First preview release of an enterprise GUI toolkit for Go: 22 interactive widgets, 3 design systems (Material 3, Fluent, Cupertino), reactive signals, zero CGO, GPU-accelerated rendering via WebGPU.
How we built a pure Go FFI library with hand-written assembly for 5 platforms, zero per-call allocations, and 88ns overhead — to power WebGPU bindings without CGO.
How we built a 6,700-line dual-execution tile rasterizer in Pure Go — Vello's 9-stage compute pipeline running on both CPU and GPU, with monoid prefix sums, DDA walking, PTCL, analytic anti-aliasing, and Euler spiral flattening.
gogpu/ui — enterprise-grade GUI toolkit for Go. Build IDEs, design tools, and desktop apps in Pure Go with GPU-accelerated rendering. No CGO, no C dependencies. 425K+ LOC ecosystem with WebGPU, shader compiler, and 2D graphics.
Karpathy says vibe coding is dead, long live agentic engineering. I wrote about this shift a month before he coined the term. Here's what his framework nails — and the critical piece it overlooks.
How we built enterprise-grade 2D/3D graphics integration without CGO, and our roadmap to a production-ready GUI toolkit
New gpucontext package enables database/sql-like GPU resource sharing. DeviceProvider, EventSource, IME support. Foundation for gogpu/ui toolkit.
122K LOC Pure Go graphics library: Canvas API, GPU acceleration, text rendering, blend modes, scene graph. No CGO. Powers IDEs and professional apps.
Linus Torvalds uses AI for Python. Anthropic writes 90% of code with AI. But there's a crucial difference between Vibe Coding and Smart Coding — and it determines whether AI makes you faster or just creates more bugs to fix.
From idea to 110K lines of Go in 8 months — Python, Rust, Go ecosystems now supported
Building a modern package manager with SAT solvers, daemon architecture, and 120x faster regex
Extracting 740 transactions from bank statements with zero errors. No CGO, no commercial license.
I built fursy, a Go HTTP router that catches API errors at compile time. Type-safe generics, 256 ns/op routing, RFC 9457 errors, OpenAPI generation - all with minimal dependencies.
Go finally has a professional graphics ecosystem. 580K+ lines of Pure Go, 5 GPU backends, 4 shader targets, GPU-accelerated 2D graphics, and a GUI toolkit with 22 widgets released.
From CPU tessellation bottleneck to GPU compute pipeline - vello-style architecture with 2,280 lines of WGSL shaders. Part of the 249K LOC Pure Go ecosystem.
gogpu/gg v0.27.1 — 155K lines of Pure Go. SDF GPU accelerator, vello-inspired rasterizer, 29 blend modes, premultiplied alpha, PDF export. Zero CGO.
How frustration with Go's GUI situation led to building 249K lines of pure Go graphics code. All 5 GPU backends complete (Vulkan, DX12, Metal, GLES, Software). All 4 shader targets (SPIR-V, MSL, GLSL, HLSL).
Why I built a database query builder with no external dependencies, and how it achieves ORM-like ergonomics without the ORM baggage.
A production-ready shader compiler in Pure Go: WGSL to SPIR-V, MSL, GLSL, HLSL. 32K LOC, zero CGO, all four major GPU backends complete.
A week after announcing coregex, Ben Hoyt integrated it into GoAWK. He found 11 bugs. Here's what happened next.
GoGPU — a Zero-CGO GPU ecosystem for Go. 380K+ lines of Pure Go. WebGPU with 5 backends (Vulkan, Metal, DX12, GLES, Software), shader compiler (4 targets), GPU-accelerated 2D graphics, GUI toolkit.
Born ML Framework releases v0.6.0 with ONNX import and lazy GPU evaluation. Training speedup from ~90s to <5s per step. Pure Go, zero CGO.
Born - a production-ready machine learning framework for Go. Pure Go, GPU acceleration, single binary deployment. No Python. No containers. Just Go.
Deep dive into building a production regex engine for Go: SIMD optimization, multi-engine architecture, and achieving 3000x speedup over stdlib on real patterns.
Hi! I've been working on a pure-Go race detector that works without CGO. Just released v0.3.0 with...