Case study: cross-session continuity with living memory

Use Cases

Case study: cross-session continuity with living memory

Jul 22, 2026 · 2 min

How we use SoluCortex to resume a 40-plus task deliverable across sessions, without re-explaining context to the agent.

The problem (before)

Working on large projects with AI agents has a hidden cost: every session starts from scratch. The agent does not remember why a decision was made, which convention the project follows, or where the last session left off, and the team ends up re-explaining the same context over and over. Across a roadmap of dozens of deliverables, that is lost time and real risk.

With SoluCortex (after)

SoluCortex is the project's living-memory engine: technical decisions, architecture, conventions, risks and learnings are recorded, moderated and retrievable by semantic similarity. At the start of a task the agent recalls the approved context; at close it proposes the new memories. The next session —or the next agent— starts with the context already in place.

A real example (this week)

This is not a borrowed story: we use it every day to build our own products. This very week, the SoluAI team shipped, over several sessions and days, a large body of work on SoluCortex itself: the financial learning layer, the billing subsystem (a 40-task deliverable), ES/EN internationalization of the API and emails, and more, plus reconciling 100-plus backlog tasks.

At no point did the agent need to be re-told the deploy convention, the infrastructure gotchas, or decisions already made: it was all in living memory and recalled at the start of each task. A 40-plus task deliverable was resumed across sessions without repeating context.

The numbers (real and verifiable)

From the SoluCortex project alone —dogfooding, we build it by using it— there are 170 living memories about its own development:

  • 78 technical decisions
  • 48 architecture notes
  • 13 conventions
  • 11 bug histories
  • and more (learnings, integrations, risks)

Platform-wide: 637 curated technical memories across 17 projects, in production since May 2026.

The point

Cross-session continuity —recalling context instead of re-explaining it— is the concrete saving. The same engine we sell is the one we use to ship our own products.

Newsletter

Practical ideas about AI agents, hardening and technical memory.

For software teams working with AI. No spam.

No spam. Unsubscribe any time.