Seated with eyes closed, one fingertip resting at the inner end of the eyebrow

B-Corps · The publishing floor

A publishing house, kept at one desk

Luminous BookFlow takes a finished manuscript, has an AI find its chapters, and from them helps make covers, posts, articles, a posting plan and courses, beside a vault for documents and a room for the team.

Begin reading
Arms opened wide to the sides at shoulder height, palms soft, chest open, eyes closed: the breath in of Opening the Chest
Arms opened wide to the sides at shoulder height, palms soft, chest open, eyes closed: the breath in of Opening the Chest

What it is

A book is finished once, and published many times

After the last full stop of a manuscript there is still most of the publishing to do. There is the cover to make and the copy for the back of it, the pictures that will sit at the head of each section, and the posts that will carry a chapter to people who will never read the whole book. There is the long article for the press's own pages, the plan that says which of these goes out on which morning, and, for a book that teaches, the course that will be built from it.

Luminous BookFlow is a desk for that work. A person uploads a manuscript, an AI reads it and draws out its chapters, and from then on the chapter is the unit that everything else is made from. On a phone the rooms are named on a bar along the bottom of the screen: Content, where posts and articles are drafted; Courses; Audio; the Vault, where documents are fingerprinted and kept; and Culture, where a team says how it is. Behind them sit a library of books with their covers, an editor that gives a book one visual theme, a reader that writes the book out as a PDF or a Word file, calendars of several kinds, and a portal where a client sees the state of their own project.

Nothing it writes is kept until somebody decides to keep it.

It is worth saying at the start what the app does not do, because a publishing house in software invites a larger claim than this one makes. Nothing in it posts to a social network. It drafts, plans, schedules and keeps the record of what is meant to go out, and a person carries each piece the last step. Its audiobook rooms keep a record of a book prepared for listening and write summaries and study guides for it, but no voice reads the book aloud inside them, and the players on those pages move without sound. Several of its analytics screens show placeholder figures, and this page leaves them out.

What remains is a great deal of real work, done by an AI where an AI does it and by plain records everywhere else. The AI reads chapters, writes drafts, proposes the structure of a course, makes pictures, reads a team's check-ins and writes reports, always at a person's request and always as a draft that waits on the screen. Nothing it writes is kept until somebody decides to keep it. That one habit, repeated on every screen, is the character of the whole app.

The arc of Painting a Rainbow, the weight settled into one leg, the gaze following the open hand
The arc of Painting a Rainbow, the weight settled into one leg, the gaze following the open hand

A day with it

Tuesday, a small press and one new book

The managing editor of a small press arrives on a Tuesday in early spring with one book on her list. The press publishes teaching books on breath and attention, and its newest, a slim manual written by a yoga teacher who has run morning classes in the same church hall for eleven years, came back from the copy-editor on Friday. Between the author's wish to be read and the press's wish to be paid sits a month of work.

She opens the Content room. The card at the top says Upload Your Book, and the place where a file should go will not take one until a title is typed, which it says in a quiet line. She types the title and the teacher's name, drags in the Word file, and watches the card say that it is extracting chapters with AI. After a minute the chapters slide in, each with its number, its title and its first line. There are nine. She chooses the third, on the breath before speaking.

The app thanks her for attuning, and the gauge on the room settles where the week has put it.

In the grid of destinations she ticks LinkedIn and sets it to three pieces, then ticks Bluesky and sets it to two. She picks the house template called Hook-Story-Offer and presses Generate. The pieces arrive in the library as cards with their hashtags. One of the LinkedIn posts is nearly right and a paragraph too long, so she presses Refine, chooses Make Shorter, reads what comes back and presses Save Changes, and the card changes in place.

For the post she cares most about she opens the LinkedIn Content Studio. Four drafts appear at once, labelled by the kind of opening each one uses, and she takes the one that begins with a story. A counter reads its length against the platform's limit of three thousand characters. She presses the refinement that ends a post with a question, removes one hashtag by pressing its chip, and asks for three pictures at the platform's proportions to look at beside the text. Then she saves the post to the library.

Late in the morning she sets the blog generator going on the same chapter: long, a how-to guide, conversational, with one keyword. The stages go by on a bar until the article sits rendered with three section pictures, a short list of suggested improvements and the AI's own estimate of how well it would rank. She saves it with its pictures. In the Book Library the manual has no cover yet, so she presses Generate Cover, and then Generate Description, which writes three paragraphs for the back of the book and a tagline that appears under the author's name. In the reader she turns two chapters with Next and downloads a Word copy for the teacher to mark up.

After lunch she opens the Calendar and starts the AI Content Calendar. She chooses the manual and the press's previous book, names the calendar for next month, sets two posts a day, and reads the strategy note before approving a dozen of the dated suggestions one by one and saving the draft. On the Editorial Calendar she sees this week laid out, with Thursday's meeting with the printer already in its square from her Google Calendar, marked as a call. She drags three library cards onto Thursday and presses Schedule, and the screen confirms three posts for that day at nine. They are a plan, and on Thursday morning she will post them herself.

At four she opens the Course Builder and presses Create Course. She gives AI Structure the subject, a beginner's level and four weeks, reads the modules it proposes and applies them. Then she opens Text to Audio, pastes the opening of chapter one, chooses a calm voice and slows it a little, and because the press has set its Azure key, the narrated file comes back as a new audio lesson in the first module, made by a synthesised voice. She renames it in the lesson editor and sets its minutes. A study group follows, named for the course and linked to it, so the first students have somewhere to talk.

Before she leaves she drops the signed author agreement into the Vault. Its fingerprint appears at once, a long line of letters and numbers, and when she presses Stamp the file is kept and an email arrives with the document's name, the time and the same line. Last, she opens Culture and does her own check-in. She chooses Activated, sets stress at seven and purpose at nine, writes that she appreciates the designer who stayed late on Monday, and asks for one quiet afternoon next week. The app thanks her for attuning, and the gauge on the room settles where the week has put it.

One arm arcs over the head, the other reaches out to the side, palm up, the body leaning gently: Painting a Rainbow
One arm arcs over the head, the other reaches out to the side, palm up, the body leaning gently: Painting a Rainbow

The book

From a manuscript to a book that can be handed on

Everything in the app starts from a book record, and the book record starts with an upload. The card asks for a title and an author first and refuses a file until a title is there, which is a small courtesy that prevents a library full of untitled manuscripts. It takes a PDF, a Word document or a plain text file, by dragging or by browsing. Once the file is up, an AI reads it and returns its chapters, each with a number, a title and its full text, and the book is saved with all of them. On the Books tab every book is listed with its author and the number of chapters found, so a person can see at a glance whether the reading went as expected.

The chapter list is where most of the app's work begins. Each chapter is shown with its number, its title, or a plain chapter number where the manuscript gave none, and the first line or so of its text, and choosing one makes it the source for every generator in the Content room. With no book yet, the list says so plainly and points to the upload.

They are the book in a form a person keeps.

The Book Library holds every book, newest first. Each card carries its cover, title, author, tagline and description, with counts of its chapters and sections. A book without a cover shows Generate Cover, which asks an image model for one from the title, the author and the first chapter's title. The style of that prompt is fixed, a soft watercolour with warm colour and room for the type, so every cover from this button belongs to one family. A small wand makes another. A book without a description shows Generate Description, which has an AI write two or three paragraphs of back-cover copy and a tagline from the title, the author and the chapter titles, and saves both to the book.

The Book Editor gives a book one visual identity and a picture for every section. On its left a person sets the book's themes, its visual style, art style, palette and mood, by typing them or by asking an AI to propose them from the title and the chapter titles. Image settings choose the pictures' shape and style. Generate All Prompts then writes an image prompt for every section from its title and its opening lines, in the book's theme, and Generate All Images makes one picture per section while a counter climbs through them. On the right each section shows its number, its opening text, its prompt in a box a person can rewrite, and its picture, with three buttons: a new prompt, a new picture, and Vary, which makes another picture from the current one with a slight shift of view. Nothing is kept until Save Changes. One plain note belongs here: the editor works over the sections a book's record already holds, and an uploaded manuscript arrives as chapters, so the app has no step of its own that divides a chapter into sections.

The reader opens any book with a row of badges to move between books, Previous and Next with the chapter's place in the whole, and a soft slide from one chapter to the next. Where a chapter has sections, each appears with its saved picture and its text, and otherwise the chapter's full text is set on the page. A Download menu writes the whole book in the browser, as a PDF with its title, author and headings, or as a Word document whose title and chapter headings are real heading levels, ready for an editor's tracked changes. Both files carry the text without the pictures. They are the book in a form a person keeps.

Feet together and soft, eyes closed, the hands gathered over the lower belly after the last breath out

A book is finished once. It is published many times, a chapter at a time, to people who will never see the whole manuscript.

In profile, one hand lifted forward at chest height and the other drawn back and down, the gaze following the front hand: the turn of Rolling the Arms
In profile, one hand lifted forward at chest height and the other drawn back and down, the gaze following the front hand: the turn of Rolling the Arms

Carrying it out

One chapter, many rooms, and the week that paces them

The Content room, which names itself ContentCraft, turns a chosen chapter into writing for other places. Thirteen destinations are grouped by kind, from the social networks to long-form writing and business copy, and ticking one reveals a count from one to ten. An optional template gives the AI a structure and a tone. Six come with the app, and a person can write their own with bracketed places for the parts. An AI reads the first three thousand characters of the chapter and writes that many pieces for each destination, each with three to five hashtags, and each piece is saved to the library carrying its book, its chapter, its platform and the template it came from.

The library keeps every piece as a card with a row of plain verbs: schedule it, refine it, repurpose it, copy it with its hashtags, or delete it. Refine opens the text with six rewrites by an AI, from making it shorter to making it more engaging, and keeps the result when a person saves. A hook generator writes twelve possible opening lines on a topic, each tagged with its kind, a sentence on why it works and the platform it suits, and a chosen line goes straight into the writing assistant.

None of these rooms sends anything. They hold a precise and shared plan.

For a single platform there is a studio. It opens with four drafts written at once, one leading with a hook, one with a story, one with practical value and one with an invitation to reply. A chosen draft moves into an editor that counts its characters against the platform's own limit, offers one-press AI refinements, and shows its hashtags as chips that a tap removes and a typed word adds. It will make three pictures at the platform's proportions to look at. Further settings add a tone, a teaching framework such as Bloom's taxonomy or Kolb's cycle, a creativity level and instructions of a person's own, and Apply & Regenerate writes the drafts again. Save to Library keeps the platform, the text and the hashtags.

The blog generator writes the long article. Its stages are shown on a bar: an outline with three title options and a meta description, the article itself at a chosen length from about eight hundred words to about four thousand, a pass for search with suggested improvements, a picture for each of the first three sections, and five questions with their answers. The scores it prints are the AI's own estimates, never measurements. The article can be copied as markdown, regenerated, or saved to the library with its pictures. The repurposer takes any saved piece into eleven other shapes, an email newsletter with subject lines, a thread, a short video script or a Quora answer among them, each with a length limit if a person wants one.

The studio, the blog generator and the repurposer read the same settings, kept on the person's own account: a preferred tone, an audience, a creativity level that runs from factual and safe to highly imaginative, notes on the brand's voice with the phrases to use and to avoid, and default hashtags. Set once, they follow a person into every room. Beside them sits an optimiser for search and for the AI-written overviews that search engines show: a pasted piece comes back with the AI's estimate of its readiness, keywords, three better titles, a meta description, five improvements and three gaps to fill. A related tool reads up to ten of a person's own blog pieces and suggests links between them, with anchor words and a keyword for each, and keeps the suggestions as a list.

The pacing happens on three calendars. The AI Content Calendar asks for one or more books, a weekly or monthly span, a start date, one to five posts a day and the platforms in play; an AI reads the chapter titles and opening lines and returns a strategy note and a dated plan, each post with its time, platform, topic and source chapter. A person approves the posts one by one and saves the calendar as a draft, and in its week view a post's label moves from suggested to approved, scheduled and published as a person presses it. The Content Schedule is a month or a list of every scheduled post. Scheduling a piece takes a date, which can never be one already past, a time in half-hour steps, any number of titled links, an optional review of a product with its stars and its affiliate link, and private notes.

The Editorial Calendar is the one a press will keep open. It shows a week, with every undated library piece waiting on the left as a card that can be dragged by pointer or by keyboard, and every day showing its meetings from the person's Google Calendar, read with a read-only permission, beside its scheduled posts. A day's pieces are scheduled together at nine in the morning. A publishing queue keeps jobs for thirteen platforms, Medium and Substack among them, with their status, and lets a failed one be put back in line. An AI agent called the social media assistant works with these records in conversation for five platforms: Twitter, LinkedIn, Instagram, TikTok and Pinterest. It writes, changes and removes scheduled posts, drafts replies to comments and presents every one for approval first, and each of its steps can be opened and read. None of these rooms sends anything. They hold a precise and shared plan.

In profile, leaning forward over a bent front knee, the arms reaching back behind the hips
In profile, leaning forward over a bent front knee, the arms reaching back behind the hips

Teaching and listening

Courses, audiobooks and the people who read together

A book that teaches becomes a course in the Course Builder. Its tabs hold the details, the curriculum and the terms, and a switch sets the course published or not. The curriculum is a set of modules, each holding lessons of six kinds: video, audio, written text, a quiz, an assignment, or an audiobook already in the app. Four AI helpers sit above it. AI Structure takes a subject, a level and a length of two to twelve weeks and drafts a whole course, its title, description, outcomes, prerequisites and modules of typed lessons with their minutes. From Audiobook builds a course from an audiobook's chapter titles and description, and can be made to follow an uploaded syllabus, with a quiz for each module. AI Generate writes lesson text, quiz questions with explanations, and suggestions of further reading. A fourth reads a lesson's text and returns a summary, its takeaways, ideas for pictures and diagrams, and a quiz of a chosen kind.

Text to Audio turns writing into an audio lesson. A person types or pastes text, or uploads a document whose text is drawn into a box they can edit, and a running count turns the words into an estimate of minutes. Nine neural voices are offered by name, with Ava first, and a speed from half to double. Where an Azure Speech key has been set in the app's settings, the text is sent to Microsoft's speech service, the narration comes back as an audio file, and the builder adds it as a new lesson in the first module. The voice is synthesised. The lesson editor names each lesson, sets its kind and minutes, takes a link or an uploaded file for video and audio, and can open a single lesson to people who have not yet enrolled.

A study group gathers around a course.

The course player lays the modules down one side, each lesson marked done, open or locked, with the learner's progress drawn above them. Enrolling creates the learner's record. A written lesson is set as formatted text with a button to mark it complete, which records the lesson and recalculates the percentage, so progress reflects lessons actually finished. Under every lesson there is a discussion: a person writes a post with a title, others open it and reply, and every reply is kept with its author's name and its date. Video and audio lessons show the face of a player in this version, and nothing plays inside it, so a course here is strongest as written lessons and conversation.

On the Courses page a search box takes a plain sentence, something on breath for anxious mornings, and an AI reads every published course and returns the ones that fit, ranked, with the reason for each. With no search showing, the same AI offers courses to continue, a next one and new ground to explore, each with a reason. A panel in the player asks an AI to name a learner's strengths, the areas to give attention to and a recommended level, and writes them into the profile that the search reads. The course analytics count enrolments, the share completed, each learner's progress and the life of the discussions, which are the numbers a teacher can use.

A study group gathers around a course. A person names it, describes it and links it to a published course if they wish, and becomes its first member and its keeper; others join with one button, and the group shows its course, its members and a discussion of its own. A reading group is described as a shared listening and can be formed and joined in the same way. For now it holds a name, a description and its members, which is the beginning of a circle rather than the whole of one.

The Audio room is headed AudioCraft. It holds a shelf of audiobook records that can be searched by title or author, with a row of the books a person has made. Convert PDF takes a PDF, has an AI draw out its text and chapters, and keeps an audiobook record with its chapters numbered and titled. No narration is made in this step, and the book's own page draws a player that moves without sound, so the room is best read as a book prepared for listening. Each audiobook's page shows its cover, its chapters and its reviews, where a person gives one to five stars and a written review that keeps a history of its versions. Generate Full Summary runs three AI passes and returns an overall summary with its takeaways, a summary and key points for every chapter, and a study guide of discussion questions, actions and related topics, which can be saved. The AI writes these from the title, the description and the chapter titles rather than the book's words, so they are a teacher's starting notes, ready for a reading group's first evening.

A small fold from the hips, knees soft, both arms sweeping down and back past the hips, eyes on the floor ahead: Rowing the Boat
A small fold from the hips, knees soft, both arms sweeping down and back past the hips, eyes on the floor ahead: Rowing the Boat

The house behind the books

Documents, clients, and the team that makes them

A press accumulates paper: agreements, permissions, proofs, invoices, the signed page that says a chapter is the author's own. The Vault keeps them. A person chooses a PDF or a Word file, or several at once, and names and describes them. If they wish, they record a value in US dollars, euros, pounds, bitcoin or ether, with a note on how it was reached. Before anything is uploaded the browser computes the file's SHA-256 fingerprint, a sixty-four-character line that any change to the file, however small, would alter, and shows it in full. Pressing Stamp keeps the file, writes a record with its fingerprint and the moment it was stamped, and sends the person an email carrying the document's name, the time and the fingerprint. The fingerprint is kept by the app and the email; it is not registered anywhere else, and its value is that anyone holding the file can compute it again and compare.

Each document then has a page of its own, set like a deed, with its details, the moment of stamping and the full fingerprint with a button to copy it, and one line that is true of every fingerprint: alter the document and the hash changes. The Vault's analytics are drawn from its own records: the total stated value, the number of documents and how many are PDFs or Word files, a bar of documents stamped in each of the last six months, and a bar of value added month by month. A log lays every stamping out in time order with its name, its tags, its moment, its fingerprint and a link to the file, and exports the whole as a spreadsheet for an accountant or a solicitor.

Under the Culture door sits a room the old description of this app never named, and it may be the part a growing press comes to value most.

The Client Portal is the press's window for the people it publishes for. A signed-in client sees only the projects that carry their own email, and an administrator sees all of them. Each project is a card that says what kind of work it is, whether a book, an audiobook, a blog, a course or an app, and the stage it has reached between intake and completion, with its progress as a bar. Opening one shows its milestones with their dates, a log of the documents filed for it with their fingerprints, and the PDF reports prepared for it, ready to download. The app has no screen of its own for opening a project, so the portal shows what has been entered into the record, and it shows it clearly.

Under the Culture door sits a room the old description of this app never named, and it may be the part a growing press comes to value most. Department Setup gives each department a name, a description and one of ten colours, and Team Setup adds each person with their role and department. A check-in takes three short steps. The person chooses a face for their mood. Then they choose one of four states of the nervous system (regulated, activated, shut down or mixed) and set five sliders from one to ten for energy, stress, creative flow, connection and purpose. Last, they write what they are appreciating today, what is hard, any insight that is becoming clear, and what support would help.

The Culture Field gathers the week. It averages the last seven days into a single figure it calls field coherence, drawn on a gauge whose needle swings into place, and beside it the count of each state, how many took part, a card for each department and a card for each person with their latest reading and, when they asked for it, the support they need. An AI reads the latest thirty check-ins as an organisational health consultant would, and returns a short assessment, the patterns it notices, recommendations grounded in appreciative coaching, anything that needs attention, and the themes of what people appreciate. Field Reports do the same for an organisation or a department over a day, a week, a month or a quarter: an executive summary, practices with their reasons and priority, action items with an owner and a date, and trends, saved and printable, and with no check-ins in the period the page says so and writes nothing. The Practice Library holds the practices to answer with, and an empty one fills in one step with six to begin from, box breathing, an appreciation circle and an end-of-day recovery ritual among them.

Lying on her back with eyes closed, the fingertips of both hands cradling the base of the skull, elbows resting wide

Every piece waits on its day with the week's meetings already there, so the week is planned once, in one place.

In profile, the waist turned and both hands raised up and back, the gaze rising past them
In profile, the waist turned and both hands raised up and back, the gaze rising past them

In B-Corps

How it sits beside the other eighteen

Luminous BookFlow is one of the nineteen applications of B-Corps, the house's own operating system. Each is installed on the house's Mac by its own method, against its own cloud record, side by side on one desk that starts, opens and rests each application and, every two minutes, restarts any that has stopped. A stock-take of every application's records, functions and agents is kept beside them. BookFlow keeps its own sign-in, as each of them does today, and on every page it quietly tells the house's registry of applications who is signed in and, after a piece is generated, what kind of piece it was and from which book.

The connections between it and its siblings are a matter of how a person arranges the desk, and none of them passes data by itself. The nearest is The Luminous Book, the studio that reads a business and writes down its voice. A person who keeps the two open side by side can paste that written voice into BookFlow's brand-voice notes, and from then on the studio and the repurposer write to it. Chapter to Blog carries a book into posts chapter by chapter as well, and a press might keep that rhythm there while it keeps the calendar and the queue here. AudioBook AI prepares a book for listening, and its prepared chapters and this app's audiobook pages, summaries and study guides describe one object from two sides. ContentSpark sets one source out across every channel and into a course, which is the same motion this app makes from a chapter.

Text moves between them by hand, and the arrangement holds.

Further out, Resonance Canon reads each book as a course with others and a coach, and a teacher who builds a course here would keep Canon open for the reading itself. Resonance Canopy carries organisational assessment and consulting, and the Culture Field's check-ins and AI field reports are exactly the reading of a team that a consultant would want beside an assessment. B-Corp, the foundation, keeps one record of the organisation and its people, and the departments and team members set up here are the same kind of record kept for one press. SquareDash is a marketing workspace for websites, and its articles and search work and this app's blog pieces and optimiser answer the same question about a site's writing. Text moves between them by hand, and the arrangement holds.

Both arms lifted up and to one side, the face turned up past the higher hand: Gazing at the Moon
Both arms lifted up and to one side, the face turned up past the higher hand: Gazing at the Moon

Who it is for

A press of one, or a press with a staff

It is for the author who has finished a book and now faces the other half of publishing alone at a kitchen table, and for the small press that publishes six or eight books a year with three people and no marketing department. It suits a teaching practice whose books are also its curriculum, where every chapter is a class and every class could be a course, and an organisation that publishes its own books and wants the books, their posts, their courses and the people who make them looked after in one place.

What it replaces is a familiar collection. There is the writing assistant in one tab and the social scheduler in another, and the course platform with its own sign-in. There is the folder of signed agreements named by date, and the quarterly survey of how the team is doing that nobody quite reads. Each of those tools knows nothing of the book, so the book is pasted into each of them and the thread between a chapter and the post it became is lost. Here the chapter is the thread. A post knows which chapter it came from, a course knows which audiobook record it was built on, and a day on the Editorial Calendar knows both the pieces planned for it and the meetings already in it.

Here the chapter is the thread.

What makes it different is a set of choices that run through it. The AI is named wherever it works, and everything it writes waits as a draft for a person to keep or set aside. The plan is kept honestly as a plan: a post marked published is marked by a person, and nothing goes out on its own. The documents are fingerprinted in the person's own browser before they leave it, and nothing more is claimed for the fingerprint than a fingerprint can bear. And the team that makes the books is asked, gently and in its own words, how it is, with appreciations asked for first. For a house that publishes on breath and attention and living systems, that last room is not an extra. It is the same care, turned toward the people doing the work.

How it is offered

Installed for an organisation, and kept

Luminous BookFlow is offered to an organisation in three ways: on its own, as part of the publishing floor, or inside the whole of B-Corps, where every application on these pages is installed together, set up in the organisation's own voice and with its own records, and kept running for as long as the organisation keeps it. The house carries each application into its suite for good, so what is set up now is still there, and still mended, in the years after.

The fee reflects the work of installing and shaping it, and the size of the organisation it serves. It is set with each organisation, after a conversation about what the application will carry for it.

Arranged in conversationLuminous BookFlow · B-Corps

Standing quietly with eyes closed, both hands resting over the lower belly, one over the other: the close of the form
Standing quietly with eyes closed, both hands resting over the lower belly, one over the other: the close of the form

The care it takes

Drafts that wait, and a quiet close

The care shows first in the drafts. Every generator says what stage it is in and shows its progress, every dialog can be closed, and the longer ones have a way back and a way out, so nothing is lost by stepping away halfway. Nothing an AI writes is kept until a person chooses to keep it, and regenerating discards only the draft on the screen. Content settings are saved to the person's own account and follow them into each room. The social agent is told to present every drafted reply for approval before anything is posted and to bring negative comments to a person rather than answer them, and every one of its steps can be opened and read.

It shows again in what the app will and will not touch. Google Calendar is read with a read-only permission, so meetings appear and the calendar is never written to, and without a connection the panel says so plainly. The audio and calendar services refuse any request that does not come from a signed-in person. A client sees only their own projects. Past days cannot be chosen for a post, dragging works from the keyboard as well as the pointer, and every empty room says what to do next, from a library with no books yet to a period with no check-ins.

The manuscript stays at the centre of the desk, where it began.

A book is kept for decades, and the work of carrying it outward is done in weeks, a chapter and a morning at a time. Luminous BookFlow keeps that work in one place, says plainly which parts an AI drafted, and leaves the last step, the sending and the reading aloud and the teaching, to the people whose book it is. The manuscript stays at the centre of the desk, where it began.

B-Corps

It begins with a conversation

Request a conversation