When a campaign changes direction, the next agent needs the new audience decision before it writes. If that decision lives only in the founder’s memory, capable agents can still work from the wrong brief. OrgX aims to make the governing decision and its source available with the work.
This April 2026 manifesto describes that direction and the product’s limits.
Make the decision available to the next person
Consider an illustrative team changing its target customer. Product work reflects the new audience, but a later marketing session retrieves an older brief. The resulting campaign can be fluent and coherent while addressing the wrong buyer.
The repair starts before another generation pass: record the audience decision, identify its owner and source, and attach it to the initiative. The next agent should inspect that record before creating the campaign. The reviewer should compare the result with the same decision.
| Handoff question | Starting record |
|---|---|
| What are we trying to do? | Initiative and related tasks |
| Why did the audience change? | Decision, owner, and source |
| Which output should continue? | Artifact version and review |
This is the intended handoff in the example. Missing or unauthorized records still need repair.
We want fewer moments where a person has to read several agent outputs, work out which one is current, rebuild the context, and write the next briefing by hand.
That does not mean every step should become automatic. It means the system should preserve the information needed to make the next step deliberate. A saved decision can carry its reasoning forward. An artifact should identify the work it belongs to. A receipt should distinguish a reported result from the checks that actually support it.
This is what we mean by continuity. It is a design goal with observable tests, rather than a promise that a company can run unattended. Ask a fresh session to find the governing decision and its evidence. If it supplies a plausible summary without those records, the handoff has not passed.
Keep judgment attached to the output
More activity is not, by itself, progress. Creating another task or producing another artifact says little about whether the work met the request.
The person responsible for the work needs a place to accept it, reject it, or ask for a change. The agent's report should not stand in for that ruling. OrgX's authority policy prevents an agent from approving its own work or increasing its own autonomy.
Cross-client recovery, capture coverage, approval flows, and versioned quality checks remain partial or beta. Use the client setup guide for the connection flow, then test the specific workflow before relying on it. A demo of one path does not establish another.
Give the founder a smaller, clearer job
The founder should be able to set direction, define an acceptance standard, and make the tradeoffs that need human judgment. The system should make the request, current state, and supporting work easier to inspect.
That does not eliminate the founder's responsibility for strategy or turn a model's quality score into business success. It creates a better record for the next decision. If an agent followed the process but produced something the owner rejected, both facts belong in that record.
The same discipline applies to OrgX itself. Claims about complete capture, universal client parity, or realized savings require evidence beyond the architecture. We should keep those limits visible while testing the workflow on real work.
Begin with one messy initiative
Choose work whose ownership, current artifact, or next approval is hard to reconstruct. Find its existing records before creating new ones. Save one important decision with its source, identify the output it governs, and record the owner's review of that output.
Then use a fresh supported session to retrieve the same records. The MCP setup guide describes client connections, and the receipt explanation describes what an agent's account can establish.
Judge the result by whether someone else can continue responsibly without rebuilding the story from your memory.