Plain Markdown. Real files.
Your original documents and new pages live in a local folder. Export a whole book, including its questions, notes, and revision history.
A little curiosity. A deeper understanding.Finally understand
Finally understand
what AI wrote.
Your AI wrote the plan.
Make the understanding yours.
Open a document. Question any passage. Follow the explanation as far as you need.
Find your first question A local reader, powered by GPT-6 Astra.Original document 4 min read
A plan for
background jobs
Our product turns customer interviews into searchable research. A long interview can hold a request open for several minutes.
How it would work
The upload endpoint saves the file, creates a job, and returns immediately. A worker picks up the job and updates its progress.
Jobs should be idempotent: processing the same job more than once must not create duplicate results.
Ask about this
The question behind the plan
A queue changes when work happens. It doesn’t remove the need to make that work safe to repeat.
Bring your own MarkdownKeep every question connectedOwn the files, always
01 / From output to understandingThe next useful layer
The next useful layer
starts with your question.
You don’t need another wall of text.
You need a way into the one you already have.
01
Bring the work you’re responsible for.
An implementation plan. A research brief. An architecture proposal. Open your files or paste what your agent just wrote.
02
Stop at the sentence that stops you.
Select a passage and ask in your own words. A quick explanation stays beside the text, right where the question began.
03
Follow the part you want to understand.
Open a worked example. Compare an alternative. Investigate a failure case. Each new page stays connected to its source.
A personal book, one question at a time
The original planMove processing to a queue
.mdQuick answerWhat does idempotent mean?
A worked exampleWhat if the job runs twice?
Research a tradeoffDo we need this complexity yet?
✳Your curiosity sets the direction.
02 / A clearer view of the reasoningA confident answer
A confident answer
isn’t the same
as good evidence.
Nested keeps the original material, generated explanations, and your own conclusions distinct. See the source passage. Notice the assumption. Leave the unanswered question open.
Examine the sample planBefore we approve
Source passage
“We expect fewer than 500 uploads per day at launch.”
A plan for background jobsStill needs evidence
This is a planning assumption. The supplied documents don’t include measured demand.
Your conclusion
Run a load test with representative interviews before choosing the queue.
Illustrative review note03 / Understanding that stays with youThe thinking is yours.
The thinking is yours.
The files are, too.
Pick up the thread.
Return to a branch, revisit an open question, or build a review brief for your agent. When a source changes, see what needs another look.
Read first. Connect when you’re ready.
Reading, editing, and notes work locally. Connect your own OpenAI API key when you want GPT-6 Astra to help you go deeper.
A few useful answersBefore you open
Before you open
the next layer.
What can I open in Nested?
Markdown and plain-text files, folders of documents, or a pasted response. A sample architecture review is included so you can explore the reader immediately.
Is this another chat app?
Nested starts with your document. Questions attach to passages, explanations become connected pages, and a book map keeps the relationship between them visible. Your reading remains the center of the experience.
How does the AI work?
Every AI feature uses GPT-6 Astra (gpt-6-astra) through the OpenAI Responses API. Connect an API key with access to that model. API usage is billed by OpenAI. There is no silent fallback to another model.
Where does my work go?
Your book is saved as local Markdown with local metadata. When you ask a question, the relevant document context is sent to OpenAI. Research may read additional pages in the current book. Nested does not publish or share your library.
What does it cost?
This build is a bring-your-own-key preview. There is no Nested subscription or payment collection. OpenAI charges for your API use. A paid personal workspace is a future experiment, not a current offer.
Can I use it on my Mac?
Yes. This project includes a packaged Apple silicon Mac app and a local browser reader. The preview is not Developer ID signed or notarized. A hosted workspace, automatic updates, and an Intel build are not included.
Start with the first
“wait, why?”
Local preview · Apple silicon · Your own OpenAI key for AI