What shipped
Newest first. Short entries — what you can now do, and whether it fully works yet. The longer stories about how any of it went live on the blog, and the how-does-this-work pages are in help.
-
Past forty plants, your oldest and your newest plant still keep their own line
When Rafe sees a big garden bed by bed, two plants keep a line of their own above the list: the one that has been in the ground longest and the one you planted most recently. The tree you have had for years and this spring’s sowing are the two a gardener asks about first, and a bed line saying “apple” is not the same as your tree by its name. A check that runs the whole suite on a later calendar day found they could drop out of view — their names had been riding along in recent journal entries, which a later day empties — so now they are named on any day.
-
Your plant settings stop running away
On a plant’s page, the two settings you occasionally need to fix — how it started, and where it lives — used to sit below the journal. That meant every time you logged watering or a harvest, they moved further down the page. On a brand new plant they were about five screens down; on one you had used for a season, closer to twelve, and nothing stopped it getting worse. They now sit above the journal, so they are in the same place on every plant no matter how much history it has. Your story of the plant is still there, just below them.
-
A photo is dated when you took it
Pictures were being filed under the moment you picked them, not the moment you took them. Take a photo on Sunday, send it on Wednesday, and your plant history said Wednesday. That matters more here than in a normal photo app: the dates are what growth comparisons and the season chart are built on. Now the app reads the date from the picture itself when it has one. It will not accept a date in the future, or one from before this app existed, because a phone with its clock set wrong would otherwise drag everything that reads it. When there is no usable date it falls back to now, which is what it always did.
-
A photo now keeps the plant you named
In the chat you can tap + and pick one of your plants before you ask a question. If you did that and sent a photo, the answer came back about the right plant — but the photo itself was filed against nothing. It did not appear in that plant's run of pictures, and the app never compared it to your earlier shots of the same plant, because as far as the record was concerned it was a picture of no particular thing. From today it is filed to the plant you picked, dated the day you took it, and it follows you to another phone. Photos taken from a plant's own page always worked this way and are unchanged. And nothing is guessed: if you did not say which plant, the picture stays exactly where it was. This was one field, and it turned out to be the reason behind several things that had looked like separate faults.
-
How tall is that? Now actually answered
Send a photo of a plant and, when the picture shows enough to judge it, you now get its height offered back — one tap to confirm, and it goes on the plant. It reads in feet and inches and it says about, because it is an estimate from a photograph rather than a tape measure. This was described here before it worked: the app knew how to record a height and what to do with one, but never asked for it in a way that produced an answer, so it quietly never happened. It does now. Just as important, it stays quiet when it should. A picture taken looking straight down into a pot cannot show how tall anything is, and for those you get no height rather than a guess — often with a note that a closer or wider shot would tell us more. If you want a height, stand back and take the picture from the side, with something ordinary in frame for scale: the pot, a hand, a boot. Age we still will not guess, and that is still on purpose.
-
It asks about more of what it saw
When the app looks at a photo it writes down several things: what it can see on the plant, how far along the plant is, and how it looks. Only the first of those three ever made it onto your plant page. The other two were being written down and then shown to nobody. Now all three appear in the same place, at the top of Right now, each with a yes and a no. It only asks about things you can settle by looking at the plant. It will not ask you to confirm which crop something is or how big the pot is, because those are not questions you answer by looking — those get filled in as blanks instead, and never overwrite what you already put there.
-
He can ask which bed now
Tell Rafe you planted something without saying where, and he will not guess — that part is deliberate, because a plant filed in the wrong bed is worse than one he asked about. Until now he did not ask either. He said he had not written anything down and left you to start again. Now he lists your beds and you tap one. He only does this when the answer is a real choice: if you have one bed there is nothing to ask, and if two of your beds have the same name he says nothing rather than offering two buttons you cannot tell apart. Naming the bed in your message still works and is still the fastest way.
-
It can tell your two sowings apart
If you sow the same crop twice in one bed — two weeks of lettuce, a second row of radishes — the app used to give up entirely. Say "the radishes are up" and it would tell you it had not written that down and offer you no way to fix it, because it only knew how to tell plants apart by their name and their bed. Both are the same for successive sowings. Now it uses the date you sowed them: it asks whether you meant the one sown Aug 14 or the one sown Aug 17, and you tap. It still says nothing rather than guessing when two really are identical — same crop, same bed, same day — because two buttons you cannot tell apart are worse than being asked to say it again.
-
He stops asking for a photo he already has
Ask Rafe about one of your plants and he gets that plant's record — when it went in, what you have logged, what it has been told since. Photographs were not in that handful, which meant he could tell you the planting date and the last note and then ask you to go outside and take a picture of a plant he had three pictures of. He now knows which photos each plant has and when each one was taken, and he is told not to ask you for one he is holding. One honest limit while we finish the other half: knowing a photo exists is not the same as looking at it. For Rafe to actually read a picture it still has to be attached to your message. What changes today is that he should stop sending you out for a photograph we already had.
-
He shows you which of your notes he used
Ask Rafe about a plant and, when the answer leans on something you wrote down, you now get those entries listed underneath it with their dates: "It used what you had written down". That list is how you tell an answer about your garden from an answer about the crop in general — both are useful, and they are not the same thing. Being straight about what changed: the app has been able to show this for a while and never had anything to show. Rafe was reading your notes and using them accurately, and simply never said which ones, so the list was always empty. He is not any better at reading your journal today than it was yesterday — he just stopped keeping that to himself. One thing worth knowing: a note written on a device that has not synced yet will not appear in the list, so a short list is not a sign he ignored you.
-
It stops asking about growth your plant is past
The app sometimes asks you to confirm what it thought it saw in a photo. Once you have told it your plant is fruiting, being asked whether you can see its first true leaves is a question about a plant that no longer exists. Those now drop off on their own. It only counts what you have actually confirmed, never what our calendar expects — if a plant is running behind, your answer is the thing that tells us, and going quiet would throw that away. So it does very little at first and more as you answer things, which is the way round we want it.
-
It stops asking which plant when it already knows
If you attach a plant to the chat and then send a photo, that photo now goes straight onto that plant. Until today it asked you which plant the picture was of — and offered every plant in your garden, including the fruit trees, when you had told it which one two messages earlier. The question itself is still there and still useful, because most photos arrive with nothing attached. It just no longer asks a question you have already answered. The note it writes on the plant uses the words you typed, which is why the filing happens when you send rather than when you take the picture. If you attach nothing, it still asks rather than guessing — putting a photo on the wrong plant is worse than asking.
-
Keep a good answer on the plant
Ask Rafe about a plant and you get an answer that is often worth having again next year — and until now it only lived in the chat. There is a button under the reply now: keep this on your Sungold. One tap and it goes on that plant's page, dated today, under Worth remembering, marked as something the app wrote rather than something you did. Next August, when the same lower leaves look alarming again, the note is right there on the plant instead of buried in a conversation you would have to find first. It only appears when we already know which plant you meant, and ignoring it costs nothing — the answer stays in the chat either way.
-
Long answers read like writing now
When Rafe gives you a long answer with sections in it, those sections now look like sections. A heading was set in exactly the same size, weight and color as a bolded phrase inside a sentence — so the only thing marking it out was a little space above it, and while you were scanning it did not read as a heading at all. It is a touch heavier now. Paragraphs also had less than half a line between them, which ran them together into a block; there is a bit more air there. Nothing shouts: a heading in a four-paragraph answer should be a signpost, not a title page. We checked all of it on a phone-sized screen rather than a laptop, because that is where you read it.
-
Nothing is true because a computer said so
When the app looks at a photo from your garden it writes down what it thinks it can see — first blooms, pods filling, lower leaves going yellow. It has been doing that for a while. What it had no way of doing was asking you whether it was right, so everything it saw sat in the record unused, and the oldest had been sitting there 28 days. That was not caution on our part, it was a missing screen. Now it turns up on the plant's own page under Right now, in plain words, with the reason it gave — and two buttons. Saying no is worth as much as saying yes: a yes lets the app use what it saw, a no means it will not come back and tells us we misread a picture, which is the most useful thing we can learn about how well any of this works. Nothing is used until you answer, and leaving it alone changes nothing.
-
Old questions stop being asked
When the app looks at a photo it sometimes asks you to confirm what it thought it saw. Those questions had no expiry, so one from four weeks ago sat on the plant page next to one from this morning. Asking how a plant looked a month ago is not a question anyone can usefully answer. Now the ones about how far along a plant is, or how it looks, stop being asked after a few weeks. Questions about something the plant did — its first true leaves, its first flower — never expire, because those either happened or they did not, and a month later the answer is exactly the same. Nothing is deleted either way. An old reading is still the best thing we know; we just stop putting it in front of you as a question.
-
One question per thing, not one per photo
When the app thinks it can see something on a plant it asks you whether it is right. It was asking once per photograph, so a plant with three pictures of it could produce three identical questions — and answering one made a card disappear while an identical one was still underneath, which looks exactly like a card coming back. Now it asks once per thing it thinks it saw, and your answer settles every copy of that question at once. Saying no works the same way: you never have to reject the same thing twice. Nothing about the record changed — a thing you have already confirmed still stands on its own, so the history of a plant that really did change over a week is intact.
-
Rain counts on the jobs you actually see
A good soaking has counted as a watering since August, so the app would stop asking you to water. It turned out to be counting against the wrong list. There are two names for watering inside the app and six for the checks you actually get shown — feel the soil where you sowed, feel the mix in the tray, feel under the mulch — and the rain was only clearing the two you never see. So it rained, and the app still asked. It clears all of them now. Nothing changed about how much rain counts: about a third of an inch, added up over three days, and a plant under glass or in a tray indoors is still credited nothing, because none of it lands on them.
-
Rafe has a proper mark now
The little symbol on Rafe's replies was a single leaf and a seed, and Nik's verdict on it was "an indistinct squiggle". It is now a young plant in a pot — two leaves, a stem, and a couple of square pixels breaking off the rim. It is a pot rather than a bare sprig on purpose: the garden tab is a bare sprig, and two marks that mean different things should not look alike. Six rounds and about sixty alternatives went into this, several of them Nik's own ideas, and the first drawing beat all of them. The mark has since become the sun.
-
Rafe now hears which bed you said
Tell Rafe you planted something and he writes the plant down for you. Some of the time, though, he would answer "got it, three basil seedlings going into Herb Corner today" and then write nothing down at all. The reason was hidden: Rafe had worked out the bed while writing his reply, but the part that saves the plant was only allowed to look in one place, and that place was sometimes empty. Now it reads the bed from what you actually said. If you name one bed, the plant goes in that bed. If you name no bed, he still asks rather than guessing, which is the behavior we want to keep. We found this by using the app rather than by reading the code: the same sentence, sent seven times, worked five times.
-
Rafe says hello now
Open the Talk tab with nothing in it and it used to be a heading and a paragraph, both talking about the app in the third person. It is Rafe's screen, so Rafe opens it: his mark, his name, and one line — "What's going on out there?" Underneath, in his own voice rather than ours, what he actually has to go on: the date, what you planted, when it went in, and what you wrote down. The starter questions are unchanged. Nothing was added that we cannot back up — he tells you what he knows, not what he might guess.
-
Rafe stops handing you homework about his own findings
You send a photo of a chewed leaf. Rafe says something has been chewing it. Then a button appeared saying "Check it for pests and disease" — which reads as go and do the diagnosis, of the plant he had just looked at. What he actually meant was narrower: holes are a symptom, not a culprit, and plenty of things leave similar marks, so from one picture he could not say which. The button now says that instead — "I can see it but not what is causing it" — and offers to take the picture that would settle it. One tap opens the camera and the photo attaches to the same conversation. If your device has no camera it offers to let you go and look and come back, which is a real answer too. And when Rafe can name the pest outright, the button does not appear at all.
-
Rafe keeps notes. They are not your conversations
When you add a photo to a plant, Rafe looks at it without being asked and writes down what he saw. That record is how the app knows a picture has already been looked at, and it is what lets it compare one week to the next. It was also turning up in your list of chats, as a conversation you never had with entries you never wrote. It no longer does. Nothing was deleted and nothing changed about what Rafe notices — the notes are still kept, they are just filed as notes rather than sitting among your questions. The other half of the same fix: those notes were written in our own shorthand, so a note could read "first_true_leaves" instead of "first true leaves". Anywhere Rafe writes for you to read, he now uses words. Where he sees something he has no plain name for, he describes what he saw and leaves the label out rather than showing you a piece of our code.
-
It tells you which photo it is sending
Ask Rafe about several pictures at once and there is a wait before the answer starts. The screen used to say only "Thinking", which was true but not the useful part: photos go up one at a time, so most of that wait is the pictures leaving your phone. It now says which one — "Sending photo 2 of 3" — and changes as each finishes. On a slow connection in the garden that is the difference between a wait you can understand and one you cannot. We deliberately did not add more steps than that. We measured where the time actually goes first, and after the pictures are up there is genuinely nothing happening on your phone worth narrating — the rest is Rafe thinking, which is what it already said.
-
The answer follows the photo, not just the notes
Last week we made the notes taken about a photo follow it onto a plant once you say which plant it is. Two things were wrong with that. The first is that it only worked when a photo was matched through the chat — matching one from your Photos screen, which is where we actually ask you to, did nothing. The second is that it moved the notes and left the answer behind, so you could get the readings about your lavender onto the lavender and not the explanation of them. Both fixed. And the card that asks you to match unplaced photos now counts a picture that has a full answer on it, not only ones with notes — one of the three waiting in a real garden had an answer and no notes, so nothing had ever asked about it.
-
The button now says who is answering
The fourth button along the bottom used to say "Ask". It now adds his name after it. Nothing moved, nothing changed about what it does, and the chat behind it is exactly the chat that was there yesterday. What changed is that you can tell there is someone to ask before you tap it. Rafe is the part of the app that answers you in words, and until today he was only ever introduced to people who had already found him — open the chat and he is there, never open it and the app looked like it had nobody in it. He is a gardener. He has no face and he is not pretending to be a person; the name is there so the thing that talks to you has something to be called. We also tried putting his mark on the round microphone button instead and threw it away: at that size the leaf read as a drip running off a melting microphone, and the microphone is the only thing on that button telling you it records your voice.
-
It offers the plant it thinks it is looking at
Send a photo without saying which plant it is and Rafe asks, with a row of buttons underneath the answer. Those buttons were your most recently touched plants and nothing more, so a photograph of beans could offer an apricot and an orange tree — with the answer directly above saying "that bean is coming along". The app had already worked out what it was looking at and then ignored itself when drawing the buttons. Now it offers the plants it thinks it is seeing. Name a bean and you get your beans. If you have two of that crop you get both, because a photograph cannot tell them apart and choosing one for you would look like an answer rather than a guess. There is always a way out: "A different plant" is there whenever the list has been narrowed, however short it is, so a wrong guess costs you one tap rather than stranding the photo. And if it cannot tell what it is looking at, it shows you everything, as before.
-
The photo now says where it went
Ask Rafe about a plant from that plant's own page and the photo is filed there — that part already worked. What it did not do was tell you. Under the reply there is now a line saying which plant it went on, with the name linked, and a "Not that plant?" button that takes it back off and asks you instead. That second half matters more than it sounds: once a photo is filed it stops appearing under "Photos with no plant", so a wrong guess used to be both invisible and unfixable.
-
The plant comes with you
Tap "Ask Rafe about this plant" on a plant's page and the plant now comes with you into the chat — you have always seen it named at the top of the box, and you can still switch it off if you meant something else. What changed is what happens to a photo you send from there: it is now filed to that plant, so it joins that plant's own run of pictures instead of belonging to nothing. Before today only the + button in the chat did that. Tapping the chat tab from a plant page still files nothing, on purpose — being on a page is not the same as saying which plant you mean, and we would rather ask than guess.
-
The plant decides what counts as wrong
When we look at a photo from your garden we write down what we can see — first flowers, pods filling, lower leaves going yellow. Your week was about to list all of it under "what is wrong", which would have told you your tomatoes were in trouble for flowering. Whether a sign is a problem depends on the plant, not on the sign: a tomato setting flower buds is the whole point, and a basil setting them means the leaves are about to turn bitter. Same thing to look at, opposite meanings. So we stopped deciding and started asking the plant's own record, which already says which signs matter for that crop and when — it is the same source your growing advice comes from. A bed is judged as a whole, so if there is basil in it, flower buds there are worth a look even when the tomato beside it is doing exactly what it should. We also fixed a fault that would have stopped the week's job list building at all once any photo was tied to a plant.