Quick Answers
Fast, meeting-ready answers to the most common questions.
Quick Answers
What is this project actually?
Likely question
What did you actually build?
Short answer
We built a Cloudflare-based news and briefing system in which AI helps select and shape publishable content, but the published data is delivered through a versioned JDBIN/JDBON + R2 + Worker path.
Longer answer
The project has four practical layers:
- research corpus layer
- editorial pipeline
- storage layer
- delivery layer
The research corpus collects raw material from RSS sources. The editorial pipeline tries to turn it into publishable signals. The storage layer versions and publishes the result into R2 using the JDBIN/JDBON model. The delivery layer exposes snapshots and other public interfaces to the UI.
What to emphasize
- this is not only about a file format
- this is about a full object-storage-based read architecture
- AI is not in the public request path
What is the actual research question?
Likely question
What is the research problem here?
Short answer
The research question is not “can a new format store data,” but whether immutable object storage can function as a queryable read layer without a separate database server.
Longer answer
The precise research question is:
Can an immutable object storage object be turned into a read-optimized query layer without a separate database server using only byte-range reads, manifest/pointer versioning, and a query planner running inside a Worker?
The goal is not to build a general SQL engine. The goal is to determine when object storage itself can act as a queryable layer for read-heavy, known query paths.
What is the research innovation?
Likely question
What is actually new here?
Short answer
The novelty is not only the JDBIN format, but the way object storage is used as a queryable canonical storage layer without a separate database server.
Longer answer
The contributions together are:
- an immutable-object-storage-based queryable artifact
- a range-read query planner for known query shapes
- manifest/pointer versioning without in-place mutation of the base object
- a canonical storage + snapshot delivery split
- a deterministic read path through a Worker
What should not be claimed
- “we invented a new database”
- “this replaces SQL”
- “this is a new SQLite”
What is JDBIN?
Likely question
What exactly is JDBIN?
Short answer
JDBIN is the binary storage artifact on top of which the Worker performs deterministic range-read queries.
Longer answer
JDBIN is not a general SQL database or a database server. It is a read-optimized storage engine for immutable object storage. It typically contains segments, indexes, a string pool, and row payloads or materialized versions of them.
What to emphasize
- the canonical binary publication artifact
- the Worker reads only the required byte ranges
- optimized for known query paths
What is JDBON?
Likely question
And JDBON, why is it needed?
Short answer
JDBON is the JSON-based control layer that tells you which base, which delta chain, and which manifest is active.
Longer answer
JDBON manages:
- which base object is active
- which delta objects belong to the chain
- which database version the chain points to
- which manifest is active through the pointer
JDBIN optimizes reading. JDBON optimizes publishing, version control, rollback, and delta-chain control.
Why two layers: JDBIN + JDBON?
Likely question
Why didn’t you put everything into one format?
Short answer
Because read optimization and publication control are different responsibilities.
Longer answer
JDBIN is the data artifact that the Worker reads through range reads. JDBON manages the base version, the active manifest, the ordered delta chain, and the pointer-based publish switch. This keeps the read path tight while keeping publish, rollback, and retention clear.
What is the role of R2?
Likely question
Isn’t R2 just a bucket?
Short answer
In this model, R2 is the canonical immutable storage layer, not just a place to keep files.
Longer answer
R2 stores:
- the active base JDBIN
- delta objects
- JDBON manifests
- the active pointer
- public snapshots
Its architectural role is based on immutable objects, byte-range reads, global availability, and Worker proximity to storage.
What does the Worker do?
Likely question
Is the Worker just an API proxy?
Short answer
No. The Worker acts as planner, range reader, parser, and response builder.
Longer answer
The Worker does not simply forward requests. It resolves part of the query logic itself. The same environment can also handle triggers, snapshot updates, and publish paths.
Why Cloudflare specifically?
Likely question
Why does this work specifically on Cloudflare?
Short answer
Because Cloudflare provides the primitives the architecture needs: Worker, R2, byte-range reads, and edge execution.