I am bullish on centralized context in agentic engineering to steer what AI produces, in a way that reduces ai slop potential across the sdl…

I am bullish on centralized context in agentic engineering to steer what AI produces, in a way that reduces ai slop potential across the sdlc. To bring in engineering context, I need integrations connected. For example, I tend to use github issues, coding rules, repo mapping + a slew of agent skills that pull that context in and execute against them. Codex + @QodoAI have these integrations supported so I know there's consistency in output quality. AI coding agents alone generate insufficient code for your codebase. The right context mitigates this. Tickets, specs and designs, and coding rules are all context sources for engineering. It should shape code before it’s written and judge it before it’s merged.

Source: Itamar Friedman (X) | 2026-08-20