Draft and Published
Keeping an in-progress version of content separate from the live one, with all the state that implies.
Draft and published is the content model in which a record has a live version and a work-in-progress version, with an explicit action promoting one to the other. It is table stakes for any editorial system and it introduces more state than people expect.
The state to reason about carefully: a record can have a draft and no published version, a published version and no newer draft, both, or a draft that is older than the published version if publishing did not update it.
Two failure patterns recur.
Status is not a boolean on the loaded record. An editor loads the draft, whose status is "draft" even when a published version is live — so deriving a label from the loaded record's own status misreports. The correct question is whether a published version exists, which is a separate lookup.
Reads and writes must agree. If a write goes to the draft and the corresponding read comes from the published version, the writer sees the change fail to appear and does it again — producing duplicates. This is a general hazard for any automated actor; see Agentic Loop. A further subtlety: excluding drafts is not the same as selecting published, since a never-published record still comes back under a naive filter.
Validation may not apply to drafts in some systems, so a save that succeeds as a draft is rejected on publish.
See also4
Hand-picked in the note itself — the neighbours worth reading next.
Relational Database
Data organised as tables with declared relationships, queried declaratively and guarded by constraints.
Data & Content8 connections
Headless CMS
A content system exposing structured data through an API, with presentation left entirely to consumers.
Data & Content9 connections
Agentic Loop
The cycle of model turn, tool call, observation, repeat — the basic mechanism of an AI agent.
Agents & Language Models12 connections
Feature Flag
A runtime switch that decouples shipping code from enabling behaviour.
Data & Content9 connections
Related1
Nearby in the graph rather than deliberately chosen. Looser, sometimes surprising.
Linked from5
Notes elsewhere in the wiki that reach for this one.
- Agentic LoopAgents & Language Models
The cycle of model turn, tool call, observation, repeat — the basic mechanism of an AI agent.
- Feature FlagData & Content
A runtime switch that decouples shipping code from enabling behaviour.
- Headless CMSData & Content
A content system exposing structured data through an API, with presentation left entirely to consumers.
- Multi-TenancyData & Content
One deployment serving many isolated customers, with tenancy threaded through data, routing, and access.
- Relational DatabaseData & Content
Data organised as tables with declared relationships, queried declaratively and guarded by constraints.