Nothing to sign up for. Nothing gets uploaded.
A file you drag in
becomes working software.
It gets an address you can send. It saves what people type into it. It keeps a record of who changed what and when. Nothing inside the file changes, and nothing in it was built for any of this.
Drop it here
It stays in your browser and nothing inside it is run. This is just so you can see what happens.
Choose a file or try one of mineEach of those becomes somewhere answers are held, and a column when you pull the results out. Send the link to five people and you get five sets of answers back, with names on them.
Why you would bother
You asked for a tool and got a drawing of one.
You open Claude, or ChatGPT, or Gemini, and you ask for a five day trip to Lisbon in October for two people, with a budget you can tick off as you book it. Thirty seconds later you have something that looks exactly like this.
Here it is twice. Read the two requests first. They are the same sentence apart from the last line, they took the same thirty seconds, and both results were designed to look like software. One of them is. Touch them both and find the difference.
What you typed 1 message
Plan a five day trip to Lisbon in October for two of us. Give me a budget I can tick off as I book each thing, and a day by day plan.
What you typed 1 message
Plan a five day trip to Lisbon in October for two of us. Give me a budget I can tick off as I book each thing, and a day by day plan. Build it as one self contained HTML file rather than a document.
The left one is a picture of a tool. Every control is drawn and none is wired. To change one number you go back to the model and make a second file, and the first one is already in somebody's inbox being read as true.
The right one is the same drawing, wired. Tick the flights and the total drops and the bar moves. Switch to Days and the trip lays itself out. Filter, clear, start again. Nothing was regenerated and the model was never asked twice.
PDF is still right for anything that has to be identical forever and provably unchanged. Contracts, tickets, receipts, anything going into a record. The mistake is reaching for the archive format when what you wanted was the working one.
And this is still not the ceiling
The one on the right only knew what you told it.
Both of those were built from whatever was already in your head. Nothing in either one had ever heard of Lisbon. You do not go and find the fares yourself and paste them in. You tell it to look, and it looks before it writes a single line. What it finds ends up inside the file.
Here is the same trip after two more messages. Not a longer prompt, a short conversation. The same thirty seconds of typing and about a minute more of waiting, and this is the only one of the three that can disagree with you.
What you typed Turn 1
Plan a five day trip to Lisbon in October for two of us. Give me a budget I can tick off as I book each thing, and a day by day plan. Build it as one self contained HTML file rather than a document.
What you typed Turn 2
Go and look all of this up before you build it. Real fares from Los Angeles for those dates, the opening hours and closing days for everything on the list, and today's dollar to euro rate. Do not ask me for any of it. Put what you find in one DATA object at the top of the file with a source and a date on each fact, and build the page from that object rather than typing numbers into it.
What you typed Turn 3
Now check my plan against what you found and give me a budget field. If anything I scheduled falls on a day it is closed, warn me at the top and give me a button that moves it. Let me tick things off as I book them and show me where I am landing against my budget as I go.
Flights, LAX to LIS Three fares found
Everything else Hours checked
The model went and found all of this before it wrote the page, then stored it in the file. Nobody typed a fare in. It was true on 6 September and it will go stale. When it does you ask for the file again and replace it: the link does not change and the old version is kept.
The model already knows how. Nobody tells it to.
You just read the three requests that produced the three artifacts above. Here they are as one block you can keep. None of it is technical. Paste it and change the subject to whatever you actually need.
Do not write me a report or a document. Build a single self contained HTML file that does it. Go and find the facts yourself first. Do not ask me for them and do not estimate. Before you write any code, look up the real prices, opening hours, closing days and dates this needs. Put everything you find in one JavaScript object at the top of the file called DATA, each fact with a source and the date you found it. Then build the page from DATA, never from values typed into the markup. Use what you looked up. If anything I planned falls on a day it is closed, or breaks a rule you found, show me a warning at the top of the page and offer the fix. Keep it self contained. No CDN links, no external fonts, no images from the web, no network calls. Everything inline, so it works with no connection. Let me put my own numbers in. Give me a field for my budget and anything else that is mine to decide, and recalculate everything against it as I go. Name the inputs. Give every input, select and text area a data-field attribute with a short lowercase name, so the answers come out labelled. Keep the state in the controls. Anything I choose or tick should live in a real input, select, checkbox or radio, not in a variable, so it is still there when I come back.
Why the DATA object matters
A file with numbers typed into the markup is a drawing again, just an interactive one. Facts in one object at the top is the difference between a page that shows an answer and a page that holds the research. The more you send it looking before it builds, the more the page can work out on its own afterwards, without you or it going back for anything.
Why self contained is not optional
Most generated pages reach out for a chart library or a font. The day that link moves, or someone opens it without a connection, the page half loads and looks broken. Every file that ever mysteriously stopped working did this.
The one that catches mistakes
Asking it to check what it looked up against what you planned is the sentence nobody writes, and it is the one that produced the warning above. A document cannot tell you that you scheduled something on a closed day. A page can.
The first two you can email. The third one you cannot.
A document that reports you something has nothing to lose. A document you operate makes something every time you use it, and a file has nowhere to keep it. Set the budget, tick four things off, close the tab, and it was never there.
It has to remember
The ticking is the work. Come back on Thursday and it is where you left it, on your phone as well as the machine you started on.
That is not a convenience. It is the difference between a tool and a thing you fill in again every time.
Other people operate it too
Send the link to four people and open it yourself. Their answers arrive within about three seconds, and the box someone is typing in gets an outline in their colour with their name above it.
Email it instead and you have four copies going four different ways. One address is one truth.
It goes out of date, and that is fine
Anything the model looked up was true the day it looked. Ask for the file again and put the new one in the same place: the link does not move and the old version is kept.
Every value also keeps its own history with an author and a time, and the whole thing pulls out as a spreadsheet, one row per person.
Most of what you generate does not deserve infrastructure. A form for this week, a plan for one trip, a mockup you throw out on Thursday. Standing up hosting for those is absurd, so people email them and they arrive broken. This is where those live, and most of them should live here forever.
2eightz keeps its own work here
Including the things that did not work, which are still up with the reason attached.