Codex: when an AI assistant becomes a universal work tool

A developer shares how Codex allows stepping beyond programming and automating a wide variety of work tasks. This became possible thanks to tool updates that made it a truly universal assistant.

Codex Goes Beyond Code

The author began actively using AI assistants for development long before Codex was released. They were applied in specialized environments designed for working with code: preparing changesets, editing repositories, automating fixes. But around November, things changed. The developer began experimenting with using AI assistants not just for programming, but also for other routine tasks. He created presentations in Slidev, used the tool as a digital secretary to transform voice notes into text, worked with PDF documents, spreadsheets, HTML files. Each experiment expanded the understanding of what could be delegated to AI.

New Capabilities Changed the Approach

Recent Codex updates made this expanded working mode feel natural for the first time. The tool still handles code excellently, but the most important change is different: now it provides a full-fledged workspace where a project can continue, evolve, and be refined.

• Long-lived sessions that remember project context
• Built-in memory of previous decisions and working style
• Direct access to computer actions
• Ability to refine tasks on the fly without starting from scratch
• Built-in review and control of results

Not One Feature, But a System of Interaction

In reality, the developer's behavior was influenced not by a single separate feature of Codex, but by their entire combination as a unified system. A long-lived conversation branch allows dialogue with AI to have full context and history — it's an evolving work session, not just an exchange of messages. Shared memory helps the tool remember your style, preferences, and decisions you've already made. Instead of re-explaining context, the AI already knows what's needed. Access to computer actions allows AI not just to recommend, but to actively do the work. The ability to guide the task on the fly saves time on iterations. And built-in review of results closes the loop — you immediately see what was accomplished.

What This Means

This approach points to the future of AI assistants. They stop being mere code generators — they become full-fledged workspaces where real work happens. For developers, this means the ability to delegate entire workflows, not just individual microtasks.