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.
-
A garlic planted five days ago is no longer asked whether its scapes have emerged
A few crops are grown to a time of year rather than to a clock that starts when you plant them. Garlic is the clearest: it goes in around the first frost of autumn and its scapes come the following summer, so everything we know about garlic’s year hangs off your frost date. That is right for a garlic planted when garlic is planted. It fell apart for one planted at the wrong time of year — a garlic put in during May was asked, five days later, whether its scapes had emerged. The plant was fine; the question was about a different garlic, one planted eight months earlier, and nothing was checking. Now if your planting date is well outside the window that crop normally goes in, we stop giving you jobs for it and keep the general note instead. "Garlic is usually scaping about now" is true about garlic. "Your garlic needs its scapes removed" would not have been true about yours. Two weeks late still counts as on time; a whole season out does not. Two things we deliberately did not do. We did not check how old the plant is — that would silence a nine-year-old apple, whose flowering and harvest genuinely do come round every year no matter how long it has been in the ground. And we did not go by whether the plant is an annual or a perennial, because our own records call garlic a perennial, which is true of the plant and useless here. Four crops are affected: garlic, horseradish, cereal rye and hairy vetch. Everything else already works from the date you gave us.
-
When the answer tells you to do something, it can become a task
Until today GardenTrack only offered to set a reminder if you asked for one, or if it had spotted a pest or a disease worth checking again. So a beta tester told to shade his potted beans from 2 to 5 PM through a heat spell had nowhere to put that, and said so: this would have been a great place for it to remember to remind me later. Now advice that tells you to do something can become a dated task in one tap. The test is whether the answer told you to DO something, not whether it named a day. Feed it, shade it, net it, repot it are all jobs; if a time came with the advice the task carries it, and if not, nothing is invented. What it will not do is offer on every answer. Beans like full sun is information rather than a job and gets no button, and we checked that against the real assistant rather than assuming it — a button under every reply is worse than no button, because it is the sort of thing people stop reading. That checking also caught our own mistake. We had first built this to fire only on advice that named a time, and the same tester pointed out that a fertilizer suggestion with no date on it should still be a task. We changed the rule, and the live test then showed the instruction we give the assistant still carried the old version — so it stayed silent on exactly the case he had raised. Four real messages, about nine cents, and a fix that no amount of reading our own code would have produced.
-
Three date boxes were reading the wrong calendar. All three now read yours
A date has to be asked of somebody's clock, and for three small pieces of GardenTrack that clock was London's rather than yours. The awkward part is that the three broke for opposite halves of the world, which is why none of them was ever reported. The planting date on a plant's own page read one day early for anyone gardening east of London — all day, every day, not only in the evening. That is the one that could do lasting harm: the date box on that page opens already filled in with it, so editing anything else on the plant and saving could write the early date back over the right one. The second was the date box on "when did it go in", which is the only thing stopping you recording a planting that has not happened yet. West of London, late in the evening, it would let you pick tomorrow — after six o'clock in Colorado, after eight on the east coast. The third was on this website: add a plant to your garden from a plant page in the evening and the My garden list said "added Aug 22" on the 21st. That one was only the line on screen. The plant was saved correctly, and the date the app records when you carry your picks across has always been worked out fresh on your own phone. All three came from the same one-line habit, so there is now a single piece of date math the whole app shares, and the build refuses any new copy of the old one before it can be committed. Worth knowing: a date already saved a day early stays a day early. If you garden outside North America, open a plant or two you added before today and check the planted date.
-
Every plant now knows its family — and it caught us handing out the wrong description
A bug started this. Our database had tomato hornworm listed on dill. Every host the source names is a nightshade — tomato, potato, eggplant, pepper — and dill is nothing like a nightshade. The right caterpillar for dill was already in our data, one row away. Nothing caught it, because "what family is this plant" was not something we knew. Now we do, for all 152 crops. We asked two independent botanical authorities separately for every one. 103 of 107 plant groups got the same answer from both; four did not, and instead of quietly picking a side we wrote down what each said and why we chose. Every plant page now names its family and points at its relatives — tomato at pepper, potato and eggplant; dill at carrot, celery and parsley. That is the fact rotation is built on: relatives share pests and diseases, so what went through one bed is waiting for the next thing in the same family. Then we pointed a check at it, and it found something better than the bug we were hunting. The same common name often belongs to two different organisms. "Brown rot" is a stone-fruit disease and also a potato disease, and they are unrelated. "Angular leaf spot" on a cucumber and on a strawberry are different organisms. We were showing the description we had sourced regardless of which one you actually had. Twenty-one of those crossings existed; all twenty-one now say so on the crop it affects. Thirteen say plainly that nobody has checked which organism yours is — which is a real answer, and a better one than a confident guess.
-
Read the note before it is saved, and see where it went afterwards
When an answer offers to write a note into a plant's story, the note now opens first. Tap the plant and a small card appears with the whole sentence in a box you can edit, and nothing is stored until you save it. The card says that out loud, because a card showing a finished sentence is exactly the sort of thing anybody would reasonably read as done. Before today the button could only fit about forty characters of it, so most of the note was never on screen — you were agreeing to words you had not read — and once a note is in a plant's story there is nowhere in the app to edit it. The moment before saving is the only chance there is. The tick afterwards changed too. It said Noted in this plant's story, which names nothing and points nowhere, and now it names the plant and links to it: Noted in Bush bean's story, open it. A confirmation that will not say where something went is the same thing as a note filed against the wrong plant — nothing looks broken and you cannot find it afterwards. Pests, diseases and harvests got the same sentence. And the conversation you have open is now part of the address, so when the app updates itself and reloads, you come back to the chat you were in rather than to a blank Talk screen. Starting a new chat still leaves the old one behind, exactly as it did.
-
It now says whether your crop beats the frost, instead of leaving you to check
A beta tester was told his wax beans would not finish before his first frost. Our own records say they will. He pushed back with the right arithmetic, the assistant apologized, and then restated the same error with the two numbers swapped. The cause was not bad math on our side. The app works out ready dates in code, from your planting date and a sourced day-count, exactly so nobody has to estimate — and it had done that correctly. What it then did was hand over two dates with the instruction never do this arithmetic yourself, and no answer. We only ever spoke up when the frost won; when the plant won we said nothing at all, which left the one question that mattered sitting there undone. Told not to compute, given everything needed to compute, and given no result, it computed. Every ready date now carries how it comes out against your frost. Comfortably clear, with the number of days. Tight, and said as tight. Landing on your frost date, said in those words. Straddling it, and you get both ends of the window and the call, because a range that opens before your frost and closes after it really does depend on the autumn. One deliberate piece of caution: a hair-thin margin is not rounded up into a yes. A day-count is a catalog figure for a good season, so anything within a week of your frost is described as tight rather than safe.
-
A plant with no date now asks for one, instead of quietly guessing
The planting date is the day every other date is counted from — when to expect it up, when it fruits, when it runs out of season. A plant without one is handed the jobs of a plant put in this morning, because that is the only thing left to assume. On a tree that means being told to stake it, mulch it and rub suckers off something that has been in the ground for years. The app had a card asking for the date. It was at the very bottom of the garden list, below routine watering, in the lowest priority band there is — which is where a question goes to die. It is now an app task instead, sitting above the garden jobs rather than buried among them, because those are two different kinds of work: things you do outside with your hands, and things you tell the app so everything else it says is about your garden rather than an average. It also has no Hide, which is deliberate and rare. Every other item in that list can be dismissed. This one stays, because skipping it costs you something you cannot see happening — months of slightly wrong dates on a plant, with no way to spot it and no way to repair it afterwards. Answer it once and it is gone for good. What has not changed: adding a plant never requires a date. You can still add one from the chat with nothing but a name. The gap becomes a job, never a wall.
-
Every plant page now says how late you can still sow it — wherever you are
The autumn part of a plant page used to be one extension service’s chart. Clemson prints fall planting windows for two regions of South Carolina, and we quote them honestly, labeled with the place they were published for. That is 24 crops out of 152, and useless to anyone outside one state. The reason nobody else publishes the same thing is that a single fall date only works where the fall is long. Minnesota can’t print one, so it prints the arithmetic instead: count back from your own first frost by how long the crop takes. We can do that. Set your zone or drop your ZIP on any plant page and it now says the last date you can still sow that crop and have it finish before your frost — 111 crops, in every state, and the date moves when you change the zone. Radish is late September in zone 6, mid November in zone 9, late August in zone 3. Two things it is careful about. It uses the fastest figure in the range, so it is the last date that could work rather than one that usually does — being wrong the other way tells somebody their season is over when it is not. And 63 of those crops say "set plants out by" rather than "sow by", because that is what our sources actually measure for them. Counting a sowing date back from a set-out-to-harvest figure would be weeks wrong in the direction that costs a season, so we say which question we answered.
-
Your shady spot is now something the app knows, and it says so on the plants that mind
A porch is not a lawn, and until today nothing in the app knew the difference. Open any spot from My garden and there is one more question under the size: **how much sun does it get?** Full sun, part sun, or shade, with the hours behind each shown the moment you pick, so you can check your own spot against them instead of guessing at what we mean. **Where the name you gave it already answers, we offer the answer and wait.** Call somewhere *Front porch* and it comes back with *"Front porch" sounds like part sun* and a button saying **Yes, that is right**. Nothing is saved until you tap it — a reading taken off a name is not an answer you gave, and once it were written down nothing could tell the two apart. Names that settle nothing get nothing: *Back garden* is not asked to mean anything, a bare *balcony* could be open to the sky or tucked under the floor above, and *shady patio* names two things at once, so it asks rather than picking one. **Then it is used.** A plant living in a spot you have marked gets one line where the light changes something for it — *Front porch gets part sun. Grows in part sun, but expect a lighter crop than in full sun.* **Most plants get nothing, on purpose.** Nearly every food crop wants full sun, so a warning keyed on that would sit under every plant in every shaded bed and mean nothing at all. What decides it is how each crop copes with less: 57 lose nothing worth mentioning and stay quiet, 22 crop lighter and say so once, and six are happier in the shade and are told so as good news rather than as a warning. Where we hold no record either way, the line says exactly that instead of predicting a failure.
-
The harvest question is drawn now, not typed out
"When will things be ready?" is one of the buttons we put at the top of Talk, so it is a promise: we knew the question was coming and we had prepared for it. It was coming back as four lines of plain text, in whatever order the plants happened to be stored in, with the assistant asked to sort your dates for you. Now the app draws it. Across the top is your season — last frost on the left, first frost hatched on the right, today marked — with each plant's picking window as a bar on it, so the ones that land with room to spare and the ones that run off the end are one glance apart. Underneath, a row per plant, soonest first: how far away it is, the dates, how it comes out against your frost, and the day-count we used with what that count is measured from. Every row opens that plant. Three things are said differently on purpose, because they are three different facts — a fruit tree or a lavender gets a picking season rather than a countdown, since those windows come round every year and say nothing about whether a young plant is old enough to bear yet; a plant with no planting date is named with a link, because that is one tap from being answerable; and a crop we do not carry says so, rather than reading as if we had lost your plant’s timing. The dates themselves have not changed. They were already worked out on your phone from your own planting dates and a sourced day-count — the only thing that changed is that the app stopped handing them to something that guesses and asking it to read them back.
-
The microphone button knows which plant you are standing at, and takes a photo
Open the round microphone button while you are on one of your plants and the box now says which plant before you say a word: About your Bush bean. Talk, and the answer comes back about that plant — its age, its bed, what you have written down about it — rather than about the crop in general. If it has guessed wrong, tap the name and it lets go; tap it again and it comes back. It is a switch, not a dismissal. We show the assumption instead of using it quietly because you might be standing at the beans and thinking about the tomatoes, and an answer confidently about the wrong plant is worse than one that had to ask — it sounds finished. On Today, Garden, Plan and Profile there is no name to show, because there is no single plant you are looking at. Take a photo is in the same box now. Point it at whatever you cannot put into words, send it, and the chat answers about that. A picture on its own is a whole question — you do not have to type anything with it — and a quieter link underneath picks one you took earlier, for the leaf you photographed in the garden and sat down at the kitchen table to deal with. Nothing is written into your garden until the app reads it back to you on a card and you agree, the picture included, and Remove takes it back off before you send. Two smaller things came with it: the link at the bottom of that box was cream text on a cream sheet in light mode and could not be read, and neither could the mic's own label or the line explaining why dictation had stopped.
-
The numbers on our help pages had quietly gone out of date, so we re-counted every one
Seven help pages and a guide told you what our plant database covers — how many crops carry a harvest cue, how many have a watering figure, how many we hold at all. Every one of those sentences was true when it was written and had since stopped being true, because the database kept growing underneath them. They said 151 crops and 3,176 varieties; it is 152 and 3,174. The catch is that the other half of each sentence had moved too, so this was not a matter of changing one digit. Harvest cues had gone from 143 crops to 150. Watering figures from 147 to 148. The count of plants whose growth stages we describe went from 846 to 852. One line said thirty-eight crops have no sowing or planting-out date and that most of them are fruit trees and berries — it is now ten, and only two are berries, so we rewrote the sentence rather than the number. Each figure was re-measured against the shipped data. We deliberately left the ones that are history: a changelog entry from the sixth of August saying we had 151 crops was true on the sixth of August, and rewriting it would turn a record into a fiction.
-
Three plants had a winter-hardiness number, and none of them lives through a winter
A hardiness zone answers one question: how cold a winter can this plant sit outside through and start growing again in spring. Sweet pea, lima bean and buckwheat all had one on their page, and all three are sown from seed every year and finished by the end of the season. Nothing is asked to live through a winter, so there is no such number to give. What was actually on those pages was the range you can grow the plant in over one summer, which is a different fact wearing the same label. Sweet pea said zones 2 to 11. Read as hardiness, that says it comes back on its own after a Minnesota winter. It does not. We took the line off those three pages and kept the figure in our own records, because "you can grow this almost anywhere in one summer" is worth knowing and we would rather move it somewhere honest later than lose it now. We cleared 59 crops of the same mix-up in August. Two of these three are crops we had already fixed then: a file got restored from an older copy to recover a different edit, and the fix went back out with it, quietly, because undoing something leaves no error message. So this time we also taught the build to refuse it. A plant that only lives one season can no longer carry a winter-survival number at all, and that runs on every build rather than waiting for somebody to remember. Plants that do come back are untouched: 85 crops still carry a real zone range, and the app still tells you before you plant a fig somewhere it will not make it through February.
-
Your young fruit tree is now told it is young — we had the number all along
Ask when things will be ready and a fruit tree gets a picking season rather than a countdown, because that window comes round every year and is a fact about your area. Alongside it, the app said it did not know how many years a young tree needs before it starts bearing. **That was wrong, and it was wrong about our own records.** We have held that figure for years: an apricot bears 2 to 5 years after planting, cross-checked against two university extension sources; a pear 3 to 10; a fig 3 to 4. The app was reading the field under the wrong name, so every count came back empty and looked exactly like a gap in the data. A founder reading his own one-year-old apricot is what caught it — it told him the apricots around him ripen from late July, and said nothing about his tree being two summers short of its first crop. Now each of those rows says both: how long that plant takes to come into bearing, and where yours stands against it — *bears 2 to 5 years after planting, 1 season in, not old enough to crop yet*. A plant too young to pick from is no longer marked as ready to pick, which is the part you could see. Past the far end it stops reassuring you and starts asking a question instead: twelve summers into a pear that bears by ten, if it has never set anything, that is worth looking into. **It reads the rootstock too.** A lemon grown from a pip takes 7 to 10 years where a grafted one takes 2 to 3 — same crop, same age, opposite answers — so if you told us which yours is, you get the figure that applies to your tree rather than the average of two. We hold this for 29 of the 39 plants that answer with a yearly season. For the other ten we still do not, and the card now names them at the bottom rather than saying it about everything.
-
A hot spell can reach your phone now, and it knows which plants are in pots
Someone on our team stood outside in 100°F heat and opened the app to ask whether the beans on his patio were getting too much sun. Everything needed to answer that was already on his phone: our records say bush beans start to suffer at 90°F and that figure comes from a university, the forecast with the 100 in it had already been downloaded, and the app knew the beans were in a pot because he had told it. There was even a card ready that said all of it. He never saw the card, because he was outside and the app was in his pocket. Frost was the only weather in this app that could reach somebody who was not already looking. Now a hot spell can too. The evening before a run of hot days, you get one alert naming the temperature and the plants — and a plant in a pot gets a different instruction from a plant in a bed, because you can pick a pot up and carry it out of the afternoon sun, and a pot runs dry in a way open ground does not. Two limits worth saying plainly. It covers 39 of our 152 crops — 37 where a university has published a temperature we can point at, and two that are our own estimate. **Corrected 26 August 2026: this said 43 and 109, and described all of them as university-published, which contradicted the very next sentence.** For the other 113 we stay quiet, and that silence means we have no number rather than that your plant is fine. And two crops we do hold a figure for — sunflower and snapdragon — are deliberately kept off the phone, because those numbers are our own reading rather than a published one. A number we can show you a source for is allowed to wake you up. A number we cannot is allowed to sit on a screen you chose to open, and no further. Sunflower’s is 78°F, which is an ordinary summer afternoon over most of the country, so that rule is the difference between an alert and a nuisance.
-
A nursery tag is two facts, and we were filing it as one
Tell GardenTrack what the label on a new tree says and it sets the plant up for you. A tag reading "apricot Mormon semi dwarf" was becoming a variety called "Mormon semi dwarf" — but the variety is Mormon, and semi-dwarf is the rootstock, which is the root system the tree was grafted onto. That is not a tidiness problem. Rootstock is a field the app already had and already reasons about: it decides how big the tree finally gets, how far apart to space it, and whether it will send up shoots from the base at all. Buried inside a variety name, none of that could be read, so the spacing and sucker advice would have been wrong for the life of the tree. The tag is now split and each part filed where it belongs, in whatever order the label happens to put them. Two things it deliberately will not do. It will not guess — a tag that names no rootstock leaves the field empty rather than assuming a standard one, because the app acts on that field and we would rather know nothing than record something invented. And it only splits for plants that actually have a rootstock: eleven grafted tree fruits and citrus. That last part matters more than it sounds. English lavender has a variety genuinely called Munstead Dwarf, and a blanket rule that pulled the word "dwarf" out of every name would have fixed one plant by quietly corrupting another.
-
A question that knows when to stop
If you never answered “any sign of it yet?”, that question came back on every list for as long as the plant existed — we measured a bush bean whose season closes on day 60 still being asked on day 700. Nothing in the app decides a plant is finished on your behalf, and that is deliberate, but the question had no other way to stop, so “you have not answered” and “this is still worth asking” were the same thing. It now stops two weeks after the plant’s last stage closes — the same two weeks the overdue warnings use. Anything still inside its season is asked exactly as before.
-
Your fruit trees have timing after all — we were asking the wrong field
Ask GardenTrack when things will be ready and it works the dates out from your own planting dates and a sourced days-to-maturity figure. For a fruit tree, a citrus or a lavender it answered "no maturity window in our data" — which reads as a gap in our records, and was the opposite of the truth. Somebody had sourced those plants, decided a days-from-sowing count is meaningless for them, and written down why: a home orange is a named variety budded onto a rootstock, so nobody grows one from a pip. We reported that finished decision as a hole, on exactly the plants a gardener is least able to work out alone. We measured how far it went before touching anything. Of 152 plants, 41 carry no day count; 39 of those say outright that the field does not apply, and 39 hold the answer in the form perennials actually use, which is a picking season rather than a countdown. The two left over are cover crops nobody harvests. So there was no data to go and gather — the lookup was asking the wrong question. You now get the season: when apricots are ready around you, or for a navel orange a winter window that lands in the following year and says so. It is equally careful about what it is not telling you. A season repeats every year and says nothing about whether your particular tree is old enough to bear, and we do not hold how many years a young tree takes before its first crop — so it says that, rather than handing you a date that looks like a promise.
-
Journal entries now say where they came from
An entry the app read for you now ends with how it read it — “Aphids — from what you told the chat”, or “Seen when the plant was added — from your photo”. Anything without one of those is something you logged yourself. It is in the words rather than an icon, because a marker on every row clutters the thing you came to read and only works on the screen it was built for. Nothing was ever written without you agreeing to it first; this is about what you can tell six months later, reading back.
-
The plant identifier can now say the kind is unknown, instead of inventing one
Ask it about a plant and it names the plant, then tries to name the kind. It had no way to decline that second half. So when there was no honest answer it returned the nearest words it could see, and two of those went into our records as gaps in our catalog. English lavender came back with the kind "English" — the plant’s own name again. Nobody can tell Hidcote from Munstead from a photograph, and the honest answer was always "English lavender, kind unknown". A mandarin photographed while asking about an orange came back as a kind of orange called "mandarin". Mandarin is its own plant and we already hold it; it just could not say "this is not an orange". Both are now available answers, and it says what would settle it — a cut fruit, a flower close up, a leaf laid flat. There is a second half that matters more to us than to you: those two were filed as requests for kinds our catalog was missing. It was not missing them. Answering them as written would have added a lavender called "English" and an orange called "mandarin", and the next time we measured our coverage we would have read our own mistake back as evidence. Those records now say what they actually are. And the rule that spots them was checked against all 3,174 kinds we hold before it shipped, because one that wrongly rejected real ones would have been worse than no rule at all.
-
Ask a question about a plant you own, and it no longer offers to add it again
Nik asked the chat about one of his own plants and got a card offering to add that plant to his garden. The information was right there — the same chat, asked what to do today, had just named all three of his plants with their own watering state. It simply was not looking. Worse than redundant: tapping the card would have made a second record for one real plant, and from then on the jobs come twice and the journal splits across both. What it was actually checking was whether you had that plant’s page open on screen. Ask from the chat, where nothing is on screen, and it concluded you must be talking about a plant it had never heard of. It now checks the garden. Two ways, because your name for a plant is often not ours: by crop, so asking about your lavender finds your English Lavender, and by what you called it. That second one is what actually fixed his case — we hold Bush Bean as a crop and carry no wax variety at all, so his Wax Bean was findable only by his own words for it. A version that matched crops alone passed every test and would have shipped without fixing the thing it was written for. And it still adds the plant when you say you planted one: a second row of beans is a second row of beans. The card only goes away when you mention a plant you already have, which is what asking a question about it looks like.
-
"Not recommended" is an answer, and we were showing it as a blank
Clemson publishes an autumn planting chart for South Carolina, split into two regions. Most rows on it are dates. Three rows say something different for the Piedmont: not recommended. That is a real answer — the extension service looked at that region and said don’t plant it there — and it is more use to somebody in the Piedmont than any date on the row. We were storing it as nothing. A blank reads as “we never checked”, which is the opposite of what the source says, and it is exactly the difference this whole database exists to keep straight. Now the plant page prints it in words, in the same list, beside the regions that do get dates. Tomato, head lettuce and winter squash all say it, all for the Piedmont. Winter squash was the one worth catching: it had no autumn calendar at all, because its refusal had nowhere to go and its coastal dates went over the side with it — so a crop whose whole row had been read looked like a crop nobody had looked up. The thing that made this take a day rather than an hour is that three separate places were dropping it, and the one we knew about was only the first: the record could not hold it, the build would not carry it, and the page itself filtered out anything without a start and end date. Fixing only the first would have changed nothing you could see.
-
It stops offering you every plant you own for a plant you do not
Photograph something and ask the chat what it is, and it can offer to write the answer into that plant’s story. A beta tester photographed a sunflower he does not grow. The chat identified it correctly and offered to add it to his garden, which is right — and then underneath asked which of his plants to log the note against, listing an orange tree, a lavender and a wax bean. None of them is a sunflower. The app had just said what the plant was and then ignored its own answer. There are three cases and only one of them behaved. If the photo matches one plant you own, the note goes there and nothing is asked — that part already worked. If it matches a plant you have several of, you now get only those, which is a question worth asking. If it matches nothing you grow, you get no plant list at all, because the only sensible thing to do with a plant that is not in your garden is add it, and that button is already on screen. We did not just shorten the list. Five unrelated plants would be the same mistake more quietly, and at a hundred plants a short list is three wrong answers instead of ninety-nine. The question is whether a plant could be the one you mean. A second fault came out with this one: where you had two plantings of the same crop, the app picked the earlier one and wrote to it without asking. A record filed against the wrong plant looks exactly like one filed right, which is why that kind goes unnoticed for years.
-
Telling the chat what you see now actually does something — and it asks when it cannot tell which plant
The app has two ways to answer a question it asked you. Tap yes on the card on Today, or just say it in the chat: "the seedlings have their proper leaves." Both were meant to write it on the plant and put back the jobs that were waiting on it. Only one of them did. The chat version needed the conversation to be attached to one particular plant, and nothing in the app ever attached it, so saying it in the chat quietly did nothing at all. Every time. It now works out which plant you mean from what you actually said, matching against the plants you are growing — by our name for them and by yours. And when it genuinely cannot tell, it asks rather than picking. Two sowings of beans and you say "the beans are flowering" — it will name both and ask which, then write it down once you answer. That is not the app being timid. Guessing would put a job back on a plant that never showed anything and tell you it had done the right thing, which is the kind of wrong record you never find. Mention a plant you do not have and nothing is written at all — it would rather record nothing than record something true about the wrong plant.
-
The free allowance is back: ten photo answers and twenty questions, once
On 17 August we opened the free tier and said there was no limit on photos or questions for now. That has ended, and the allowance it always had is back: ten photo answers and twenty written questions, counted from the day you make your account, not refilling each month. The two are separate, and Profile tells you where you stand as a plain number — there is no bar or gauge, because watching a number fall is not something this app asks you to do. We would rather you spend all ten on the two weeks that actually matter than save them for a month that never comes, which is why it is a one-time grant and not a monthly trickle. Refusals and errors do not count against you. Everything that costs us nothing stays free forever — your plants, your dates, your journal, your reminders and the plant database — and writing in your journal never needs AI at all, so running out stops you asking questions, not recording your garden.
-
Ask if it is too late to sow something, and it counts back from your own frost
A single fall sowing date only means anything where the fall is long. The published planting tables we hold are one southern state’s, and printing them at somebody in Minnesota would be worse than saying nothing. So instead of looking a date up, it does the arithmetic the northern extension services print: count back from your own first frost by how long that plant needs. Ask whether it is too late to sow radishes and you get the date and the days left. Seven plants get a straight refusal instead, and that is the part worth explaining. For artichoke, broccoli, cabbage, cauliflower, sage, sweet potato and tomato, every source we hold measures the days from setting a young plant out — not from sowing a seed. Working a sowing date back from that would mean guessing how long the seedling stage takes, and being wrong there costs a season rather than a packet of seed. So it gives the set-out date and says plainly that it cannot give a sowing one. We would rather answer half the question honestly than all of it confidently. And for things you dig or cut, it says that sowing past the date still gives you something smaller rather than nothing — that date is not a cliff for every plant.
-
Choosing the plant now saves the note, and the note says something
Ask the chat about a photo and it can offer to write what it found into that plant’s story. Two things were wrong with that, and a beta tester found both in one go. First: when the app could not tell which plant the note was for, it showed your plants as a row of buttons — and tapping one did not save anything. It swapped that row for another button, in the same place and the same shape, and waited for a second tap nobody had been told about. Answering the question looked exactly like finishing it, so the note was never written. Choosing the plant now writes it there and then, which is what the rest of these buttons already did when there was only one plant it could be. Second, and stranger: the note could contain the word "note". The button is offered either when we tell you something is normal for the season, or when the plant simply looks fine. The first case always had words to save. The second had none, and the placeholder went into the journal instead. The button would not say so beforehand either — it read "Note this on your Wax Bean" and left you to find out. It now shows the sentence it is about to save, and when the answer carries no short summary of its own it saves the useful part of what was actually said. One more thing we tightened while we were in there: a save that cannot find a plant to write to now says so on screen. It used to return quietly, which is the beginning of an app telling you it did something it did not.
-
Tell the chat you planted the seeds, and it now answers the plan you made last night
If you add a plant with a date in the future, the app treats it as a plan and asks on the day whether it happened. That question could only be answered in one place — the card on Today. Say it in the chat instead and nothing connected the two. Nik did exactly that: he entered four beans one evening for the next day, then told the chat "just planted four of those seeds in that bucket and then we mulched" — and the only thing offered back was Mulched. One fact of three. The chat had no way to say the planting happened. It could offer to add the plant, which would have given him a fifth bean and left the plan still waiting, or write the word "planted" into the journal as a note, which records nothing the app can count from. That matters more than a tidy list: every date for that plant — when to expect it up, when it is ready — is counted from the day it actually went in, so until something confirms that day, all of them are guesses presented as facts. It now offers the real thing, and it is the same button as the card on Today, so answering in one place clears it from the other. If you said how many, it takes that too — and it will not overwrite a number you typed yourself.
-
A card that says keep them in no longer explains why they should go out
Hardening off is the ten-day ramp that gets an indoor-raised plant used to being outside, and its last night is the one instruction in GardenTrack that can kill a tray rather than cost you a day. When tonight’s low breaks a plant’s own cold floor the card refuses, and it refused correctly: Too cold today, keep the tray in today, 38 degrees tonight is too cold for this one, and no button offering to put them out. Open Why this underneath, though, and it carried on giving the day’s instructions — where to stand the tray, what time to bring it back in, and on the final night the worst of them: 38 degrees forecast tonight, which is above the line for these. That is the same number the card is refusing on, described as safe. The sentence was written for the night that passes and shown on any night we had a forecast at all, which includes every night cold enough to do the damage. A refused day now carries the refusal, the plant’s floor and the reason for it, and nothing that tells you to do the thing we just said not to. It still tells you which step you are on. Two things worth saying plainly: the check itself was never wrong — the headline, the temperature and the missing button were all right, and only the explanation underneath disagreed with them. And it was not only the last night. Every step of the ten has a bring-in time, so a refused day one was telling people to set a reminder for a tray that was staying indoors.
-
Ask about "tomatoes" and it now knows you mean tomatoes
Someone asked whether sweet potatoes with shoots on them could be planted. The chat said we had not written that up. We had: our record for sweet potato says it is grown from cuttings, sourced from Utah State and NC State, and shoots on a stored root are exactly what those cuttings are. The answer existed and nothing handed it to the chat. All 152 of our plants carry how they are started — from seed, from cuttings, from cloves, grafted, divided — and none of it was reaching the conversation. It does now. Chasing that turned up something bigger. When you named a plant, it looked for the exact name in our list, so "tomato" matched and "tomatoes" did not. That was not a small gap: 136 of our 152 plants could not be found by their own plural. "Sweet potatoes" happened to work only because that is literally how sweet potato is written in our list, which is why it looked handled. So a question like "can I still grow tomatoes this year" reached the part that works out whether there is time — and did not know which plant you meant. Plural and singular both work now, everywhere that reads what you typed. One limit worth knowing: what it tells you is what our record holds, not everything possible. A method we have not written down means we have not written it down, and it says so rather than letting you read it as "you cannot".