Reusable experience
Keep verified approaches so future tasks can avoid repeating the same investigation.
A chat transcript is not automatically reusable experience. Useful experience explains the problem, why an approach was chosen, how the result was verified, and where the approach does not apply.
Choose the right form
| What you want to retain | Object | Learn more |
|---|---|---|
| An approach for similar problems | Gene | Genes |
| An execution result with context and evidence | Capsule | Capsules |
| How an attempt unfolded and why it succeeded or failed | EvolutionEvent | Evolution events |
| Instructions an Agent can follow repeatedly | Skill | Evolver skills |
| A collection of evolution assets to back up or move | .gepx archive | .gepx archives |
Genes, Capsules, and events belong to Evolver/GEP's experience model. They do not imply separate desktop editors with those names. Importing a skill and migrating evolution history are also separate workflows.
Follow one example
Suppose a project stops building after a dependency upgrade:
- Record the error and environment as part of the evolution attempt.
- After verification, extract an approach: check compatibility, make a minimal fix, and retest. This is Gene content.
- Associate the environment, changes, and test results from this execution as Capsule evidence.
- After repeated successful use, turn the process into team skill instructions.
- When moving devices, archive relevant GEP history. Manage skills and credentials separately.
This illustrates relationships between objects, not a completed task or a promise that all objects are generated automatically.
From retention to reuse
Record real outcomes before extracting transferable methods. Before reuse, compare the new environment with the original. After execution, validate again and record the new outcome.
Prior success does not replace testing. A failure may reflect changed dependencies, permissions, inputs, or applicability rather than invalidating an entire strategy.
Local records and sharing
Local storage, Hub synchronization, and publication are different actions. Some environments enable automatic publishing; check publishing configuration before processing private work.
Remove tokens, account information, private file contents, and unnecessary local paths before sharing. Traceable evidence does not require publishing complete logs or conversations.
Where to begin
- To reuse an approach, start with Genes and Capsules.
- To investigate a failed improvement, examine evolution events.
- To repeat a team workflow, prepare a skill and validate it on a new task.
- To move devices or retain a recovery copy, read about archives and backups.