The Codex Experience
A living record of my agentic coding journey—configurations, experiments, and lessons learned while automating research workflows.
- Published 14 Jan 2026
- Focus: Agentic Workflow Hub
Episodes in this series
Each entry shares a blueprint, artifacts, and lessons learned from building an AI-assisted workflow.
Series ongoing
Codex Configuration Log
Tracking specific tweaks in config.toml, including enabling web search and managing agent features.
- Updated 14 Jan 2026
- Focus: Configuration Documentation
Beyond the Usage Bar: Estimating Codex Tokens with ccusage
When OpenAI's subscription dashboard only shows a percentage, here's a CLI-based approach to approximate daily and monthly tokens and spend.
- Published 11 Jan 2026
- Focus: Usage estimation + ccusage
- Tool: ccusage CLI
Auto Calendar: Converting Emails into Outlook Schedules
Turning sprawling project update emails into a shared Outlook calendar with prompt-classified summaries, validation layers, and Microsoft Graph automations.
- Published 10 Oct 2025
- Focus: Email-to-calendar orchestration
- Stack: Azure Logic Apps + Microsoft Graph
How I Switched from GitHub Copilot to ChatGPT Pro + Codex
Swapping Copilot for Codex, documenting AGENTS.md rituals, and turning TODO.md into the command centre for agentic coding.
- Published 9 Oct 2025
- Focus: Agentic collaboration
- Stack: ChatGPT Pro + Codex CLI
How I Built an AI Calendar Generator
Inside the scheduling assistant experiment coordinating fieldwork—architecture notes, automation safeguards, and adoption checklist.
- Published 7 Oct 2025
- Focus: Workflow orchestration
- Stack: LLM orchestration + calendar APIs
AI-first meeting notes assistant
Placeholder entry outlining the planned teardown of auto-summarising collaboration rituals.
DraftingPromptable syllabus designer
Placeholder entry describing how lesson planning templates evolve with structured prompting.
ResearchingSeries roadmap
A living backlog of experiments I am prioritising as teaching terms, research sprints, and team needs shift.
- Stability guardrails: Placeholder text about evaluating when to pair deterministic scheduling logic with LLM suggestions.
- Data hygiene: Placeholder text about privacy reviews, consent artifacts, and secure storage pipelines.
- People workflow: Placeholder text about onboarding collaborators and communicating automation boundaries.