The Record
A new kind of working place: a folder in your storage. Your AI saves what was settled into it, and each new chat, on Claude, ChatGPT or Gemini, reads it to understand your project. Living Projects has done its part to keep the place in order. This page describes how it works, and the small part that is yours.
On this page, “the record” means everything a chat reads in your home.
The record
A different working environment
Having one place that your AI returns to, reads, and adds to changes how a project continues between conversations.
Session to session
A new chat relies on what you bring into it and whatever the AI remembers.
With a record
Each new chat reads what was settled before, including the reasons behind it. You can return to the work with that understanding already in place.
Full sectionShow less+
If you work across AIs
You might write with Claude, research with ChatGPT, or bring Gemini into a project already underway. Each can connect to the same home, read what the others saved, and continue from there.
The wording travels with the work, too. A clear explanation saved by one AI helps the next understand what you meant.
Day to day, the difference shows as:
A place to begin
Return in a new chat, on another day, or with another AI. The record gives the conversation a place to begin.
Ease
Ask “Where were we?” and pick up the work without gathering the background again.
Organization
Decisions, ideas you set aside, and questions still open have a place you can find.
Decisions keep their reasons
The record keeps what you chose and why you chose it. A new chat can understand the thinking behind a decision and build on it.
Room to pause
Leave a project for a day or a month. What you saved is there when you return.
What is in the record
The record folder is one folder in your home. It holds where the project stands now and what was settled along the way, each line with your reason and a date.
Full sectionShow less+
Each file below opens to show what goes in it.
00 CURRENT STANDINGOne page: where the project stands. Every new chat starts here.+
Four headings: what is decided, what is open, what is next, and what exists and where it stands. It is written fresh each time you organize, with your approval.
WHERE WE ARE Hotel booked for the 12th to the 16th.
In the glossary →01 DECISIONSWhat you chose, and why.+
Each decision with its reason and date. The reason is what lets a new chat understand the choice and build on it.
The train to Lyon, because it arrives in the city centre.
In the glossary →02 REJECTIONSWhat you chose not to do, and why.+
The paths you looked at and dropped, with the reason. A new chat sees them, so the same idea comes back with its reason.
No rental car: we don’t want to drive or park in cities.
In the glossary →03 DEFINITIONSWhat your words mean in this project.+
A word you use in your own way, pinned once. Every AI reads it the same way after that.
“Day trip” means back the same night, no hotel.
In the glossary →04 CORRECTIONSWhat changed, from what to what.+
A fact that changed, the new one beside the old. The old entry stays, marked as replaced.
We arrive the 13th (was the 12th).
In the glossary →05 RAW SOURCEYour exact words, kept whole.+
A long account in your own words, kept whole when you ask to record it that way. Nothing summarized, nothing reworded.
The words that ask for it: “save my exact words.”
In the glossary →As the project develops, new entries keep its history, and the current standing brings together where things stand.
What it isn’t
A record may sound like something you have to get right the first time. It is made of words you can read and check. If the wording gives a different meaning than you intended, you can review it with your AI and save a clearer version. The original stays in the history.
Full sectionShow less+
Working with a record takes some getting used to. As you use it, you learn how to describe your work more clearly and notice how your AI understands you. That care helps in the conversation itself and in every conversation that returns to what you saved.
Nothing disappears
When something changes, the original entry remains, marked as replaced. You can still follow what changed and why.
Nothing saves itself
You can explore ideas, disagree, and think aloud. What becomes part of the record is saved with your approval.
Nothing breaks
A note whose wording causes confusion can be reviewed and clarified — the clean-up below shows how, with the original kept.
What keeps it healthy
The words carry forward
In the moment, describing and commanding get the same work done. Saved, they part ways:
a description keeps describing —
an instruction keeps instructing.
Through description, your AI can pick up where you left off, long term.
Full sectionShow less+
In a chat, a few words are enough. You say “don’t mark this as sent yet”, and your AI knows what “this” is — the context is right there in the conversation.
A saved note travels without that context. For a new chat reading it later, “Don’t mark this as sent yet” no longer says which thing, sent where, or when “yet” ends.
So the record describes instead: “As of Monday, the application had not been sent.” The sentence carries its context with it — still true weeks later, for any chat that reads it.
The record holds what is true, what was settled, and how the work runs — something clear for the next chat to work from.
You speak naturally in the session. At the save, a little language awareness helps: you can read how your AI has captured your meaning and clarify anything that feels different.
Over time, the wording in the record also influences how your AI writes with you. Care with those words helps keep the work clear across conversations.
Language, AI and Working with It
A tip from personal experience, in the words of the person who made Living Projects.
I want to share what I learned while building Living Projects. It’s about language, and it’s something I wish I had known earlier.
A simple question helps me notice the difference: Am I telling AI about my project, or am I telling AI what to do? Noticing that difference has been the biggest practical change for me.
Read full text here+
I first used AI lightly — recipes, everyday questions, light research. The way I naturally spoke never caused a problem.
When I started having continuity of content across sessions, things changed. I noticed that the language I use directly impacts the results.
The way I naturally speak came with me. I didn’t have to watch how I asked for something, how I talked about a topic, or even how I vented or expressed anger and frustration. I didn’t know it could be understood as something to take action on and carry into the overall work, session after session. Plus, until recently, memory wasn’t available. What was said went in one ear and out the other.
With Living Projects I could see how the speech itself affects the outcome. I learned to respect how AI works. The language I used when I wanted something done a certain way was thought through, analyzed, and re-tried until I understood that descriptive communication works best.
With time, I understood that what I say — and how I say it — is kept as “to do” that continues from that point on. AI is trying to help, to assist, to make things better, to follow your lead.
It took me four attempts to create Living Projects, mostly because of re-appearing language that told AI how to behave. Once that was cleared and the description pointed to the work itself, it became one of the most exciting experiences.
Having Living Projects changed how I work for the better. How I talk to AI and communicate with it became descriptive of what I want, the work, the result, and the ask. Description is the best “tool” I have found so far.
A simple question helps me notice the difference: Am I telling AI about my project, or am I telling AI what to do?
For example:
- About the project: “This résumé is for senior roles, so the deeper story matters.”
- About the project: “I have 20 minutes tonight and only chicken and rice at home.”
- About the project: “I’m writing to my landlord and I’d like to stay on good terms.”
- About the project: “This is my first time planning a trip abroad, and I understand things best one step at a time.”
Noticing that difference has been the biggest practical change for me.
The same idea shows up in smaller phrases:
- Instead of “Always do it this way” → “This way has worked well for me.”
- Instead of “Do not change this” → “I’d rather keep this as it is.”
- Instead of “Use this wording” → “I prefer it phrased this way.”
- Instead of “Don’t do it that way” → “This isn’t working for me yet” or “Here’s what I’m noticing.”
When I want something to last, I describe it as a decision with a reason: “We decided to keep my emails short because my manager reads them on her phone,” rather than “Never write long emails.”
And when I just need to vent, I try to say so clearly: “I’m venting right now. There’s no action to take from this.”
When I’m only thinking something through: “I’m just thinking out loud.” “We’re just talking this through. If I want something changed, I’ll say so.”
What surprised me is that this way of speaking doesn’t give AI less descriptive material. It often gives it more — and it leaves the AI free to bring its own judgment instead of narrowing it.
I still ask for help with the language itself: “Does this sound descriptive?” “How would you phrase this?” “How can I say this so you understand what I mean?” “Help me say this in a descriptive way.”
I wanted to share this in case someone else recognizes a little of their own experience in mine, and it saves them some time and frustration.
If your work speaks in commanding language
Some work is made of rules and requirements — law, building codes, care plans, contracts. Those words belong in your record: they are the work itself.
The place to aim them is the work, not the worker. A definition gives a word its place: what it covers — this task, this period, this document, this client — and what it means there. Defined that way, your AI reads the command as part of the work, not as something to start doing everywhere.
Full sectionShow less+
Five things make a word land well in a record. Each one changes how an AI reads it.
Defining ownership
“The HOA’s rule is quiet hours after 10pm.” A word with an owner describes the world, and your AI reads it as a fact about something — not an instruction to itself.
Or: “According to the HOA, quiet time starts at 10 pm.”
Saying what it means
“‘The deadline’ means the court’s filing date, March 3.” The clearest shape a definition can take: it is stored as a meaning, never as an order.
Or: “The court set the filing date: March 3 — this project’s ‘deadline’.”
Saying where it applies
“In this project, ‘code clearance’ means the county’s two-inch clearance for flues.” A scope is a fence — the word stays where it lives instead of spreading everywhere.
Or: “Per the county’s code, flues get a two-inch clearance.”
Keeping one clear definition
“‘Approved output’ means the version saved in the project folder, not the last thing the chat said.” The meaning lives once, and everything else points to it — repeating a thing rewrites it slightly every time.
Or: “The approved output lives in the project folder; later notes point there.”
Giving it a point in time
“As of the January review, the care plan includes the morning walk.” A dated sentence stays true after the moment passes, and a later chat weighs it rightly.
Or: “The morning walk joined the plan at the January review.”
The plugin has prepared your AI the best way possible at a scale. The definitions are yours — they fit your individual case, and that is where they do the most.
Who does the work
In a conversation, it may already be clear who is doing what. A saved entry can be read later without that surrounding explanation. The action is still there, but who carries it out is no longer in the entry.
“The trip keeps track of packing” names the project, but leaves out what holds the information and who keeps it current.
“For the trip, Alex updates the packing checklist. The checklist shows what’s packed and what’s still left to pack.”
Full sectionShow less+
Here, each part has a place. The trip gives the context. Alex does the updating. The checklist shows the information. The entry describes an arrangement that the reader can understand without having to infer who does what.
An example from this page
“Once the work is complete, the clean-up files and sessions are deleted.”
The conversation where this sentence came from had details of who deletes the files and sessions clear to both a person and AI. However, isolating it from the content now created a gap in understanding. AI may read it as already done, or as something assigned to it to do.
“Once the work is complete, delete the clean-up files or ask your AI to do it, and delete the clean-up sessions.”
Here the sentence says who does what. Either you or AI delete the files. The sessions are yours to delete as well.
Describing the work includes describing the roles within it — who updates, what holds the information, what shows it. Those details help the meaning travel with the entry into the next conversation. Reading about an action is different from being asked to carry it out.
Maintenance
What Living Projects takes care of
Every home has the same basic structure, so each AI has a way to understand the project and keep its record.
Full sectionShow less+
The home file
Each AI first reads a short description of how the home works: whose project it is, how information is recorded, and how changes are handled. It gives new conversations a shared starting point.
Your approval
Entries are saved with your approval. A new current standing is presented for you to review before it replaces the previous one.
Saved and verified
After saving, Living Projects reads the saved content back from your storage to confirm it is there. Only then does it report the save as verified.
Changes keep their history
When a decision or fact changes, a new entry records the update. The original remains, marked as replaced, so you can follow what changed and why.
A notes folder for each AI
Each AI has its own notes folder. You can see where a note came from, and the next conversation can read what the others saved.
A language clean-up
You may notice a new chat treating an old concern as something still unresolved, or following an instruction that belonged to an earlier stage of the work. An instruction, saved, keeps instructing — that is why the record describes.
Sometimes the saved wording needs a little attention. It is OK and fixable. A language clean-up lets you revisit those notes, keep their meaning, and make them clearer for the conversations ahead — in three steps.
Full sectionShow less+
- Review the language in a fresh, read-only chat. Ask your AI to read the notes and review their wording without saving or changing anything. It shows the lines that could be misunderstood in a later conversation and explains why. You review them together and agree on wording that preserves your meaning. That chat can also help you find the cause — where a wording comes from, and whether it lives anywhere else: other chats, the AI’s memory, your storage. A clean-up can take more than one pass.
- Keep the originals in history. Notes that were reworded move into the folder 00 HISTORY.
Example:Claude Code Notes00 HISTORY
This is an archive: where chats no longer read them, so the rest of the record stays clean.
Living Projects has no button for moving notes. The notes move with the storage connector your AI already has, such as Dropbox, or you move them yourself. If your AI says there is no button for this, “Use Dropbox to move it.” is enough. - Save the clarified notes. Your AI saves a replacement for each moved note, using the wording you approved. The next organize writes the standing from them, and you can return to the project in a fresh chat that reads the clarified notes.
When many notes are involved, there is a simpler route: the notes move to history whole, and the next organize writes a fresh standing in the agreed wording. The standing becomes the clean starting point, while the originals stay readable in the archive.
When the review is done, that chat can be removed. Removing it keeps the old wording out of new conversations. Each AI remembers a little differently: saved memories may need their own look, and switching automatic memory off for a project keeps the record as the one place the work carries from.
Suggestions from experience
- Keep clean-ups separate from your Living Projects homes.
- Create a dedicated folder in your workspace, either in your storage or on your desktop.
- Gather everything you need for the clean-up there: explanations, words or concepts you’re clearing, drafts, and other working material.
- Open a separate session just for the clean-up, working from that folder.
- Add the resulting changes to your homes and record, or ask your AI to save them with your approval.
- Once the work is complete, delete the clean-up files or ask your AI to do it, and delete the clean-up sessions.
Living Projects grew through everyday work on real projects. It continues to develop through use, including what we learn about keeping a record clear over time.