is stepping into a pivotal moment where portable compute means portable capabilities, not just portable code. In this episode of WebAssembly Unleashed, Joel Moses and Matt Yacobucci sit down with Mendy Berger from Cosmonic to unpack the latest shift in the WASI graphics ecosystem: why WASI WebGPU is being treated as the stable foundation, while higher-level display and surface work is moving into the more experimental wasi-gfx track.
Mendy explains the core rationale behind the split. WebGPU, though new to many developers, has years of design maturity and broad vendor input behind it, making it a strong candidate for a stable interface. By contrast, “surface” and presentation APIs outside the browser don’t have the same level of polish yet, and need room to evolve quickly. The separation allows WebGPU to advance without being held back by the still-emerging display story, and it also supports non-graphics use cases where GPU compute matters without ever rendering a pixel.
The conversation covers how WebGPU in WASI fits into the component model, why the WIT interface is so large, and what breaks when you push WIT into GPU-style corner cases like bitwise flags. They also dig into AI inference as a killer use case, including early work porting projects like llama.cpp and ONNX Runtime, and where WebGPU is a better fit than alternatives like WASI-NN or WebNN.
Finally, Mendy shares where the specs stand today, what’s happening with Preview 3 and async, and how people can get involved through the WASI WebGPU repo, the wasi-gfx runtime work, and the weekly community meetings. If you care about GPUs, graphics, or bringing AI closer to where code runs, this episode lays out what’s changing and why it matters.
Chapters:
00:00 Welcome to WebAssembly Unleashed
00:27 Wasm, Graphics UI, and AI
01:15 Meet Mendy Berger: wasi-gfx + Renderlet + Cosmonic
01:58 Why split wasi-gfx from WebGPU: Stability vs experimentation
03:46 How wasi-gfx and WebGPU overlap (and where they don’t)
04:25 What WebGPU solves: “Wasm for GPUs” + isolation
05:52 Use cases: graphics, AI inference, parallel compute
07:49 WASI component model fit: 1000+ lines of WIT + resources/flags
10:57 Killer demos: Llama.cpp + ONNX Runtime watermarking
12:50 WebNN vs WebGPU vs WASI-NN: What runs where
15:39 OpCodes: Wasm, WebGPU, or WebNN
17:26 Is WASI WebGPU the answer to controlling LLM token generation cost?
19:17 Are there use cases where WASI WebGPU isn't ideal?
20:07 Where we are: WebGPU in Phase 2, P3/async, implementations
22:41 Conformance test results: Can we use this now?
24:25 How to get involved with wasi-gfx and WASI WebGPU
26:38 The future of WebAssembly
Read The future of wasi-gfx and wasi:webgpu by Sean Isom & Mendy Berger: https://wasi-gfx.dev/blog/posts/future-of-wasi-gfx/
Get involved with WASI WebGPU: github.com/WebAssembly/wasi-webgpu
Get involved with wasi-gfx: github.com/wasi-gfx
For more from F5's Office of the CTO visit the following sites:
Blogs - f5.com/company/octo
Reports - f5.com/services/resources/reports
Meet Your Hosts:
Joel Moses | LinkedIn | @Joel_Moses
Oscar Spencer | X | LinkedIn
Matthew Yacobucci | LinkedIn
⬇️⬇️⬇️ JOIN THE COMMUNITY! ⬇️⬇️⬇️
DevCentral is an online community of technical peers dedicated to learning, exchanging ideas, and solving problems - together.
Find all our platform links ⬇️ and follow our Community Evangelists! 👋
➡️ DEVCENTRAL: community.f5.com
➡️ YOUTUBE: youtube.com/devcentral
➡️ LINKEDIN: linkedin.com/showcase/f5-devcentral
➡️ TWITTER: twitter.com/devcentral
Your Community Evangelists:
👋 Jason Rahm: linkedin.com/in/jrahm | twitter.com/jasonrahm
👋 Buu Lam: linkedin.com/in/buulam | twitter.com/buulam
👋 Chase Abbott: linkedin.com/in/chaseabbott1
GPU acceleration is no longer “just for graphics,” and WebAssembly is stepping into a pivotal moment where portable compute means portable capabilities, not just portable code. In this episode of WebAssembly Unleashed, Joel Moses and Matt Yacobucci sit down with Mendy Berger from Cosmonic to unpack the latest shift in the WASI graphics ecosystem: why WASI WebGPU is being treated as the stable foundation, while higher-level display and surface work is moving into the more experimental wasi-gfx track.
Mendy explains the core rationale behind the split. WebGPU, though new to many developers, has years of design maturity and broad vendor input behind it, making it a strong candidate for a stable interface. By contrast, “surface” and presentation APIs outside the browser don’t have the same level of polish yet, and need room to evolve quickly. The separation allows WebGPU to advance without being held back by the still-emerging display story, and it also supports non-graphics use cases where GPU compute matters without ever rendering a pixel.
The conversation covers how WebGPU in WASI fits into the component model, why the WIT interface is so large, and what breaks when you push WIT into GPU-style corner cases like bitwise flags. They also dig into AI inference as a killer use case, including early work porting projects like llama.cpp and ONNX Runtime, and where WebGPU is a better fit than alternatives like WASI-NN or WebNN.
Finally, Mendy shares where the specs stand today, what’s happening with Preview 3 and async, and how people can get involved through the WASI WebGPU repo, the wasi-gfx runtime work, and the weekly community meetings. If you care about GPUs, graphics, or bringing AI closer to where code runs, this episode lays out what’s changing and why it matters.
Chapters:
00:00 Welcome to WebAssembly Unleashed
00:27 Wasm, Graphics UI, and AI
01:15 Meet Mendy Berger: wasi-gfx + Renderlet + Cosmonic
01:58 Why split wasi-gfx from WebGPU: Stability vs experimentation
03:46 How wasi-gfx and WebGPU overlap (and where they don’t)
04:25 What WebGPU solves: “Wasm for GPUs” + isolation
05:52 Use cases: graphics, AI inference, parallel compute
07:49 WASI component model fit: 1000+ lines of WIT + resources/flags
10:57 Killer demos: Llama.cpp + ONNX Runtime watermarking
12:50 WebNN vs WebGPU vs WASI-NN: What runs where
15:39 OpCodes: Wasm, WebGPU, or WebNN
17:26 Is WASI WebGPU the answer to controlling LLM token generation cost?
19:17 Are there use cases where WASI WebGPU isn't ideal?
20:07 Where we are: WebGPU in Phase 2, P3/async, implementations
22:41 Conformance test results: Can we use this now?
24:25 How to get involved with wasi-gfx and WASI WebGPU
26:38 The future of WebAssembly
Read The future of wasi-gfx and wasi:webgpu by Sean Isom & Mendy Berger: https://wasi-gfx.dev/blog/posts/future-of-wasi-gfx/
Get involved with WASI WebGPU: github.com/WebAssembly/wasi-webgpu
Get involved with wasi-gfx: github.com/wasi-gfx
For more from F5's Office of the CTO visit the following sites:
Blogs - f5.com/company/octo
Reports - f5.com/services/resources/reports
Meet Your Hosts:
Joel Moses | LinkedIn | @Joel_Moses
Oscar Spencer | X | LinkedIn
Matthew Yacobucci | LinkedIn
⬇️⬇️⬇️ JOIN THE COMMUNITY! ⬇️⬇️⬇️
DevCentral is an online community of technical peers dedicated to learning, exchanging ideas, and solving problems - together.
Find all our platform links ⬇️ and follow our Community Evangelists! 👋
➡️ DEVCENTRAL: community.f5.com
➡️ YOUTUBE: youtube.com/devcentral
➡️ LINKEDIN: linkedin.com/showcase/f5-devcentral
➡️ TWITTER: twitter.com/devcentral
Your Community Evangelists:
👋 Jason Rahm: linkedin.com/in/jrahm | twitter.com/jasonrahm
👋 Buu Lam: linkedin.com/in/buulam | twitter.com/buulam
👋 Chase Abbott: linkedin.com/in/chaseabbott1
