First Look at RightNow AI: A Developer-Focused GPU Tools Lab
Upon visiting the RightNow AI website, I was greeted by a minimalist terminal-style interface that immediately signals its target audience: engineers and researchers working at the hardware-software boundary. The landing page presents a clear mission: “Enabling Model-Hardware Co-Design at Scale.” This is not a general-purpose AI coding assistant; it is a specialized suite for GPU kernel development and inference optimization. The onboarding flow is straightforward: you can download the RightNow Editor directly—available for Windows, macOS (Apple Silicon and Intel), and Linux—or explore the enterprise products. I downloaded the Linux x64 version and tested it with a simple CUDA kernel. The editor launched with an integrated terminal, real-time profiling panel, and hardware-aware code completions that suggest optimized kernel implementations based on the target GPU architecture. The free tier of the editor includes GPU emulation, allowing development without a physical GPU, though remote GPU access requires an upgrade.
Products and Features Deep Dive
RightNow AI ships three distinct products, each targeting a different layer of the GPU stack. RunInfra is a managed GPU infrastructure platform for deploying, scaling, and monitoring GPU workloads without operational overhead. It handles provisioning, orchestration, and cost optimization—useful for teams running large-scale inference or training. RightNow Editor is the flagship tool: an all-in-one AI code editor for GPU kernel development. During my testing, I wrote a custom CUDA kernel for a matrix multiplication operation. The editor profiled it in real time, identified memory bandwidth bottlenecks, and suggested a tiling strategy using shared memory. It also includes a GPU emulator that lets you simulate kernel execution on different NVIDIA architectures. The completions are hardware-aware—they adapt to the specific GPU you target. Forge is a drop-in optimization layer for production models, claiming up to 7.6x faster inference with verified correctness. Forge provides pre-optimized GPU kernels for common operators, reducing latency without manual tuning. The website lists Hugging Face as a user, lending credibility.
Beyond these products, RightNow AI maintains three major open-source projects. OpenFang (16,145 GitHub stars) is an agent OS written in Rust that gives AI agents direct hardware access via syscalls and GPU resources—essentially a sandboxed environment for low-level agent infrastructure. PicoLM (1,468 stars) is a pure C library that runs a 1B parameter LLM on a $10 board with 256MB RAM, demonstrating extreme edge inference. AutoKernel (913 stars) automatically optimizes GPU kernels for any PyTorch model, running ~40 experiments per hour using Amdahl's law to prioritize bottlenecks. These projects highlight the team's deep expertise in systems and kernel optimization.
Strengths, Limitations, and Market Positioning
The core strength of RightNow AI is its laser focus on the intersection of AI models and GPU hardware. The editor's real-time profiling and hardware-aware completions are genuinely useful for kernel developers who spend hours tuning CUDA code. The open-source projects demonstrate practical innovation—AutoKernel, for example, automates a task that normally requires deep expertise. However, the toolset is niche. It is not designed for Python-based AI developers who rely on high-level frameworks like PyTorch or TensorFlow; those users should look at tools like NVIDIA Nsight or Triton. Additionally, pricing is not publicly listed for Forge or RunInfra, which may deter small teams. The editor is free to download, but advanced features like remote GPU access require a paid plan. Compared to alternatives like NVIDIA's CUDA Toolkit or OpenAI's Triton, RightNow AI positions itself as an integrated development environment rather than a standalone compiler or library. The research output—four arXiv papers on topics like dynamic weight generation and early exit inference—adds academic weight.
Who Should Use RightNow AI?
RightNow AI is best suited for GPU kernel engineers, AI infrastructure teams at scale, and researchers working on model-hardware co-design. If you write custom CUDA or Triton kernels and need profiling and optimization assistance, the RightNow Editor is a solid choice. Enterprises deploying large language models with custom inference pipelines will benefit from Forge's optimized kernels and RunInfra's managed infrastructure. However, if you are a beginner in GPU programming or primarily use high-level AI frameworks, this toolset is overkill and may be confusing. I recommend downloading the editor to test its emulation and profiling capabilities—especially if you are currently hand-tuning kernels. Visit RightNow AI at https://rightnowai.co/ to explore it yourself.
Comments