June 2026 — Ongoing

One Year AI Journey

A documented year of deep exploration: building prototypes, studying the AI stack, and thinking seriously about how this technology transforms how we work and organize.

Digital Brain

As I learn more, the nodes in the digital brain grows.

My approach

Dedicated deep exploration of AI

As a sociologist, economist, and on-leave manager in the public service, I am drawn to the enormous transformational potential of AI — as a normal technology. Throughout my career, I have been involved in various projects studying the trends and impacts of technology on society and the economy from a statistical and research perspective. For the next year, I will be shifting my focus towards deep exploration of the capabilities, possibilities, and risks of this quickly evolving technology. I want to be able to not only understand the ecosystem that enables this technology, but also to be able to fluently use and architect realistic business solutions with it.

...through learning by doing

Over the next year (and probably longer!), I will be building functional prototypes, experimenting with new workflows, and thinking deeply about how AI can transform organizational strategy and operations. My work will likely fall along several buckets:

  1. building products with AI (e.g. old-school app)
  2. products with an AI integration component (e.g. LLM API integration or something more local)
  3. evolving my workflow to incorporate more agentic loops that change how I work
  4. abstracting from these experiments to imagine different organizational models and processes

...as well as study and reflection

As a non-IT professional, there is a lot for me to learn about the technical ecosystem in which AI technologies are embedded. To do this, I am diving deep into the end-to-end AI stack, including but not limited to:

  1. data and metadata management (e.g. cloud solutions, local solutions)
  2. AI product design
  3. foundational AI engineering concepts (see: Chip Huyen's "AI Engineering")
  4. responsible AI (e.g. security, ethics)
  5. others to add...

...in order to avoid hype and act more effectively

I want to find that middle-ground where I can contribute ambitious, innovative ideas and solutions that are well-thought-out and grounded in technical understanding.

My current workflow

This diagram maps how I currently experiment and develop prototypes — the autonomous vs. manual aspects of the decision-making, execution, and delivery parts of my workflow.

          %%{init: {"securityLevel": "loose", "flowchart": {"padding": 20, "nodeSpacing": 50, "rankSpacing": 70}, "theme": "dark"}}%%
          flowchart LR
          IDEA["💡 Ideas & Research"]
          DECIDE["🧠 Decide"]
          EXECUTE["🛠️ Build with AI"]
          DELIVER["🚀 Ship & Test"]

          IDEA --> DECIDE --> EXECUTE --> DELIVER
          DELIVER -.->|feedback loops back| IDEA

          linkStyle default stroke:#D9B08C,stroke-width:2px;

          class IDEA,DECIDE,EXECUTE,DELIVER brand
        
          %%{init: {"securityLevel": "loose", "flowchart": {"padding": 30, "rankSpacing": 80, "nodeSpacing": 60}, "theme": "dark"}}%%
          flowchart TD

          subgraph DECIDE["1. Decision making"]

            subgraph STREAM1[Stream A - Direct Human Prompting]
              H[Human with\nDomain Knowledge]
            end

            subgraph STREAM2[Stream B - Autonomous Idea Pipeline]
              M[(Obsidian Vault)]
              N[Daily Review Agent]
              M -->|Daily scheduled review| N
            end

            R[Research and Analysis]

            R -->|Informs and shapes| H
            R -->|Captures insights| M
            H -->|Direct prompt| O
            H ~~~ N
            N -.->|"Suggested prompt"| H
            O[Antigravity AI]
          end

          subgraph EXECUTE["2. Execution"]
            UI[System and UI Design]
            DM[Data Management]
            MCP[Model Context Protocol]

            subgraph EXISTING[Handled by Existing Systems]
              LLM[LLM APIs and Orchestration]
              SEC[Security and Optimization]
            end
          end

          UI --> LLM
          UI --> DM
          UI --> MCP
          DM --> LLM
          MCP --> LLM
          LLM --> SEC

          subgraph DELIVER["3. Delivery"]
            G[Local Testing and Debugging]
            DP{Deployment}
            I[Vercel and Netlify]
            J[Hugging Face Spaces]
            K[GitHub Codebase / CI]
            P[Portfolio Site Showcase]
            W[Web3Forms Contact]
            L[Iterative Feedback]

            G --> DP
            DP -->|Web Apps & UIs| I
            DP -->|AI Models & Python Backends| J
            DP -->|All Source Code| K
            DP -->|Curated Highlights| P
            P --> W
            I --> L
            J --> L
            K --> L
            P --> L
            W --> L
          end

          DECIDE ~~~ EXECUTE
          EXECUTE ~~~ DELIVER

          O -->|Scaffold and generate code| UI
          SEC --> G

          R <-->|"Refine direction"| L
          M <-->|"Capture new ideas"| L
          H <-->|"Direct follow-up"| L
          H <-->|"Test results"| G

          %% Interactive Tooltips and Clicks
          click MCP "javascript:void(0)" "Example Project: statcan_mcp"

          %% Classes are now styled via external CSS in style.css for Light/Dark mode support

          class R,H human
          class M obsidian
          class N agent
          class O ai
          class UI,DM,MCP execute
          class LLM,SEC existing
          class G,DP,I,J,K,P,W deliver
          class L feedback
        

My journey so far

A living log of milestones, prototypes, self-study, and reflections. I'll keep adding entries as the journey unfolds.