Agentic Coding and Writing Tools — A Direction Worth Watching

I’ve been thinking about agentic coding lately — not just as a development practice, but as a way to think about product evolution. If AI agents can autonomously write, test, and deploy code, what could they do for a writing tool?

WriteDaily has been a traditional web application for 12 years: human-written code, human-designed features, human-maintained infrastructure. The idea of an agent contributing to its development — or even to the writing experience itself — is new territory.

What Agentic Coding Could Mean for WriteDaily

I’m exploring three conceptual directions:

1. Agent-Assisted Development

Could an AI agent handle routine WriteDaily maintenance? Dependency updates, security patches, test coverage expansion — the kind of work that keeps a side project healthy but doesn’t require creative decision-making.

I’ve experimented with coding agents on other projects, and the pattern is promising: the agent proposes changes, runs tests, and presents a diff for review. For WriteDaily’s maintenance workload, this could reduce the monthly operational burden from a few hours to a few minutes of review time.

2. Agent-Powered Writing Analysis

The AI writing feedback prototype I’ve been developing uses a static model for pattern recognition. An agentic approach could go further: dynamically querying your writing history, formulating specific observations, and presenting them as a conversation rather than a report.

Imagine asking: “How has my writing changed since last year?” and having an agent compile the answer from your entry history, LIWC data, and stylistic metrics — all running locally, all respecting privacy.

3. Agentic Writing Companions

This is the most speculative direction. An agent that doesn’t just analyse your writing but engages with it — asking follow-up questions, suggesting directions, acting as a thoughtful reader rather than a grammar checker.

I’m not building this. But I’m watching the space. Tools like Claude’s artifacts and ChatGPT’s canvas hint at a future where AI is a collaborator, not just a tool. A daily writing practice with an AI companion that remembers everything you’ve written and helps you explore your own patterns — that’s compelling.

What I’m Not Doing

I’m not announcing an agentic feature for WriteDaily — this post sits squarely in the same speculative space as my AI experiments and LLM analytics exploration. The technology is too new, the reliability challenges are real, and the privacy implications of autonomous agents accessing personal writing history need careful consideration.

But I’m paying attention. Agentic coding is changing how I build software. It may eventually change how I think about the software I’ve already built.

For now, writedaily.co remains what it’s always been: a private, reliable tool for daily writing. The agentic future is something I’m exploring, not something I’m shipping. Yet.