julián medina

Digital DNA

How to maintain context when working with AI on long projects

By Julián Medina · Director of SourcingUp and creator of CommerceUp

Keep the goal, current decisions, completed work and next step in a record you can update. Continuity improves when you can retrieve the context that matters without rebuilding the entire conversation.

Record the current state of the project

A useful project summary explains the goal and what remains before it is complete. Include constraints, relevant files and decisions that should only be reopened with new evidence. Keep the context that guides the next piece of work, rather than copying the whole conversation.

Separate decisions from ideas still being explored

During a project ideas appear that are later discarded. If they are mixed with current instructions, the AI ​​can return to abandoned paths. Mark what was decided, why, and what condition would justify reviewing it. Keep the alternatives when they explain a criterion, without presenting them as pending tasks.

Check before continuing

When you return to a project, ask the AI to identify its current state and check the relevant references. A file may have changed since the last summary. Update the record after a meaningful step forward. This does not replace checking the results, but it reduces the need to remember every detail or explain the project from scratch.

Put it into practice

The thinking behind this guide

A guide by Julián Medina. The scenarios are illustrative, not measured customer outcomes. More ideas on Medium.