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.
-
The question waits for the answer
Send Rafe a photo and, when he cannot tell which plant it is, a short row of buttons appears under the reply asking you to say. That row was appearing too early — as soon as the reply started arriving rather than once it had finished. So the buttons and the words were landing at the same time, and a beta tester could not tell whether tapping one had caused the rest of the answer to appear. It waits for the answer to finish now. Nothing else changed: the same question, the same buttons, and ignoring it still costs you nothing.
-
Sending the same picture twice
If you sent the same photo more than once — a second tap, or picking the same file again days later — the app stored it as two separate pictures. Ten of the sixty-five pictures in one garden were copies of something already there, one of them five times over. Now it recognizes a file it already has and keeps the one you sent first, with its original date. It matters more than tidiness: once a picture is matched to a plant, the app looks at it properly, and five copies of one picture is five of those. Nothing you sent is lost, and a genuinely different photo is always its own picture, even of the same plant a minute later.
-
Two doors that were paragraphs
On a plant's page, the way in to the symptom guide said "Something looks wrong?" followed by a sentence explaining itself. It now says "Match the symptoms yourself — free, every line quoted with its source". Same door, same place, and the two things worth knowing before you tap are on the face of it. The second one mattered more. When we had read a photo for you, the card said "What we made of your photo — read it, and tell us which parts are right". That asked you to check our work at the moment you most wanted to trust it, which is a job we handed you rather than a feature. It now just says what it is and takes you there. Correcting a reading has not gone anywhere — the buttons for it sit on the reading itself, where they always were, and that is the right place for them: something you can do if you want to, not something asked of you before you have finished reading.
-
Walk to a plant, not just to a bed
When you pick where you are going on a garden walk, each growing space now has a small arrow beside it. Tap it and the plants in that space appear underneath, and any of them can be a stop of its own. Tapping "Back bed" means you will photograph the bed; tapping "Sungold" beneath it means you will photograph the Sungold, and tapping both gives you both, in the order you tapped them. The reason it matters is what happens afterwards: a picture tied to one plant goes into that plant's own run of pictures and can be compared with the next one, while a picture of a whole bed belongs to the bed rather than to anything in it. If you do not want any of this, ignore the arrows — the walk is exactly what it was, and you are still never asked which plant a picture was of while you are out in the garden.
-
We described the height offer too broadly
Earlier today we said a photo of a plant gets you its height offered back. That is true when you ask Rafe about the picture in the chat, and it is not true yet when you add a photo straight to a plant from its own page. Those are two different routes through the app and only one of them measures. The other one is watching for what is happening to the plant — new leaves, trouble, change since last time — and it has nowhere to put a number. We have corrected the help pages to say which is which. If you want a height today, ask about the picture in the chat, and it helps to stand back and shoot from the side with the pot or your hand in frame so there is something to judge the size against. This is the second correction to this feature in one day: we described it before it worked, and then we described it more broadly than it worked. Both are on us, and we would rather say so than leave a page you cannot trust.
-
We spotted something in that picture
When you send a photo the app writes down what it can see, even before it knows which plant it is. Since last week those notes follow the photo onto a plant the moment you say which one it is. What was missing was any reason to bother: the card on your profile just told you how many pictures had no plant, which is a chore with a number on it. Now it leads with what is actually in them — we spotted something in three of your pictures, tell us which plant and what we noticed goes on that plant. It only says that when it is true. When there is nothing waiting it goes back to the quiet version, because most pictures with no plant have nothing in them worth chasing you about.
-
What it noticed waits for the plant
When you send a photo, the app looks at it and notes what it can see even before it knows which plant it is. Those notes were being kept, but they had nowhere to go: they were attached to the picture and not to any plant, so no plant page could ever show them. Now, the moment you tell the app which plant a photo is of, everything it already noticed about that picture comes with it. Nothing has to be looked at twice, and a photo you file a week later brings its notes along. Notes that already belong to a plant are never moved — whoever put them there knew more than this does.
-
When an answer gets cut short, it now says so
Every answer has a length limit, and once in a while a long one runs into it and stops mid-sentence. The app knew when that happened and said nothing, so a reply that ended in the middle of a word looked like your phone had dropped the connection. Now there is one short line under it: that answer was cut short. It does not tell you to ask again for the rest, because asking again starts a fresh answer rather than picking up where that one stopped, and we would rather say nothing than promise something the app cannot do. If you want the rest, asking a narrower question is the thing that actually works.
-
When it misses something, it says so about itself
If the app cannot write down something you told it, it says so under the answer rather than letting a cheerful reply stand. Two of those messages were blaming you. One said you had not asked it to write anything down, and another said what you mentioned was not something you said. Both of those appear when our side loses your words on the way through — you did ask, and there is no way for you to know otherwise. They now say what actually happened: it did not catch it. Same information, and it stops telling you something untrue about yourself on the one line whose whole job is owning up to a mistake.
-
When we are sure, we say so
Take a photo of a plant and we tell you what we spotted on it. Every one of those used to arrive as a question — including the obvious ones. A gardener photographed a healthy young bean and got asked, five times, whether it had leaves. Now it depends on how sure we are. When we are confident, the card says what we noticed rather than asking you to confirm it: "Rafe noticed first blooms", with "Sounds right" to agree and "Actually, no" if we got it wrong. When we are genuinely unsure, it still asks properly, because there your answer really does decide something. Either way it takes one tap and it stops coming back. And when the same thing shows up in several photos of one plant, you now see it once instead of once per picture.
-
Which plant was that?
Send a photo into the chat and you get an answer about what is wrong with it. Until now the picture itself was filed against nothing, so it never appeared on any plant's page and nothing could be compared to it later. Now, when we do not already know which plant it was, there is one short line under the answer asking, with a button for each of your plants. Tap one and the photo goes on that plant's page, dated the day you took it. Ignore it and nothing is lost — the picture stays exactly where it was, and you can still put it on a plant later from Photos with no plant. It never asks when it already knows: a photo taken from a plant's own page, or one where you picked the plant with the + button, is filed without asking. And it never holds up the answer — it appears underneath, after the reply you were waiting for.
-
A greenhouse is a place, not a pot
When you tell the app what kind of space a bed is, it used to offer four answers: a raised bed, in the ground, a pot, or a greenhouse. Three of those say what the roots are sitting in. The fourth says what is over your head, which is a different fact — and putting them in one list meant somebody with a potted lemon under glass had to choose which half of the truth to tell us. Picking the greenhouse cost them the pot, and the pot is the half that changes the watering. So the question now offers three, and the greenhouse moves up a level to sit with the back yard and the front porch — the places a growing space lives in. Nothing you already recorded was thrown away: a space you marked as a greenhouse keeps its mark and its size, and rain falling on your garden is still not counted as having watered it, because glass sheds rain. The cold is deliberately unchanged. An unheated greenhouse still freezes on a hard night and we cannot know whether yours is heated, so frost warnings treat it exactly as before. Being a place rather than a kind does not make it warm.
-
A label that guessed who did it
When something cannot be saved to our servers, your profile lists what is stuck. A journal entry in that list was labeled "Something you logged" — which was true for as long as you were the only one who could write one. Rafe writes journal entries now, so for some of them it was simply wrong, and being told you did something to your own records that you did not do is worse than a vague label. That list only knows what kind of thing is stuck, never who made it, so it names the thing instead: "A journal entry". While we were in there, "A bed or spot" became "A growing space", matching what the rest of the app calls it. Two names for one thing on one screen is how you end up wondering whether they are two different things.
-
A map of the whole app, behind a button
There is a menu button at the top left of every screen now. Open it and you get everything the app can do, grouped, with a line under each one saying what it is for. That last part matters more than it sounds: there were eighteen places to go and five of them had a button anywhere. Garden mode, catch-up, your photo roll, the week in review — all built, none reachable unless you stumbled out of another screen onto them. There is a search box too, and it matches what a screen is for rather than only its name, so typing "frost" finds the screen for pinning your garden on the map even though the word frost is not in its title. The bottom row of buttons has not changed. That is for the things you tap every day; this is for everything else.
-
A photo can stand in for a planting date you never gave
For most crops our calendar can say nothing at all about where a plant should be without knowing when it went in — 110 of the 152 we hold. If you never typed a date, those entries recorded nothing. Now, if you photographed the plant and we could see it was a seedling, we work backwards: only a narrow band of sowing dates fits a seedling on that day, and we use the middle of it. The entry says "planting date estimated from a photo" wherever it appears, so you always know which kind of date you are looking at. It never replaces a date you gave us — if you have entered one, this does not run at all. And it is never used to judge the photo it came from. That would be our own guess agreeing with itself: we worked the date out FROM that picture, so of course the two match, and a journal saying "just as expected" on the strength of it would look like confirmation and mean nothing. It only describes days after the picture was taken, where the plant can genuinely surprise us.
-
A photo you send now lands on the plant
Send a picture of something you are growing and ask about it, and the picture now goes onto that plant, not into a pile you have to go looking for. It sits in the plant's own photo strip, and opening it says what we saw — including which stage the plant is at, which we had been working out and then throwing away. There is a link back to the conversation too, because the useful half of an answer is usually the part no form has a box for: that your mulch is doing its job, or that the nibbled leaf is not worth treating yet. We are not going to invent a field for every nice thing that can be said about a plant. We would rather keep the words and point you at them. One caveat worth stating plainly: we only file the photo when we are sure which plant it is. If you grow two things with the same name, we leave it unfiled rather than guess — a picture on the wrong plant is a quiet mistake you would have no reason to check.
-
A setup task that makes you actually use the microphone
The microphone on every screen has always gone straight to Rafe, and almost nobody knew. There is a setup task for it now, and it does not clear when you press a button saying you understand — it clears when your words actually reach a conversation. Speaking or typing both count. That sounds stricter than it is, and it is the point: a card telling you the microphone talks to Rafe is exactly the thing that was already not working. Using it once teaches it in a way reading about it does not. If your browser has no microphone, or you declined the permission, the task clears on its own and tells you to type instead — you are never left holding a job you cannot finish.
-
A way back, from the place it happened
When Rafe writes something down or makes something, the chat has always said so in a line under his reply — "Already done, added 6 Sungold, watered". That line now ends with "Change or undo", which takes you to the list of everything he has done and a way to reverse any of it. That screen has existed for a while, and there was no route to it from the chat, which is the one place you would think to look because it is where you just watched it happen. We considered adding a count of what changed instead, like the +411 −30 badge a code editor shows. The line already names them, so a number beside it would have been decoration. The missing thing was the way back, not the arithmetic.
-
Bolting is only a problem if you wanted the leaves
A lettuce running to seed is a disaster if you are picking leaves and it is the whole point if you are saving seed for next year, and until today we only knew how to tell you the first one. Now a plant’s own page asks, once its picking window opens or we start expecting a flower stalk: are you growing this for food, for seed, or some of each. Say seed and running to seed stops being counted against the plant. The hot and cold spell cards say it is coming instead of warning you about it, and they stop ranking it above things you can actually do something about. Your journal stops marking it as a setback. And it moves out of the list of things to watch out for and into the list of things to look for, because it is the thing you are waiting for. Say some of each and we call it neither good nor bad — it is the moment the leaves are done and the seed is coming, which is a decision rather than a verdict. Say nothing and nothing changes: we treat a plant as grown for food unless you tell us otherwise, because that is what nearly everyone is doing and the warning is right for them. It is set per plant, not per crop, so two lettuces in the same bed can be doing two different jobs.
-
Empty screens that explain themselves
Before you have made a growing space, or grouped anything into a garden zone, or walked your garden, those parts of the app used to show a line of gray text or an empty list. An empty list tells you nothing you did not already know. They now show a card that says what the thing does for you, links to the help page about it, and offers the one way to start. It goes away the moment you have one — it is not a nag and there is nothing to dismiss. Alongside that, plants in your weekly report now carry a small label: on track, behind, or "can't tell yet". That third one matters most. We can only say a plant is on track when what you photographed and what our calendar expected use the same word for it, and most of the time they do not. It would be easy to leave those plants unlabelled and let the screen look tidier, but a blank space reads as "fine" — so they say plainly that we could not tell, and why.
-
If frost does not reach you, we stop counting down to it
Gardeners in Santa Monica, Hialeah and the Keys pin their address and are correctly told there is no average frost date worth giving them. The top of the Today screen then counted down to one anyway — 51 days to frost — with an "add ZIP" link beside it. The number came from a typical zone 6b garden standing in until we know better, except we did know better: they had just told us exactly where they are. And no ZIP was ever going to produce frost dates that do not exist, so that link was an instruction you could not carry out. It now says "frost is rare here" or "no frost here", and offers nothing to fix. The season chart and day count carry on as before — you still have a growing season, it is just not bounded by frost.
-
It now tells you when it did NOT write something down
Rafe writes his reply before the app has worked out whether the write is allowed, so he could say "Got it, I'll write that down" about something that then did not happen. We found exactly that while testing against a garden with three pole beans in it: "I watered the beans-pole" could have meant any of the three, so nothing was written — which is the right call — and the sentence said it had been. Nothing on the screen contradicted it. That is the worst thing this app can do: not a wrong answer, but a wrong reassurance about your own records, which you have no reason to go and check. The app now says so in its own line, above the answer: "Not written down — I could not tell which plant you meant." It appears whenever something was wanted and did not happen, whatever the paragraph above it claims, because it reads what the app did rather than what Rafe said. The line telling you something WAS written has worked that way for a while; this is the missing other half. Say which plant you meant and it goes straight in.
-
He says what he cannot do, instead of saying he will
Ask the assistant to delete a plant and he used to answer "I'll delete that plant for you." He has no way to delete anything — he never has — so that could not come true, and nothing on the screen said otherwise. We found it today, testing the assistant end to end for the first time. It is worse than a wrong answer: you pulled the plant out of the ground, you were told it was handled, and it would have gone on making jobs for a plant that is not there. The fix is not telling him to choose better words. He now has a way to say "I have no tool for that", and when he does, the app writes the reply rather than the assistant: Rafe cannot remove plants, and here is where you do it. The same for changing a date and for undoing an entry. It names both ways of ending a plant, because they are different and the difference matters — "It is done" keeps the whole record and takes it off your list, and "Delete this plant" is for something you added by mistake. A plant that really grew is worth keeping the record of.
-
One word for the thing you grow in
A bed, a pot, a row, a windowsill — the app used to call these your "places". That was fine until there was a level above them. A garden zone is a part of your garden that holds several of them: your back yard, your front porch. With both on one screen, one word was doing two jobs and there was no way to tell which level anything meant. So the thing a plant is rooted in is a growing space, everywhere — the heading on your garden screen, the button that makes one, the question asking which one a photo is of, and the help page about them. Nothing moved and nothing was renamed in your garden; only what we call it changed.
-
Setting your location on a brand-new phone now works
On a device that had never saved anything yet, pinning your location did nothing at all. You would see your frost dates previewed, press the button, and the screen would sit there — no error, no explanation, nothing stored. Your garden is created the first time something is saved, and the save code was checking whether one already existed and giving up when it did not, which is exactly the moment it was meant to make one. Three ways in were affected and all three are fixed: looking your dates up, typing your own, and saving a spot that has no frost dates to look up. Two questions in the setup flow had the same problem and dropped your answer the same way. If this happened to you nothing was corrupted — it simply never ran. Set it again and it will stick.
-
Rain in inches, and a clearer line about what Rafe did
Two small honesty fixes. The weather days showed a chance of rain — 43% — when what you actually need is the amount. A 90% chance of a passing shower wets nothing; a 40% chance of a thunderstorm can drop half an inch. The forecast has carried the amount all along and no screen read it, so you now see "0.4 in" where we have it and the chance where we do not. Second: when Rafe has added something, the line explaining it used to name only the permission — "you let it write things down" — which reads as though Rafe acted on a switch you left on. He cannot. We checked the code rather than assuming: Rafe only writes when you said the thing, in that conversation. So the line now says "you asked, and you let it write things down". Both halves are true and the first one was missing.
-
Entries written from a photo now say so
Journal entries whose words the app wrote have always carried a small "read by the app" mark. That was right, and it never answered the question you would actually have: read off what? A sentence written from a photograph is a different thing from one written from what you just said — you watched the second one happen, and the first you might not remember at all. You took a picture, and a sentence appeared in your records describing something you never typed. Those now say "read from your photo". Everything else the app wrote still says "read by the app", and anything you typed yourself carries no mark. Your journal is your record, and in three years the difference is part of how much weight to give a line.
-
Resting now means resting
Seven crops have a curing step — garlic, luffa, onion, pumpkins, shallot, winter squash and sweet potato. Curing is what happens to the crop after you pick it: skins hardening in a shed for a week or two so it keeps. It is not the plant having a rest, because by then the plant is finished. We had it filed under the same heading as dormancy, which is a real rest — an apple tree in January is asleep and comes back — so the season chart said "Resting" over a squash sitting on a shelf. Curing now shows as Cropping, since you are still getting your crop in and that is the last step of it. And Resting now means exactly one thing: a plant that is asleep and will come back. Forty-five of our crops do that. The rest never will, and their chart simply ends after cropping — which is the year finishing, not something missing.
-
One name and one pronoun for the helper
Small copy change with an embarrassing cause. The app referred to the helper (now Rafe) three different ways in three different places: "she" in the setup screens, "its" in your profile settings, and "his" in one of the help pages. You could meet all three inside your first week. None of them was decided; they were just written at different times by different hands. Every one of those now uses his name, or nothing at all. The settings heading changed from "How much … can do on its own" to "How much … can do without asking", which is also just a better description of what the setting does. Whether he is a character or a tool was a real question, and we should not have been answering it by accident, three ways, in shipped text. The name is Rafe now, and the pronoun is he.
-
Setup asks what you came for
Setting up your garden used to ask two things: where you garden, and one thing that is growing. Everybody then got the same six crops to pick from — tomato, lettuce, basil, pepper, carrot, bean — whatever they had come for. There is now a question before that one: what do you most want from your garden? Food, or to stop losing plants, or to get better at it, or a place that looks good. It is one tap, and the point is what it changes. Answer "a place that looks good" and the next screen offers marigolds, sunflowers, zinnias, nasturtiums, lavender and cosmos instead. It also decides which job sits at the top of your list afterwards: somebody who came to stop losing plants gets asked how much sun each spot gets, because sun is what the watering advice is worked out from, and watering is what kills a plant in a pot. You can skip it and nothing is held against you. Two more screens joined the end: one asking which way of paying would suit you — nothing is charged, no card is asked for, and we do not show a price because we have not set one — and a last card that hands you to Rafe, who can add the rest of your garden while you talk to him.