Skip to content
GardenTrack GardenTrack
Start free

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.

  1. A pepper you grew from seed and one you bought are thirty days apart

    Ask the app when things will be ready and it counts forward from the day you planted, using a sourced figure for how long that crop takes. The catch is what the figure is counted FROM. A pepper takes 100 to 120 days from sowing a seed, or 70 to 85 from setting out a plant you bought — the difference is the six or seven weeks somebody else spent raising it. We hold both numbers for 56 of our crops, and the answer was using whichever one the crop record happened to list first, for everybody. If you had sown pepper seed you were told it would be ready a month before it will be. It now picks based on what YOU did, because that is the only place the answer can come from — no amount of cleverness in the plant data can know whether you sowed a seed or bought a plant. Sow it and you get the from-seed figure; buy it and you get the from-transplant one. The sentence says which, so you can see the reasoning rather than just the date. And where we only hold one of the two numbers, it still says plainly which one it used rather than quietly counting from the wrong event.

  2. A basil sown yesterday no longer says "Ready to pick"

    Sow basil from seed, look at your Garden screen, and it said **Ready to pick** — with nothing above the soil and sixty days to go.

    **Radish sown the same afternoon was right**, which is what made it findable. The difference is in our own plant records: radish measures every stage from the day you sow it, and basil measures its later stages from the day a seedling gets planted out. **Sow basil straight into the ground and that day never happens**, so those stages fell back to the typical calendar for the crop — and in late August, the typical basil is ready.

    **The app now refuses to say a plant is at a stage it cannot possibly have reached.** If a seed has not been in the ground long enough to have come up, nothing about picking it can be true, whatever the calendar says about the season.

    **What this does not fix yet**, and it is worth saying: for those crops the app still cannot tell you when your own basil will be ready. It stops guessing wrongly; it does not yet know. That is the other half and it is open.

  3. A tree planted last year was still being told to settle in

    The app knows how many seasons ago each plant went in, and it uses that to stop giving you first-year advice for ever — stake the young whip, keep it watered every few days, strip the first fruit so it builds structure. All of that is right in year one and wrong every year after. It turned out the setting worked for a plant two or more years old and did nothing at all for a plant exactly one year old. Read ten days either side of its planting date, 151 of our 152 crops gave a one-year-old plant word-for-word the same advice as one put in that morning. The cause was two corrections of the same size pointing opposite ways. One pushed the plant's calendar back a year, correctly. The other, written for a different problem entirely, pulled today back a year to find which season a date belongs to — and landed right back inside the window that had just been moved away. They canceled out exactly, which is why it looked like the setting simply had no effect. Now a date you gave us is taken at face value and never second-guessed; only a date we had to estimate from the typical calendar for that crop gets adjusted, because that is the only kind missing a year. Fixing it turned up a second thing we were not looking for. Parsley flowers in its second summer, not its first, and the same guesswork was telling a fifteen-day-old parsley seedling that it was flowering — a stage a full year away, reported as happening now. Three crops still repeat a first-year stage they have finished, and we have left those alone deliberately: the obvious fix silences fennel and scallion for the entire year, and a living plant we have nothing to say about is the worse of the two problems.

  4. If somebody answers your report, the answer turns up on your report

    The pencil button has always worked one way. You wrote what was wrong, it reached us, and that was the end of what you could see — the screen could honestly tell you the note had arrived and nothing more, because there was no way for anything to come back. Now there is. An answer lands on your own report, in Profile, sitting under the words you wrote, with a name and a date on it. Nothing goes to your email and nothing needs an app to be open. Two honest limits worth saying out loud. **Most notes will never get a reply**, and that is not the app failing you: a bug that gets fixed does not need an answer, and you will see the fix rather than a message about it. And **you cannot write back to a reply yet** — it lands, and the conversation stops there. If you have more to say, send another note from the same pencil button and it stays in one place. The line at the bottom of that screen promising that anything needing an answer would come by email is gone, because it is no longer true.

  5. "Any bean sprouts?" — the app can finally ask, and listen to the answer

    Every plant carries a rough day count for when its first seed leaves should be up. When yours goes past it, the app has a real doubt — a cold spring, deep sowing or old seed all push it later — and until now it just kept counting silently, because there was no way for you to tell it. That is the odd part: the app has always known what to do with the answer. Say a plant has come up and it stops wondering, stops nagging, and treats that plant as further along. It has been able to do that since the day it was built. Nothing ever asked the question. It asks now, once, naming the crop and the thing you would actually look for: "Any bean sprouts?" rather than "anything up yet". The difference matters more than it looks — the second one is really the app asking whether its own records are stale, and leaving you to work out what to go and look at. Saying yes ends the question for that plant for good. Saying not yet records nothing at all, deliberately: there is no such thing as a plant that failed to come up on schedule as far as we are concerned, and we are not going to write one down on your behalf. Some plants will never ask, and that is on purpose. The question only exists where there is a yes-or-no thing to look at — seed leaves, flower buds, tiny fruit forming. Where the plant record is waiting on something like "growing fast", there is no answer you could give, so we stay quiet rather than putting a question mark after a phrase and calling it a question. Second thing, same idea. If you add a plant with today's date, the app now asks you to plant it. That sounds too obvious to mention, and it is exactly the job the app should never be silent about — but it was, for 149 of our 152 crops, because the reminder depended on a quirk of each plant's record rather than on the date you gave us. Add beans the night before and you would open the app in the morning, standing over the bed, and see nothing about beans. Ticking it is also what starts that plant's clock, so every date afterwards is counted from the day it really went in.

  6. Courgette, aubergine, capsicum, rocket and coriander now find the plant

    Type **courgette** into add-a-plant and you got nothing. Same for **aubergine**, **capsicum**, **rocket** and **coriander** — the names these crops are sold under across the UK, Ireland, Australia, New Zealand and much of Europe.

    **We had every one of them all along**, under the American name, with varieties. The search was only ever looking at one name per crop.

    All six now resolve, including **swede**. If you know a crop by a name we still miss, tell us with the pencil button — that list is finishable and we would rather be told than guess at it.

    **Misspellings are still a miss** — type *brocolli* and you get nothing. That is a different fix and it is next.

  7. Saying your tomatoes have gone orange now offers to log the harvest

    Tell the chat "my Sungolds have gone properly orange, are they ready to pick?" and you would get a good answer and no button. Describe the same plant using the word ripe and the offer to log a harvest appeared. That is a distinction no gardener should have to know about. The reason is worth explaining, because the app was doing something right and then dropping it: when you describe what you can SEE, the chat records it as an observation — "the fruit is coloring" — which is more precise than the word ripe and is the right thing to keep. But the harvest button was only watching for the word. It now reads the observation as well, through your crop's own stages: orange fruit on a tomato falls between ripening and harvest window, and since both of those mean pick it, the answer is the same either way. Where the two possible stages would disagree, it still says nothing rather than guessing. And the part that protects you is unchanged: the app checks its own calendar for your garden first. If your plant is nowhere near its harvest window, describing it as ripe gets you no button — the chat does not get to decide that on its own, because only the app knows when the plant went in.

  8. Four plants in the list had no picture at all, and it had nothing to do with your signal

    Nik went out to the garden with his phone in aeroplane mode and found that a lot of plant pictures were missing. Most of that is about what the app downloads when you install it, and that is being fixed properly. But four of them were not missing because of your signal — cucamelon, lentil, nasturtium and sweet alyssum had no picture at all, on any device, online or off. There were 148 pictures for 152 plants and nobody had ever counted. All four now have one, drawn in the same faceted style as the rest. The lentil took three goes: the first was a brown dome that looked like every other brown dome in the list, so it is now a split red lentil you could not mistake for a potato.

  9. The app stops assuming your seed came up

    Sow a kale in the ground and leave it a month, and the app used to tell you to weed around the kale and check the kale's leaves. **It had never asked whether the kale came up.** Every job it gave you rested on the assumption that a seed became a plant, three weeks after a one-week germination window, with nothing anywhere checking.

    **Now it asks, and that question is the job** — not a note beside the list. While it is waiting, anything that talks about a part of the plant is held back: no checking leaves that may not exist, no thinning seedlings that may not have come up.

    **Watering never stops.** A dry seedbed is the main reason seed fails, so that one is never held back for any reason. **Weeding does not stop either** — it gets more important while you wait, because weeds pull easiest small and take the water your seedling needs the moment it emerges. But it says something different now: *keep the row clear, go gently, work the edges*. You cannot tell a seedling from a weed yet, and rooting about disturbs the seed. That is real advice for the state you are actually in, not the usual advice with the plant's name taken out.

    **And if you never answer, nothing goes quiet.** Once it is past the point where that crop should have shown, the app assumes it came up and gives you the full list again — still asking, quietly, at the bottom. A pointless question costs you a glance. A real plant going silent for a season would cost you the plant.

  10. It asked whether your seeds were up nineteen minutes after you sowed them

    Sow broccoli at 9:35 in the evening and the app put "Any sign of the broccoli yet?" on your list at 9:54, suggesting a quick look at the row would settle it. In the dark, nineteen minutes after the seed went in. The same thing happened with radish and basil, so it was the rule rather than one crop with odd data. What makes it worse is that the app already knew better and was saying so on another screen: open that broccoli and its own page reads "First sprouts in 5-10 days". We were asking a question our own records said could not have a yes for four more days. The question now waits until the earliest day this crop could actually show — the same figure the plant page is quoting, so the two can no longer disagree. It waits for the earliest day rather than the average, because the fast half of a sowing is real and staying silent through it would be its own kind of wrong. And a crop with no sprouting window in our records is not held back at all: not knowing when something sprouts is not evidence that it sprouts instantly.

  11. Add something you are planting next month, and the app gave you jobs for it today

    You can add a plant with a date in the future — next weekend, next month — and plan the bed before you dig it. The app was then treating that plan as a plant. Across every crop we hold, 68 of 152 crops handed you work today for something dated thirty days out, and 42 of them told you to harvest it. It would also tell you the seeds never came up: 566 different ways to be informed that a plant still in its packet was overdue, one of them 191 days overdue. All of that is now silent until the day you said you were planting, and on that day everything starts properly, including the reminder to actually go and plant it. Three separate things were causing it, and only one was obvious. A plant record carries a few dates — when you sowed it, when you set it out — and where you have not given us one, the app falls back to the typical calendar for that crop across the country. So half of a plant could be running on your date and the other half on a national average that knew nothing about you. The second was arithmetic: an asparagus bed has an establishment period two years long, and the code that works out which year you are looking at was quietly rolling forward into the next one to make today fit inside it — so the app believed you were 335 days past a planting you had not done. The third was our own first attempt at the fix, which silenced everything for a future planting and would have stopped telling you to water a tray of seedlings during the two weeks you spend hardening them off. A transplant date in the future means the plant has not been moved yet, not that it does not exist. The chart still shows you the whole year ahead for anything you are planning — that part was always the useful bit.

  12. Every invited tester was told they had no questions left. They never had a limit at all

    If you were invited to the beta and opened Talk, it very likely said **"You have used all your free questions."** Profile said something like *6 photos and 0 questions left*. Neither was true. A beta account has never had a cap, and the server would have answered every one of those questions — **nothing was ever blocked.** You were just told it was, which stops people just as effectively, because nobody types a question after being told they have none left. What went wrong is small and worth explaining. Your phone keeps the last count the server sent it, so the number survives a reload. When an account is given no limit the server sends no count — correctly, there is nothing to send — and the old number was never cleared. So the last thing you were told while you had a limit stayed on your phone after you stopped having one. **An account with no limit now says nothing at all.** Not "no limit", not a big number — nothing, because a ceiling is only worth a sentence when there is one. And while we are in beta that is everybody: the founder has opened the free and no-account tier to the same limits as paid for now, so nobody meets a wall while we are still finding out what this should be.

  13. "Nothing logged" to somebody who had just done everything

    Work through every card in garden mode and the closing screen could still say "Nothing logged." The count was honest — notices have nothing to write down, and answering "not yet" records nothing on purpose — but it read like a score for the morning.

    Get to the end having dealt with every card and it now says so, and explains why the journal is empty instead of leaving you wondering where your answers went. The same mistake sat one line further on: acknowledging three notices told you "3 left for later" when there were none.

  14. A basil sown yesterday says when it will sprout, not "nothing scheduled"

    Earlier today we stopped the app claiming a seed sown yesterday was ready to pick. **For a few crops that left the Garden screen saying "nothing scheduled" instead** — next to two identical sowings correctly saying *Sprouting in 4d*.

    **That is our word for the state of our list, not a fact about your plant.** The basil had a sprouting date four days out the whole time; the screen had simply worked out what to show before it worked out what to hide.

    **Fixed, and it is worth naming what happened**: one wrong answer was traded for another, and the second was quieter, so it would have lasted longer. If a screen ever tells you there is nothing to say about a plant you can see growing, that is us being wrong — tell us with the pencil button.

  15. Setting your location now works with no signal, and the number beside Today means today

    Two fixes from a day of testing on real phones.

    **You can set your location without a signal.** The app turns a place into your frost dates using a file of map coordinates, and that file was being left out of the download by mistake — a rule written to keep the plant lists out had swept it up with them. So somebody who installed the app, walked out to the garden and lost signal could not finish the very first step. **The whole point of this app is that it works out there**, and it fell over in exactly the place it exists for. It is now downloaded with the app.

    **And the number beside Today is today's.** It was counting everything on the screen — jobs that surfaced today but can wait were counted under Today and then listed again under This week with their own number. One garden read four when it meant one; another read eleven when it meant three. **Eleven jobs and three jobs are different mornings.** The heading now counts only what is actually today, and every row is counted once.

  16. Profile and Talk went blank for signed-in accounts, and are back

    If you were signed in this afternoon, Profile and Talk showed nothing at all — not even the tabs along the bottom. **That was ours and it was live for a few hours.** We put the previous version back the moment it was reported, and this is the proper fix.

    **What happened.** Both screens ask "how much of your allowance is left?". That question was also quietly tidying up an out-of-date answer stored on your phone — and on these two screens, saving something is what makes them ask again. So they asked, tidied, asked, tidied, and never got as far as drawing anything. The tidying now happens when your plan actually changes, which is once, instead of every time a screen looks.

    **Why we did not catch it.** It could only happen once you were signed in — signed out, there is no plan to check and the whole thing never runs. Everything we test automatically runs signed out. **So this could not have been caught before somebody opened it**, which is a gap in how we check our work rather than bad luck, and we are closing it.

  17. The sign-out button now says what signing out actually does

    Next to **Sign out on this device**, we told you *"your garden stays here either way."* **That was not true.** Signing out clears your garden off the phone — every plant, bed, note and photo — and it always has.

    **The behavior is right and has not changed.** If you hand your phone to somebody, or sign in on a shared tablet, the next person must not be able to read your garden. Your account keeps everything and signing back in downloads it all again.

    **It was the sentence that was wrong, and wrong in the worst direction** — it is exactly what you read right before signing out casually to check something, and it removed the caution the moment deserves. It now says the garden leaves the phone, and that it comes back.

    **And if anything has not reached your account yet, it tells you that too.** Sign out while offline and whatever is still waiting to upload goes with the wipe. That warning only appears when there is genuinely something waiting.

  18. Telling the chat something is broken now actually reaches us

    If you told the chat that something in the app was not working, it answered you kindly and then nothing happened. Not "we read it later" — nothing. The message was never filed anywhere we would see it. That is fixed: say a button does nothing, a number looks wrong, a plant has vanished, and it lands in the same place a bug report from the report form lands, with your own words kept exactly as you typed them. You do not have to find a form, pick a category, or know that a form exists. Nothing changes on your screen — you get a normal answer, and the report happens quietly behind it. The part that took the care was teaching it the difference between "this is broken" about the app and "this is broken" about a plant, because people use the same words for both. "I planted beans two weeks ago and nothing has happened" is gardening. "I tapped save and nothing happened" is us. The chat is not deciding whether you have found a real bug or how serious it is — it only says which of the two you were talking about, which is the one thing it can see and we cannot. Everything after that is our job.

  19. The app now says which jobs it is holding back, by name

    When the daily list asks whether a seed has come up, it explains what it is staying quiet about while it waits — and it names the actual jobs rather than saying "some tasks are paused". The sentence is built from your plant, so it says two jobs for one and four for another, and says nothing at all when nothing is being held back.

    There is a second kind of waiting, where a job is held until the first proper leaves show. Answering that one does not yet release the job, so **the app does not ask about it** — a question whose answer changes nothing is worse than no question.

  20. The daily list knew the tree was too young. The calendar beside it did not.

    Every plant page shows you two things: what to do today, and the handful of dates that make up its year. Only the first was checking how old the plant is. So a three-year-old pear got a correct daily list that said nothing about fruit, and a calendar right beside it booking a fruit-thinning for June. Seven plants did this — apricot, asparagus, grape, kiwi, peach, pear and rhubarb. Asparagus was the one worth hurrying for: it was scheduling a spear-cutting in the establishment years, and cutting those first spears is how a bed that should crop for twenty years gets ruined. The cause was small and easy to miss — the two lists were being handed two different copies of the plant record, and only one of them carried the age. They now read the same copy, so they cannot disagree again. A second gap went with it: the check knew about ripening and picking but not about the stage where the fruit first forms, which is the one that carries the thinning job. Twenty-three plants were affected. And then the fix gave something back that we had not gone looking for. The old check hid a young tree’s entire late-summer stage — and on an established tree that stage is the only one still open, so it was taking the watering reminders with it. A three-year-old pear, peach or persimmon was being told nothing at all about water, for the whole year, which is the opposite of what a young tree needs. Now only the jobs that need fruit are held back. Pear went from 47 days of watering advice a year to 138.

  21. The chat and the This week screen were answering the same question two different ways

    Same tomato, same afternoon. The This week screen said "will not finish before your frost — it would start ripening 14 days after your first frost". The chat, asked about the same plant an hour earlier, said "you are tight but you have a real window" and then "fifty-two days is plenty of time for early fruit, and you might be surprised". One of those says ripening starts two weeks after the plant is dead. If you read both, you cannot act on either — and whether the thing you just planted will give you anything is the question this whole app exists to help with. The cause was not that the two numbers disagreed. The app already had a single answer: it works out, for every plant you have, whether it finishes before your frost, using your dates and your sowing day. The screen was reading that answer. The chat was not — it was handed the raw growing times instead and did the arithmetic itself, taking the optimistic end of a 52-to-90-day range. Now the chat is given the same verdict the screen shows you, in the same words, and told plainly that it is settled and not to work it out again. Tested against the real complaint: without that, three tries out of three produced the encouraging answer, including the words "plenty of time". With it, three out of three say it will not finish. And for the genuinely close cases the app now says both halves — it can work, and an early frost ends it — rather than picking the cheerful half.

  22. The chat described a watering screen that does not exist, and every number in it was real

    Someone asked the chat about watering and was told to press buttons for "one, two, three, six, twelve or twenty-five gallons". They went and looked: there are no buttons. Watering is the one job the app deliberately never asks an amount for, because nobody standing there with a hose knows whether they put on one gallon or two, and a guessed number is worse than a blank one. The interesting part is that the chat had not made the numbers up. One, two, three, six, twelve and twenty-five are the real buttons — for a HARVEST, counting fruit, which is a different screen entirely. The chat took a true row from one part of the app, moved it to another, and changed the unit from fruit to gallons. Everything in the answer was ours. Only the arrangement was invented, and that is worse than a guess rather than better, because an answer built out of real parts reads exactly like knowing. So there is now a seventh rule in the list of things it never does: never carry a fact about one part of the app across to another. If our own writing does not cover the screen you are actually looking at, it has to say it is not sure what that screen looks like and ask you what you see. We would rather it asked than described. But the rule on its own only fixed it two times in three, which is not a fix — one roll in three is a gardener. What actually closed it was taking the six numbers out of what the chat is given at all. They were in a help page that named them and ruled watering out in the same breath, and that turns out to be exactly the shape that gets half-remembered. It cannot move a number it was never handed. Twelve tries since, all clean.

  23. Every evening the chat thought it was tomorrow, and argued with you about it

    From about six in the evening in Colorado — earlier on the west coast — the chat was told the wrong date. Not a display problem: it reasoned from it. Somebody said the watering advice was wrong and the chat disagreed, explaining that they had planted everything "yesterday, the 17th" and watered "twice yesterday and again this morning". All of it had happened that same afternoon. The person was right and the app talked them out of it using history it had invented, which is worse than not answering. The cause is that the server built every date in UTC, and after six in the evening in Denver it is already tomorrow in UTC. So one afternoon got split across two days: a watering at 4:27 read as today and one at 7:22 read as tomorrow. Every interval it quoted was off by one too — days since planting, days to harvest, days to frost. Now your phone tells the server which timezone you are in, and every date in the conversation is the date on your wall. "Two days ago" also means two dates ago rather than forty-eight hours, which is what a person means by it. If you ever see the chat and the Today screen disagree about what day it is, that is a bug worth telling us about — and telling the chat now reaches us.

  24. The chat invented a week away that never happened

    Somebody who had been in the app all evening — planting things, watering things — asked "what did I miss?" and the chat opened with "You were away from the garden yesterday and today is the first day back. Let me walk you through what happened while you were gone." There had been no absence. Everything on the list had gone in that afternoon. We first read this as the date bug we fixed the same day, and it was not. Fixing the dates and asking again still produced "You have been away for a bit." The real cause is worth explaining, because it is the thing to watch for in any app like this. Hand a fluent model a record it cannot make sense of and it will not say so — it will invent whatever makes the contradiction disappear, and here that was a gardener who had gone somewhere. It is a confident, friendly, completely fictional explanation, and the person it hurts most is the one who genuinely has been away for a week, opening this screen for real, with no way to tell the true half from the invented half. So there is now a standing rule: the chat will never tell you that you have been away, that you are back, or that you have not looked at something in a while. It cannot know any of that. Nothing here records when you last opened the app or last stood in your garden — and a gap in the written record is not a gap in your attention, because almost everything a person does to a plant never gets written down. "What did I miss?" is a question about the garden, not a confession.

  25. A check that quietly stopped checking every evening

    The app has a guard that catches an impossible date: a seed cannot come up before it was sown. Log a sprout for a bean you planted next week and it should stop you and ask which of the two dates is wrong. It was doing that all morning and quietly giving up every evening. The reason is the same one behind yesterday's chat fix, in a different place. The guard worked out what day the sprouting happened on using UTC — the clock at Greenwich — and then compared it against a sowing date recorded in your own local time. For most of the day those agree. After about six in the evening in Colorado they do not, because it is already tomorrow in London, so the sprouting date came out one day too high and stepped over the line it was supposed to stop at. Nothing broke and nothing showed a warning. It simply let the impossible entry through, and only after dark. It now uses the same date math as the rest of the app, which is the whole point of having one copy of it. We also taught our own build checks to spot this specific mistake, so the next one gets caught before it ships rather than after — and that check found this one.

  26. The free tier is open — no limit on photos or questions for now

    A free account is not being metered at the moment. Ask as many questions as you like, send as many photos as you like, and nothing will stop you halfway through working out what is wrong with a plant.

    **This is a decision for now, not forever, and we would rather say that than let you find out later.** The design is a one-time grant — ten photo answers and twenty questions, not refilling each month — and it is switched off rather than removed. When it comes back the help pages and the app will say so first. **You will not discover it by being refused.**

    **Profile now says nothing about limits**, which is the only honest thing it can say while there are none. It used to print whatever figure was last left on your phone, which is how a beta tester opened Talk this morning and was told they had no questions left on a plan that has never had a ceiling.

    Everything that costs us nothing stays free regardless — your plants, your beds, your dates, your journal, your reminders, and the whole plant database.

  27. The invite box only appears if you actually need one

    Signing in asked everybody for an invite code, and read it from almost nobody. The code is only checked when an email address has no account yet — that is the moment it creates one — so for every gardener signing in for the second time or the twentieth, it was a box that took your typing and quietly dropped it. Somebody half-remembering that they were told they needed an invite would go and find the code, type it in, and get no sign either way that it mattered. A box that ignores you is worse than no box.

    **Now it is not there until it is needed.** Put your email in, type the six digits we send, and if your address is new the app asks for the invite code right then, with the box already focused. If you already have an account you never see it at all.

    **We could not simply look your address up**, and deliberately so: an app that will tell you whether an email is registered will tell anyone, about anyone. So the app does not ask that question. It waits until you have typed the code we emailed you — which means you are reading that inbox — and only then does it say an invite is needed.

    **And six digits now signs you in on its own again.** The screen submitted by itself once you finished typing the code, but only if the invite box had something in it, which meant the one person who correctly left it empty was the one made to hunt for the button. Every time. That was backwards and it is fixed.

  28. It stops asking whether the radish came up once the radish is ready to pull

    Earlier today the app stopped assuming a seed had come up, and started asking instead. That was right, and it went one step too far on a few crops.

    **A radish is ready about three weeks after sowing.** But the point where we would normally stop waiting and assume it must be up is a week later than that. So for that week, if you never answered the question, you had a radish ready to pull and an app that would not mention it — the exact thing we were fixing, running the other way.

    **Now the question stops as soon as your own dates say the crop is pickable.** Not the calendar's dates — yours. If you sowed it and the time it takes has passed, that is a better reason to stop waiting than any average.

    **Where the timing does not come from your dates, nothing changed.** Some plants have picking seasons worked out from the local frost rather than from when you sowed — sage and anise hyssop among them. A season being open says something about the year, not about whether your seed came up, so those still wait for the answer or for the usual point where we assume it.

  29. A crop page stops saying it is too late for a plant you already have

    Open a crop you are already growing and the page led with **"Too late this year"** — about something already in the ground, sometimes germinated an hour earlier.

    **It now leads with your plant.** *You have one of these growing*, with a link straight to it. The sowing window is still there, below, asking whether you were thinking of sowing another — because a second row is a real thing to want and we should not delete the answer.

    **And the sentence was claiming more than we had checked.** That verdict is worked out from a crop's spring sowing dates, so *too late this year* was a statement about the whole year made from one window — and plenty of crops take a second sowing in late summer that those dates say nothing about. It now says **the spring window closed around…**, which is what we actually looked at.

  30. The growth stage saved with every journal entry was about a typical plant, not yours

    Every time you log something, the app quietly writes down roughly where that plant should have been in its life on that date. The point of it is comparison: months later you can see that your tomato is running three weeks behind what is normal, which is genuinely the most useful thing this app could notice. It was working that stage out from your frost dates and nothing else — not the day you actually sowed the thing, not how many seasons it had been in the ground. So a tomato you sowed in late July was recorded with the stage a tomato sown in May would be in. About a third of the entries carried a stage the plant was not in. On some crops it was worse than wrong, it was missing: basil, broccoli and cabbage recorded no stage at all for the whole of January to September, because their stages are counted from the day you sow and the sowing date was the one thing not being used. Across the catalog that was 68 crops recording nothing for nine months. It is now 4, and those four are ones we genuinely cannot place. We have not gone back and repaired the old entries, and that is deliberate rather than laziness. What makes the stamp worth anything is that it was written at the time. Recalculate a season of them from the planting date and what you are looking at is our own model checked against itself, which tells you nothing about your plant. So the old ones stay as they are and the new ones are right — which is exactly why we did this one before several more visible bugs. Every day it waited wrote more records that could never be corrected.