Maturity Active development 33 of 120 roadmap tasks complete (1 of 6 phases). Not a released product.
Editorial position
Episodic is built for production teams who remain answerable for what they publish. That answerability sets every constraint below.
Generated copy fails in ways that are cheap to produce and expensive to catch.
A model will assert a date it never read, attribute a quotation to the wrong speaker, or resolve two contradictory sources by averaging them into something neither said. None of these look like errors. They look like fluent prose, which is precisely the problem: the failure mode is confident and well formed.
Editorial teams already have a process for this. Sources get weighed, claims get checked against them, and someone signs off. What breaks when software joins the process is not the judgement - it is the evidence. If a draft arrives with no record of which sources it drew on, which it discarded, and why, the checking has to start from nothing.
Episodic treats that record as the primary artefact. The canonical document carries its own provenance, so a claim can be traced to the source that produced it without reconstructing the run.
Commitments about how the platform is built. Each says where it is already load bearing, and where it is still a design.
Every transformation stays traceable to its inputs. Source material is retained whether it was preferred or rejected, because an audit that only keeps the winners is not an audit.
Source weighting decides which material takes precedence. It does not establish that the material is correct, and the site does not describe it as though it did.
Model calls are explicit, budgeted, and interruptible. A run that cannot afford its next step stops rather than quietly spending more.
Malformed structured output, invalid timing, and stale revisions stop the run. Publishing something plausible is worse than publishing nothing.
The system may draft, compare, estimate, and flag. Deciding what reaches an audience is reserved for a person, and there is no automated path around it.
Hexagonal architecture only holds if something checks it. Import rules are a lint gate in continuous integration rather than a convention in a style guide.
The division of labour, stated as a matrix rather than a reassurance.
| Activity | The system | A person | Status |
|---|---|---|---|
| Weighting and selecting between conflicting sources | Computes priority scores from quality, freshness, and reliability, and applies the weighting matrix. | Sets the weighting coefficients per series profile, and can inspect every rejected source. | Available |
| Estimating episode runtime | Produces a deterministic estimate and records the estimator metadata behind it. | Decides whether the estimate is close enough to commission the recording. | Available |
| Scoring a draft for accuracy | Returns per-claim findings with severity and remediation guidance. | Decides whether a finding is correct and what to do about it. A score is not a verdict. | In progress |
| Spending money on model calls | Reserves budget before a billable call, commits on success, and releases on failure. | Sets the per-episode budget and the tier of model each task may use. | In progress |
| Approving an episode for an audience | Surfaces findings, suspends the graph, and preserves state while the decision is pending. | Approves, requests changes, or redirects the run. This decision has no automated path. | Planned |
The last row is the one that matters. Approval has no automated path - not a threshold, not a confidence score, not a quiet default after a timeout. A run that reaches the checkpoint waits there.
Four things the platform is not, said plainly so the rest can be believed.