I use AI as part of a deliberate stack: coding in the editor, reasoning in the browser, models on the desktop, and automation in the background. Here is the tooling I rely on, how it fits together, and what I have shipped with it.
My main vehicle for coding and day-to-day AI-assisted work inside the repo. Cursor connects to models from major providers you enable in settings (for example Anthropic, OpenAI, and others depending on plan and region). There is no single fixed “model under the hood”; you choose what runs per task.
General-purpose reasoning: search adjacent ideas, analyze long text, compare options, and filter signal from noise before I commit to an approach in code or design.
Image and visual workflows: generation and manipulation when the brief needs fast visual iteration beyond static comps.
Running LLMs locally for privacy-sensitive experiments, offline prompts, and apps that call a local API without sending data to the cloud.
Workflow automation for chaining services, schedules, and webhooks so repetitive glue work happens without babysitting scripts.
The AI assistant I use from the terminal for agent-style runs (opencode run "…"): multi-step tasks, repo-aware edits, and automation-friendly workflows alongside the editor.
Cursor and your project files sit at the center. Around that: local models (Ollama), automation (n8n), research (ChatGPT), terminal agents (OpenCode), and visual tools (Higgsfield). Experiments like Google’s newer surfaces sit beside this loop.
Google’s agent-first development surface. I use it to explore how autonomous agents, editor, and browser loops compare to my usual Cursor + terminal workflow.
Long-form reasoning and canvas-style collaboration inside Gemini for drafts, structured breakdowns, and visual planning alongside other tools.
Google’s UI experiment tool for turning prompts into interface directions. Good for quick explorations before refining in Figma or code.
Generative AI programs on Coursera (IBM, SkillUp). Open each PDF to view the full certificate.