Live in Flow

Privacy, in plain English

The detailed sections below are the full policy.

Live in Flow

Your child’s data

This page covers Kid’s Flow Academy: what is stored about a child, what leaves this system and who receives it, what you agreed to, and how to get all of it deleted. It is written in the same words you are shown when you agree, because a policy that paraphrases its own consents is a second document that can drift from the first. Below the Kid’s Flow Academy sections, a second part covers a member’s own Finance Fire: the Money Command Center, the AI Family Office and its document vault.

The short version

Deleting his record

Open his page in Kid’s Flow Academy and use the erase option there. It removes every row about him across every table that holds one: his profile and intake, his sessions and transcripts, his word bank, his flags, his assessments and written work, his handwriting photographs, his placement runs and answer traces, his badges, his councils, his team memberships and raid attempts, his coaching records, and the consent rows themselves. You are shown the count from each table rather than a confirmation message, so you can see it happened.

Three things deletion here cannot reach, and they are listed rather than summarised because each one has a different company and a different answer to who is on the contract. The first is anything already sent to an AI provider. Every AI request a member makes runs on that member’s own AI key and own account with that provider, so what went out is your relationship with that provider and their retention, governed by their terms and not this system’s. Live in Flow never pays for or routes a member’s AI requests through its own account. The one exception is the owner’s own record, which is not a member. That transmission happened, and this system has no ability to recall it. The second is anything he said on a live guide call, which streamed to Daily, Tavus and OpenCharts. Unlike the AI path, that account is Live in Flow’s rather than yours, there is no agreement with any of them covering your child, and nothing here can reach in and erase what was sent. The third is any sentence already sent to the speech company to be read out loud. That account is Live in Flow’s too, there is no agreement with that company covering your child, and nothing here can reach in and erase what was sent.

Withdrawing a consent is not the same as deleting. Withdrawing stops anything new being stored. Deleting is its own button, and it is deliberately yours to press.

What is stored about your child, real voices and live guide calls

I am this child's parent or legal guardian, and I can decide this for him. I agree to him being enrolled in this home school module. What is stored about him: his first name, the hero name he goes by on screen, his age, his grade, the state he lives in, the day and the month of his birthday, whether he is a boy or a girl, the realm he picked for the quest world, the browser voice he picked for the tutor, the word he uses for whoever teaches him, a family code the children in his family share, and a 4 to 6 digit PIN he uses to sign in. When he calls a guide, this system also stores that the call happened, how long it ran and which guide he chose. Two of those are worth a sentence each. His state is stored because it decides which standards his work is written against and which home education law I am answering to, and for nothing else. His birthday is a day and a month and never the year, because the year would be the most identifying thing in this file and the module does not use it for anything; if a year is ever needed, a parent will be asked for it separately and told why. What a parent writes in the intake: his interests, how long he can hold attention, what motivates him and what shuts him down, his reading level, any test scores a parent uploads such as MAP or Lexile, and a sample of his writing. What he produces as he works: full turn by turn transcripts of him talking to the tutor, the words he misspells and the words he meets, the passages he flags, his quiz answers and written responses, photographs of his handwritten work that a parent uploads, and his progress: grade placement per subject, promotions, XP, badges and scores. His PIN is never stored. Only a one-way hash of it is kept, which can check a PIN and cannot reveal one. Where it lives: encrypted under a key this system uses for nothing else, in a database hosted by Neon in the United States. His hero name, his family code, his age, his grade, his state, the day and month of his birthday, whether he is a boy or a girl, his realm, his voice choice, the word he uses for his teacher and his progress numbers are all stored plainly rather than encrypted, because they are what he types to sign in, what decides which work he is set, and what the screens count. His real first name is encrypted, and so is what a parent writes in the intake. Who can see it: no screen in this system lets anyone at Fathers on Fire open my child's record. There is no coach view and no admin view of a child. That is enforced in code, on every read and every write. It is not a claim that no human being could ever reach the database, because somebody holds the credentials and the key, and I would rather be told that plainly than promised otherwise. If the law ever compels disclosure, the parent who enrolled him will be told unless telling them is prohibited. How long: as long as it is kept. Nothing here deletes on a timer. The parent who enrolled him can erase everything about him, at any time, from his own page, and is shown a receipt of exactly what was removed, table by table. Erasing stays with that parent. Real voices and live guide calls: these are part of Kid's Flow Academy, and this one yes covers them. They are not separate switches on his page. Once a parent agrees to this wording they are on for him, and they stop if it is withdrawn. Being read to in a real voice: the tutor's replies are read to him out loud, and so is the hello a guide says when he presses a face to pick it, in a real voice made by an outside speech company. That is OpenAI or ElevenLabs, whichever this system is set up to use, on an account Fathers on Fire holds rather than one of mine. What is sent is the words the tutor has just said to him, as text, and which voice he picked. His hero name is replaced with the word hero before a tutor reply leaves, and that is enforced in code. The one exception is the hello: it is "Hi", his hero name and "How are you today?", sent as they are, because a guide who cannot say your name cannot greet you. The tutor's words can quote or refer back to what he wrote a moment earlier, so something he wrote can leave inside the reply to it. Audio comes back and is played to him. It is not saved. Live video calls with a guide: he picks a guide, and when he presses Meet your guide the guide appears as a face that speaks and listens. It is a real video call with a computer, not a person. His camera is never turned on, so the guide can hear him and cannot see him. His microphone audio is streamed live for as long as the call lasts, to a video room run by Daily (daily.co) for Tavus, the company that makes the guide's face and voice. Fathers on Fire reaches Tavus through OpenCharts. The server's requests to OpenCharts to open and close a room say which guide was called and nothing about him. His browser then connects to Daily directly. The guide is handed, through his browser, into the same room: his hero name, what is on his lesson screen (the step and its goal, the tutor's latest question, his last answer, the subject and session number, the XP he has earned, the struggle he is working on today, the skills due for a warm-up, and whether a retell is open), and what his guides kept of his last few calls. If a parent has also agreed to the AI tutor wording that names his teaching notes, it is handed how to teach him as well: teaching moves only, with no scores, no counts, nothing a parent wrote and no fixes. Tavus and Daily receive all of it. His real first name is never sent. Reading his screen during a call: while a call with his guide is live, and only then, his browser reads the words showing on the Kid's Flow Academy screen he is on and hands them to the guide in the same room, so the guide can answer what he asks about the page and walk him through it one small step at a time. That is the full text visible on his quest screen: the page title and headings, the lesson and its goal, his chat with the tutor as it shows, questions and choices, scores and progress, the book or chapter named, notices, the names of the buttons, and what he is typing in an answer box, including words he has typed and not yet sent. It reads this app's own page and nothing else: not another tab, not another window, not another program. A PIN, the family code and any password box are never read, and his real first name is replaced with his hero name before the text leaves, in code. Tavus and Daily receive this text. The guide is told never to give him the answer to scored work and to give no hints on a test, but it does see the question and what he typed, and nothing in this system checks what a guide says on a call. Fathers on Fire does not save the text it reads from his screen. The reading stops when the call ends. Showing his screen: during a call he may be offered a button to show his guide his screen. If he presses it, his browser asks him which screen, window or tab to show, and then streams live video of it to the same Daily room for Tavus, for that call only, so the guide can look at it. He chooses what is shown, so it can take in things outside Kid's Flow Academy if he picks them, such as another window or a message that pops up. It is his screen and never his camera. A notice stays on his screen for as long as it is being shown. It stops when he presses Stop showing, when the call ends, or when he leaves the page. Fathers on Fire does not record or save that video. Daily and Tavus receive it. Tavus's own privacy policy says its service is not intended for anyone under 18, and Fathers on Fire has no agreement with Tavus or Daily about what they keep of what he shows or for how long. Where his browser allows it, he is offered the tab he is already on first. The button appears only once Fathers on Fire has made it available, and this yes covers it from then. What those companies keep, and for how long, is their business under their terms. Fathers on Fire has no agreement with OpenAI, ElevenLabs, OpenCharts, Tavus or Daily covering my child, and neither of us can say what they keep. Once a sentence has been spoken to a vendor or he has said something on a call, that has been sent. Erasing my child's records here does not reach into their systems and does not recall it. What this consent does not cover: sending his work to an AI provider, his browser listening to his microphone so he can talk to the tutor, the guide keeping and summarising the words of his calls so it can remember him, and showing his hero name to other families. Each of those is its own decision, with its own wording and its own switch on his page, and saying yes here says nothing about any of them. I can withdraw this at any time, and so can his other parent if one is linked to him here. Withdrawing stops anything new being stored about him, and it stops real voices and live guide calls. It does not delete what is already there, because deleting is its own button and it stays with the parent who enrolled him.

The AI tutor

I am choosing to let my child work with the AI tutor. Here is exactly what that sends and where it goes. What is sent, every turn: his hero name, his grade level and writing stage, the book and focus points he is working on, his word bank, the passages he has flagged, the full transcript of the session so far, and his teaching notes, described below. When a parent uploads a photograph of his handwriting, the whole unedited image goes with his hero name and writing stage. What is never sent: his real first name. He is only ever the hero name he chose. That is a deliberate design decision and it is enforced in code. His teaching notes: his tutors keep a short list of notes on how to teach him, built only from what he has actually done. The lines come from his retells and their scores, his struggle reps and how many he got, his scored sessions (tries per step, his own too easy or too hard answer, measured minutes), the misconceptions a parent confirmed, the summaries of his guide calls, and what a parent wrote in the intake. Every line carries where it came from and the date. Old lines fall away. Which of those lines leave: only the measured ones, from his retells, his reps and his scored sessions. They carry his scores and counts. The text tutor is handed them at the start of every turn so it teaches him the way the last session taught it to. And between sessions, after a session ends or a guide call is summarised, a short merge call sends those same measured lines to the AI provider and gets back a tidier set of teaching instructions, which is checked against the evidence before it is kept. Which lines never go to the AI provider: what a parent wrote in the intake, any misconception fix a parent confirmed, and the lines drawn from his guide calls. The tutor is never handed them and the merge call never sees them. That is enforced in code, by the kind of line and by its key. The teaching moves from these notes (no scores, no counts, nothing a parent wrote and no fixes) are also handed through his browser into his live video calls, so his guide teaches him the same way, and Tavus and Daily receive them there; the calls are set out under live video calls. What his video guide remembers of his calls is a separate decision with its own wording, under live video calls. Where I see all of it: every line, including the ones that never leave, is on his Teaching page. I can clear any line, and so can any other parent with access to him here, and a cleared line stays cleared through every later merge. Clearing them all starts the notes over from his evidence. Ask Young David: in Kid's Faith Academy he can type or say questions to Young David, a pretend Bible character played by the same AI tutor, on the same key described below. What is sent with each question: his hero name, his age, his question, and the last few lines of that chat. The chat itself is not saved here. If his words, or the AI, suggest he may be hurt or unsafe, the story stops, he is told to come and find a parent or another trusted grown-up right away, and that one message is saved to his tutor log for his parents to see. Where it goes: to Anthropic or OpenAI, whichever the parent who enrolled him chose on the Kid's Flow Academy page, on the family's AI key, held on the account of the parent who enrolled him. Every AI request made for his family runs on that key and that account with that provider. Fathers on Fire never pays for or routes those requests through its own account. If no key has been added, nothing is sent and the tutor does not run until the parent who enrolled him adds one. The merge call runs on the same key. Fathers on Fire is not a party to that parent's relationship with the provider, and what the provider keeps, and for how long, is governed by their terms and not by this system's. There is no monthly limit from Fathers on Fire. The one family that runs on a Fathers on Fire account is the owner's own, which is not a member's. Two things I should know. Fathers on Fire always writes the prompt, chooses the model and decides what is transmitted, so using the family's own key does not pass responsibility to the parent who holds it by itself. And both providers' developer terms require the account holder to be an adult and place responsibility for any data about a child under 13 on that account holder, who is the parent who enrolled him. This part cannot be undone: once a session, or a merge call, has been sent, that transmission has happened. Erasing my child's records here does not reach into a provider's systems and does not recall what was already sent. I can withdraw this at any time, which stops the tutor from running, stops the merge call, and stops his notes being handed to anything outside this system. Everything else in the module keeps working without it.

Talking out loud

I am letting my child talk to the tutor out loud instead of typing. What happens: his browser listens to the microphone and turns his speech into text. This system never records, stores or transmits audio, and there is no recording of his voice anywhere in it. Where his voice goes: the browser does the listening, and in Chrome that means his microphone audio is streamed to Google's speech service and text comes back. That is Google's service under Google's terms. Neither Fathers on Fire nor I have any agreement with Google covering my child's voice, and neither of us can say what Google keeps. He can always type instead, and everything works the same way if he does. I can withdraw this at any time, which turns the microphone off for him.

The guide remembering his calls

I am letting my child's AI guide remember him from one live video call to the next. Live calls with the guide he chose are part of Kid's Flow Academy for every enrolled child, so this wording describes the whole call. My yes here is what turns on the keeping and the summarising of his calls. Until I withdraw this, the guide is also handed, at the start of each call, how to teach him and whatever his guides already kept of his calls, as it was before this wording, whether or not I have said yes to it. What happens on a call: he picks a guide, and that guide appears on screen as a face that speaks and listens. It is a real video call with a computer, not a recording, and not a person. He starts it himself by pressing Meet your guide, on his lesson screen or on the guide's own page. His camera is never turned on. Nothing on his side of the call asks for a camera, so the guide can hear him and cannot see him. No picture of him leaves this device. What leaves this device while he is in a call: his microphone audio, streamed live for as long as the call lasts. His browser asks before the microphone opens. What he says and what he asks are processed by outside companies on their own systems. Where his voice goes: to a video room run by Daily (daily.co) for Tavus, the company that makes the guide's face and voice. Fathers on Fire reaches Tavus through OpenCharts, the company the guides were set up with. How a call starts and ends: the Fathers on Fire server asks OpenCharts to open the room, and tells it when the call is over. Those requests say which guide was called and nothing about him. Not his name, not his age, not our family code. During the call his browser connects to Daily directly. These companies hear his name only if he says it out loud. What the guide is handed, through his browser, into the call: at the start, how to teach him (the teaching moves from his notes, with no scores, no counts, nothing a parent wrote and no fixes) and what his guides remember of his last few calls. During the call, what is on his lesson screen: the step, its goal, the tutor's latest question, his last answer, his hero name, the subject and session number, the XP he has earned, the struggle he is working on today (in kid words, from the Before today check filled in for him) and its boss, the skills due for a warm-up, and whether a retell is open. All of it goes into the same room his voice goes to, so Tavus and Daily receive it. It is the same memory whichever guide he picks. What this system keeps, so the guide can remember him: that a call happened, how long it ran and which guide he chose; the words of the call as text, both sides, what he said and what the guide said, as the guide's company heard them; and a short summary of each call. The video and the audio are never kept. The words and the summaries are encrypted. How the summary is written: when a call ends, the words of that call are sent to the AI provider, on the AI key the AI tutor decision describes, and a short summary comes back. Without the AI tutor decision there is no summary, and the guide is reminded with the last few lines of the call instead. Where I see it, and how I make the guide forget: his page lists every call, what the guide remembers of it, and a Forget button. Forget removes that call's words and its summary for good, and the guide is never handed them again. I can press it, and so can any other parent with access to him here. Erasing him removes all of it. The guide's background is removed on this device, in his browser, so the guide can stand on the page. Nothing is sent anywhere for that. This is more than the "talk to it out loud" permission. That one sends speech, gets text back and keeps nothing. This one opens his microphone to outside companies for the length of a conversation, a face answers him, and what was said is kept here as text so the guide can bring it up next time. What those companies keep: that is their service under their terms. Fathers on Fire has no agreement with OpenCharts, Tavus or Daily covering my child, and neither of us can say what they keep or for how long. The guides run on an account Fathers on Fire holds, not one of mine. This part cannot be undone: once he has said something on a call, it has been sent, and once a call has been summarised, that send has happened too. Erasing my child's records here does not reach into those companies' systems and does not recall it. He never has to use it. Every lesson, every trial and every subject works exactly the same way with the guide turned off. I can withdraw this at any time. Withdrawing stops the memory: nothing new is kept from his calls, nothing is summarised, and nothing already kept is handed to the guide again. It does not delete what is already kept, because Forget is its own button and I should be the one who presses it. His calls themselves keep working, and the guide starts each one fresh.

Shared Fire Team boards

I am putting my child on a shared Fire Team board with other families' children. What the other families see: the hero name he chose, his level, his XP and his Flow Score. That is the whole list. What they never see: his real name, his age, his grade, anything he wrote, anything the tutor corrected, his scores on a trial, or any part of his profile. That is enforced in code, not by convention. Why this is opt in and nothing defaults onto it: a leaderboard is the game these children are already playing, and it is still other people's children looking at my son's number. Nobody is put on one without a parent choosing it. I can withdraw this at any time, which removes him from every board and every team raid immediately.

Win texts to a parent's phone

I want texts on my own phone about my child in Kid's Flow Academy: a text when he has a big win, and a short daily lesson report after each day he does a lesson. Who gets it: me, at the mobile number I type here. Nobody else, unless the other parent says yes for themselves, as the next part explains. The other parent on this membership can also turn this on from their own login, with their own number, and neither of us can turn it on for the other. I can also add the other parent's mobile. That number gets one text asking them to reply YES. Nothing else is sent to it unless they reply YES, and they can reply STOP at any time. If that number already gets win texts from an earlier yes, it gets one text asking them to reply YES to the daily lesson reports too, and no report goes to it unless they do. Win texts: when he beats a struggle boss, or sets a new personal best telling a story back in his own words. Something like: "Fathers on Fire: Blaze beat The Summary Serpent today. Ask him to tell you about it at dinner." Daily lesson report: after each day he does a lesson, one short text with what went well, what was hard and what to practice next. It is built only from what his lessons recorded that day: the subjects, the minutes, the blocks he finished, his XP, how the lesson felt to him, his story retell, any boss he beat, and the skills he worked on. Something like: "Fathers on Fire: Blaze finished reading today, 32 min, all 4 blocks. Good: retold chapter 8 in his own words. Hard: catching the key details in his retell. Work on next: ask him to read chapter 9 aloud and tell you the 3 big events." A part the lessons did not record is left out, never guessed. What the texts say about him: his hero name, the name he chose for the quest. His real name is never in a text. A daily lesson report names, in plain words, a skill he found hard and what to practice, because that is what a report is for. No text ever carries a retell or session score, anything he said, anything he wrote, or who checked what on the Before today list. How often: at most three texts a day to my number, win texts and reports together, and at most two of those are lesson reports. Never between 8 at night and 8 in the morning in our own time zone. A text due in those hours waits for the morning. Where it goes: the text is sent through Twilio, the company that delivers texts for this system, on an account Fathers on Fire holds. Twilio sees my number and the words of the text. That is their service under their terms. What is stored here: my number, encrypted, and a record of each text that was sent, so the same win or the same lesson day is never texted twice. The words of a text are not kept. How to stop: turn it off on his page, which also forgets my number, or reply STOP to any text. Either one ends it right away. Reply HELP for help. The texts never go to my child. Cost: Fathers on Fire does not charge for these texts. Message and data rates from my phone carrier may apply. Full terms for these texts: os.fathersonfire.org/sms. My mobile number is never sold or shared with anyone for marketing.

Your own finances: Finance Fire, the Money Command Center and the AI Family Office

This part is about a member’s own money records, not a child’s. Each section below is the exact wording you are shown when you agree, for the same reason as above. Nothing is stored or sent until you have agreed to the section that covers it, and each can be turned off again. No screen in this system lets anyone atLive in Flow open your accounts, transactions or documents.

Finance Fire: the figures you enter

I'm choosing to enter financial information here: passive income, active legacy income, life insurance benefit, liquid assets, real estate equity, how I want a distribution split across categories, which of the four Finance pillars matters most to me and why, and which steps on each pillar's path I have checked off, with the date and any note I add. Who can see it: nobody at Fathers on Fire can view these figures through any screen in this system. There is no coach view and no admin view. Fathers on Fire does not read the database to look at member financial records. If the law ever compels disclosure, I will be told unless telling me is prohibited. Where it lives: encrypted, under a key used for nothing else in this system, in a database hosted by Neon in the United States. How long: as long as I keep it. Nothing here deletes on a timer. I can delete any year's figures myself from the Dynasty Trust screen and any pillar answer from the Finance Fire screen, I can untick any path step or clear its note where the path is shown, and I can turn this off at any time from the Finance Fire screen, which stops anything new being saved. This tool is a framework for thinking through a distribution plan. It is not legal, tax, or investment advice, and it has not been reviewed by a CPA or attorney.

Reading your statements and returns with an AI

I'm choosing to upload my own financial statements or tax returns so an AI can review them and draft questions for my bookkeeper or CPA. What happens to the files: each file is read into memory once, sent to the AI provider whose key I save here (Anthropic), and then dropped. Fathers on Fire does not store the files. The provider handles them under its own terms for API use. Tax returns carry Social Security numbers and account numbers, and I can black those out before uploading because the review does not need them. What is kept: the findings and the email draft from each review, plus the file names, encrypted under a key used for nothing else in this system, in a database hosted by Neon in the United States. Nobody at Fathers on Fire can open them through any screen. I can delete any review myself, and I can turn this off from the Finance Fire screen. Whose key: mine. The review runs on my own AI key, so the provider bills me, not Fathers on Fire. What this is not: the review is an AI reading documents. It can be wrong, it is not tax, legal or accounting advice, it has not been reviewed by a CPA or attorney, and nothing it says should be acted on without my own licensed professional agreeing.

The Money Command Center: connecting accounts

I'm choosing to connect my bank, card, loan and investment accounts to the Money Command Center, or to upload statements from them. What is collected: the names of my accounts and the last four digits of each, their balances, every transaction with its date, amount, merchant and description, my loan balances, rates and minimum payments, and my investment holdings where the institution shares them. Also what I add on top: categories, notes, tags, budgets, goals, rules, and anything I type to the advisor. Who can see it: me, and anyone I later invite with a share I can end at any time. Nobody at Fathers on Fire can view my accounts or transactions through any screen in this system. There is no coach view and no admin view. The owner's screens show only whether a connection is working and how many rows it holds, never a transaction. Fathers on Fire does not read the database to look at member financial records. If the law ever compels disclosure, I will be told unless telling me is prohibited. Plaid's role: the connection to my bank is made by Plaid, a separate company. I sign in to my bank inside Plaid's window, and Fathers on Fire never sees my bank password. Plaid handles my data under its own End User Privacy Policy, which is shown to me when I connect. Fathers on Fire holds the key Plaid issues for my connection encrypted, under a key used for nothing else in this system, in a database hosted by Neon in the United States, and uses it only to pull my accounts and transactions. What this is not: the Money Command Center is education. It sorts my money, shows me the numbers and the formulas behind every score and every suggestion, and lets me run what-ifs. It is not legal, tax, or investment advice, it has not been reviewed by a CPA or attorney, and my attorney and CPA decide. How I stop: I can disconnect any account from the Accounts screen, which tells Plaid to drop the connection. I can delete everything the Money Command Center holds about me in one step from the Household screen, which also tells Plaid to drop every connection, and I get a receipt with the counts. I can export all of it first. I can turn this consent off at any time from the same screen, which stops anything new being collected.

The Money Advisor: questions sent to an AI model

I'm choosing to let the Advisor's Ask answer my questions with an AI model. What gets sent: the question I type, the conversation so far, and the figures the Advisor works out from my own accounts to answer it. Those figures include totals, ratios, my category names and my account names with only the last two digits of an account number. When a question needs my transactions, they include the merchant name, the amount and the date of the transactions it looks at. Memos, notes, tags and bank logins are never sent. Where it goes: to Anthropic, the company that makes the Claude model, using the Anthropic AI key I saved on the Finance page. If I have not saved a key, nothing is sent and the Advisor answers from its fixed rules instead. If I invite a partner and she asks the Advisor about our household, her questions are sent the same way, under my key, and the cost is mine. Fathers on Fire is not a party to my relationship with Anthropic, and what Anthropic keeps, and for how long, is governed by its terms and not by this system's. This part cannot be undone: once a question has been sent, that transmission has happened. Deleting my money data here does not reach into Anthropic's systems and does not recall what was already sent. What this is not: the Advisor is education and not legal, tax or investment advice. The model never does the math: every number it shows was worked out by this system and can be traced. How I stop: I can turn this off at any time on the Ask screen. Questions then go back to the Advisor's fixed rules and nothing more is sent. Turning it off does not touch my connected accounts.

The Family Office document vault

I'm choosing to store my own documents in the AI Family Office vault: tax returns and the forms behind them, statements, receipts, wills, trusts, powers of attorney, insurance policies, deeds and entity papers, whichever ones I upload. What is stored: each file, the text read out of it, the line items I confirm, and a private map that ties each masked identifier back to the real one. All of it is encrypted under a key used for nothing else in this system, in a database hosted by Neon in the United States. The findings, questions and packets the seats build from my documents are kept too, encrypted under the Finance Fire key. Who can see it: me. Nobody at Fathers on Fire can view my documents or findings through any screen in this system, and there is no coach view and no admin view. If I give my spouse a partner grant, my spouse can read my vault and my binder, and cannot change or delete anything. If the law ever compels disclosure, I will be told unless telling me is prohibited. What it is used for: only to run the Family Office seats for me and to build the packets I choose to send. Fathers on Fire will not use my tax return information for marketing, will not sell it, and will not share it with anyone except a professional I name myself. How long: as long as I keep it, unless I set a shorter time. I can delete any document, or everything in the Family Office at once, at any time from the Family Office screens. Deleting is never blocked and leaves a receipt that lists counts, not contents. My choice: I do not have to agree to use the rest of Fathers on Fire. I can withdraw this at any time from the Family Office settings, which stops anything new being stored. I can download this wording and the date I agreed from the same screen. What this is not: the Family Office lists questions for my own CPA, attorney and other licensed professionals. It has not been reviewed by a CPA or attorney, it can be wrong, and it is not tax, legal, accounting, insurance or investment advice.

The Family Office and the AI provider

I'm choosing to let the AI Family Office send text read from my stored documents to the AI provider whose key I save here (Anthropic), so it can find the line items on my returns and word the seats' questions for my professionals. What is sent: redacted text only. Before any AI reads a page, Social Security numbers, taxpayer and employer identification numbers, bank account and routing numbers and similar identifiers are replaced with masked tokens. The file itself is never sent. Numbers in the findings are computed by software, not by the AI. Who receives it: the AI provider of my own key, under its own terms for API use, which I can read in my own provider console. Fathers on Fire keeps a fingerprint and the token count of each request, never the text sent. Whose key: mine. The provider bills me, not Fathers on Fire. Without a saved key, the checks and math still run and nothing is sent. How long: until I withdraw it. I can withdraw this at any time from the Family Office settings, which stops anything new being sent. Text already sent stays with the provider under its terms. I do not have to agree to use the vault or the rest of Fathers on Fire. What this is not: an AI reading documents can be wrong. Nothing it writes is tax, legal, accounting, insurance or investment advice, and nothing should be acted on until my own licensed professional agrees.

Sending packets to your professionals

I'm choosing to send Family Office packets to professionals I name myself, such as my CPA, my attorney, my bookkeeper or my insurance agent. What a packet holds: the findings and questions for that professional, the math behind each estimate, the return lines it refers to, and identifiers masked to the last four digits. It never holds a full Social Security, taxpayer, employer, account or routing number. How it travels: through a private link to that one professional, which expires after 30 days, can carry a passcode I deliver separately, is read only, and which I can revoke at any time. Every open is logged and I can see when it was opened. The professional may reply on the same link. Who receives it: only the person I name for each packet. A professional listed in the Fathers on Fire directory receives nothing until I send a packet to that person myself. Fathers on Fire receives no referral fee for a packet. How long: each link lasts until it expires or I revoke it. I can withdraw this consent at any time from the Family Office settings, which stops any new packet being sent. What a professional already downloaded stays with that professional. What this is not: the packet is software output for a licensed professional to confirm. It is not tax, legal, accounting, insurance or investment advice.

Plaid, who connects your bank

When you connect a bank, card, loan or investment account, the connection is made by Plaid Inc., a separate company that acts for you and for us in moving the data. You sign in to your bank inside Plaid’s window and Live in Flow never sees your bank password. Plaid receives what it needs to make the connection and pass your accounts and transactions back, and handles that data under its own End User Privacy Policy, which is shown to you when you connect and is here: plaid.com/legal. The key Plaid issues for your connection is kept encrypted here, under a key used for nothing else, and is used only to pull your accounts and transactions. Disconnecting an account, or deleting everything from the Household screen, tells Plaid to drop the connection. What Plaid already holds is governed by its policy, and this system cannot recall it.

Anthropic, the AI model behind the Advisor and the document reading

Where you have agreed to it, text you type to the Money Advisor, the figures it works out for the answer (including merchant names, amounts and dates when a question needs them) and text read from documents you chose to have read go to Anthropic, the company that makes the Claude model. It goes under the Anthropic key you saved, so the account is yours and Anthropic’s terms govern what it keeps and for how long. If you have not saved a key, nothing is sent. If you invite a partner, her questions to the Advisor are sent the same way under your key. Deleting your records here does not reach into Anthropic’s systems and cannot recall what was already sent.

Text messages and your mobile number

Your mobile number and your agreement to receive texts are used only to send the texts you turned on. No mobile information is sold, rented or shared with third parties or affiliates for marketing or promotional purposes. Text messaging opt-in data and consent are never shared with anyone. The only company that sees your number is Twilio, which delivers the texts. There are two kinds. A win text says the child beat a boss or set a new best, and never mentions anything he finds hard. A daily lesson report, sent only to a parent who agreed to the wording that names it, says after each lesson day what went well, what was hard and what to work on next. So a report does name, in plain words, a skill the child found hard that day; that is what it is for, and a parent who has not agreed to reports never gets one. No text carries his real name, a retell or session score, or anything he said or wrote, and the words of a text are not kept here. At most three texts a day per number, win texts and reports together, at most two of them reports, and none between 8 at night and 8 in the morning; a parent who agreed to win texts only gets at most one a day. Message and data rates may apply. Reply STOP to stop or HELP for help. A parent’s own number is also held until that phone replies YES to the one text it is sent when she types it; nothing else goes to it before. If a parent adds the other parent’s mobile, that number is stored encrypted and gets one text asking the person to reply YES. Nothing else is sent to it until they do, they can reply STOP at any time, and the parent who added it can remove it. A YES to win texts is not a YES to reports: that number gets reports only after its own YES to them. To make sure a number is never asked twice in a day, a keyed hash of it and the time it was asked are kept for one day, then deleted. A reply to our number is matched to a stored number by a keyed hash, never by reading numbers in the clear, and the number itself is never written to a log. Full terms: text message terms.

Questions, and getting data out

Questions about your data: reply to any email we send you. If you want a copy of everything held about your child rather than a deletion, ask, and you will get it.

Back to Kid’s Flow Academy