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.
-
Your garden, zone by zone
A few more pieces landed on top of today’s nesting. A breadcrumb sits under the header the whole time — Garden, or Garden › your zone’s name once you are in one — and every part of it but the one you are on is a tap back up.
Once your garden holds a dozen plants or more, a search field sits at the top of the Garden screen; below that it is a small button so a smaller garden pays nothing for it. Type a plant’s name, its crop, or where it lives, and every match comes back as a flat list with the path to it — tap one and you are on that plant.
And the zone’s own line stopped being a percent. It now reads the way you would say it: “Bean Pot is dry”, “Bean Pot empties tomorrow”, “all fine”, or nothing at all when there is nothing to track yet. The gauge itself stayed on the plant’s own page, one tap in.
Renaming or removing a zone now happens right there on the zone: a ⋯ beside its name, with Remove zone giving you a minute to undo. Moving a growing space to a different zone is the same ⋯, on the space’s own row, or a long press on it.
-
Your walk comes back as a letter
The garden walk’s report was a plain list, and it could say three different things about the same walk at once. It is rebuilt today as a letter from your garden.
It opens on the best thing the walk saw — a photo, one line under it, and how many stops were read, never a number first. No usable photo, and it opens on the words instead: what evening it was, how many stops, how many read — still never a stale image standing in for the walk.
Then, in order: what needs you, named plainly, with a job you can add in one tap; what moved on; what we could not tell, with why, for every stop the walk could not read — said once, never repeated as a card for every place that was not on the walk; and what is coming from the forecast. The letter never says nothing needs you unless every stop it read agrees — the report used to say that over stops nobody had read yet, and it does not any more.
-
Hanging baskets, wall planters and towers — and what holds a plant up
Two things the app could not say until today. A pot can now say where it sits — hanging basket, wall planter or stacked tower — under its Details, next to its size. Its page says “Hanging basket” in the heading, and its watering job explains that a basket loses water through its sides as well as its bottom, so it dries faster than a pot on the ground; if one has dried right out, stand it in a tub of water for an hour or two, never longer.
And a plant can say what holds it up — stakes, a cage, a trellis, an arch, or nothing — under its own Details. That does a job: the “set a stake or cage while the plant is small” task is done the moment you name one, everywhere it shows. It lives on the plant rather than the bed because the plant is what climbs.
-
A crop’s page, and What Rafe did, in the app’s own style
Open a crop from Add a plant and its page now looks like the rest of the garden: the crop’s picture beside its name, the “can I plant this now?” answer as a card that glows amber when the answer is yes, and its facts — spacing, pot size, sun, water — laid out the way the season page shows yours. What Rafe did, in the journal, now reads as the journal’s own rows, with Undo as a button on each. And on Add or log, the “Write it down” button no longer sits over the note box on a phone before you have picked a plant.
-
A pot can say what it is made of, whether it drains, and whether it waters itself
Every pot used to get the same watering advice, ending in “water until it runs out of the drainage holes.” A pot with no hole can’t do that — and that instruction is exactly how such a pot drowns its plant. So a pot’s own page now asks three optional questions next to its size: what it’s made of, whether it has a drainage hole, and whether it’s self-watering.
Answer them and the job changes. No hole: water a little at a time until the top two inches are damp, then stop, and give it less than a pot that drains. Self-watering: check the reservoir instead of the soil, every couple of days, and fill it through the tube when it’s low. Terracotta, fabric or wood: still looked at daily, but the reason now says these dry out faster than plastic or glazed, so look sooner in heat.
Leave a question unanswered and nothing changes — the app never guesses a pot has no hole. What you say stays with the pot on every phone you sign in on.
-
Setup is five quick things — the plan and your account come first
Opening GardenTrack for the first time now starts with picking a plan and making your account, and only then asks about your garden. So the setup run no longer ends by asking which plan you want — it was the same question twice. It is five screens now: where you garden, your dates, one plant, its spot, and what the app already knows, drawn from sunrise to one dark night at the end.
-
Calendar is its own page now — right on the bar at the bottom
Your season, your tasks and what still fits used to live stacked on one page you had to scroll to find. They’re three views on their own page now: **Calendar**, on the bar at the bottom, in the spot the journal tab used to sit. The journal didn’t go anywhere — it moved into the menu, one tap from the leaf at the top right.
Calendar opens on **Week**: a grid you scroll sideways, a column per week with its date on it, your plantings drawn as bars grouped by bed, this week’s forecast sitting right over today’s column. Grab a planned bar’s handle and drag it to a new day — the crop’s own planting window lights up as you go, the same as it always has.
The grid’s own head carries three more controls. A **Weeks | Season** toggle zooms the same bars out to the whole year at once, no scrolling. **What still fits** opens as a sheet next to it — every bed, how much room is left, what would still finish before frost — and a bed with room to spare says so right on its own row too. **Filter** holds sort and show behind one button, with a badge when they’re off their defaults. Two chips sit next to Week: **Month** is a wall calendar of the days something happens — a plant going in, a first picking, a job due — tap a day for its list. **Tasks** is a task list: search at the top, then Overdue, Today, This week and Later, one line per job with a tick on every one you can do in a tap — the same tap Today uses — the crop and the bed as tags, and Done with all of these on the overdue pile. And on the week grid, the … on a row opens **Add succession**: every so many weeks, how many more, the dates previewed, one tap to plan them all with one undo.
A tap on a bar opens a sheet now instead of jumping straight to the plant: its bed and its own date, **Stage now** and **Next stage**, and its **Next job** with the same tick your task list uses. **Open the plant** and **Ask about this** sit underneath, so the plant’s own page is still one tap away.
A round **+** at the end of Week · Month · Tasks opens Talk with "Plan " already typed, cursor waiting — finish it with something like "tomatoes in Bed 1 for October 3" and it lands on the grid outlined, since nothing’s gone in yet. On Month, a day still to come with nothing on it says "Nothing yet — tell Rafe what to put in ›", the same door for that day.
Each row carries a photo now, not just a name. The Week grid’s name column and Month’s day list both lead with your own latest journal photo of that plant — the same picture its own page and the garden pages already use. A plant with no photo yet keeps the small crop mark it always had.
And from a first red-team pass the same day: dragging a planned bar to a day the grid calls too late now snaps it back and says which day it kept, instead of saving it anyway; a filter that matches nothing says so; a row menu near the bottom of the screen opens upward; and a garden of only perennials is told where they live (the Year Wheel) rather than to add a plant.
-
A whole plan for the rest of your season, priced before you plant it
There’s a new page, **Garden Planner**, in the menu next to the leaf at the top right — or open it straight from the Calendar’s What still fits sheet, the line at the bottom that reads “Plan the whole rest of the year ›.”
It opens on **The rest of this year**: how many days you’ve got left to first frost, and about what date that lands on. Three chips underneath — **Most plants**, **Least work**, **A bit of both** — flip between them and the whole plan below remakes itself on the spot. A bit of both is the default.
Above the plan sits its price, so you see the cost before you commit to anything: how many jobs, about how many hours a week that comes to — a floor and a ceiling, never a bare number — the very first job coming up, and the single busiest week. All of it comes from your own places and your own frost dates, run through the app’s own arithmetic. No AI decides what goes where.
Below that, every place you grow in gets its own line: the crop, when to sow it or plant it out, when it starts to pick, and “tight” on the ones where the day falls late in that crop’s window. A place with nothing new to add says why in one line instead of showing nothing — full, nothing left that would finish, or no frost window to check against. One button at the foot, **Plan it**, writes every planting on the screen at once — outlined on the Calendar until you confirm it’s in the ground — with one Undo for the whole batch.
No beds yet? The Garden Planner opens on **Let’s build your garden** instead of a plan: pick what you have — raised beds, in-ground beds, pots and containers, seed trays — say how many, then walk through each one: a name, a size, how much sun, where it sits, what’s already growing there. Each place gets written the moment you finish it, so the plan starts filling in as you go, and “Rather just say it? Ask” is one tap away on every step, chat already started from what you’ve picked. All of it works offline.
Don’t like a crop the plan picked? Tap **swap** on its row. You get up to three others that still finish in that place, each with what it changes in the plan’s own units — “5 fewer jobs”, “about ¼ hour a week more” — and “Nothing — the room stays.” Pick one and the price on top moves with it; “undo swap” puts it back. Nothing is saved until you tap Plan it.
And when you tell Rafe about a new bed or pot, he now keeps the sun you mention — full sun, part sun, or shade — on the record, and “two pots on the porch” become two pots, each with its own name, so you can put different things in each.
And a plan is something you can keep now. **Save for later** puts it on **Your plans** — posture, jobs, hours, the day you saved it — with nothing planted. Open one again and it remakes itself from your garden as it stands; clone it to try a different version; pick any two and **Compare** them on jobs, hours a week, the busiest week and the first job — never on how much food, which nobody can count honestly; plant one when you’re ready. Saved plans sync like everything else.
Next year is here too. Tap **Next year** under the header and the plan runs from ten weeks before your last spring frost to your first fall frost. Every crop goes in on the first day of its own window — greens a few weeks before the last frost, beans and tomatoes a week after — and anything you start indoors shows its sowing day as well, weeks ahead. Made in September, the first job on the price is next spring’s first sowing, not something due this week.
And Rafe can read a saved plan back to you: “What’s my plan for next year?”, “When’s my first job?”, “What did I plan for Bed 2?” He reads your saved plans, newest first, remade from your garden as it stands, and never invents one — with nothing saved, he says so.
And the plan as a picture: tap **Chart** and the same plan is drawn on the Calendar’s own week grid, place by place across the season, every planned row outlined beside what’s already growing there. Tap a bar to swap it; nothing is written until Plan it.
Still ahead: where in the bed each plant stands, and which variety.
-
The plans have prices now
When setup asks which plan suits you, it now says what each one costs. (Pricing has since changed. See /pricing.) Either one gets you the whole app: every plant, every bed, the garden walk, and Talk with no cap on what you can ask.
Until now that screen showed the two shapes and no numbers, and said so — we had not set a price, and putting one on the screen to make it look finished would have been telling you something we did not know.
**Nothing has changed about what happens today.** If you are in the invited beta, your invite is still what opens the app, nothing is charged, and no card is asked for. There is still no checkout. You can see what you would be choosing before you choose it, which is the whole point of showing you the number.
-
Two headings you could not see, and a setup page that said the wrong thing
Two screens each had a heading drawn under the bar at the top of the screen — there, in the right words, and invisible. Both now sit below the bar like every other screen’s title.
Behind it, the check that looks for exactly this — something a person is meant to read, drawn underneath something else — had been reading the sign-in screen for weeks, because it never got past the door. It opens the door itself now and looks at all twenty-one screens, and it has learned the difference between a control hidden under the bar and one that is simply further down a page you can scroll.
And the help page for setting up your garden said the run ended by asking which plan you would pick. It has not since 16 September: you pick a plan, make your account, and setup is five screens about your garden. The page says so now.
-
“What can I plant right now?” is readable again by day
On the Add a plant screen, the button that lists what still finishes before your frost was showing its crop names in cream on cream by day — you could see the headings and almost nothing else. The block was designed for a dark card it no longer sits on. Every line in it now takes the right ink for the ground it is on, day or night, and the “too late” heading is red rather than orange so the three verdicts read apart at a glance.
-
You see the price before you make an account
Opening GardenTrack without an account now starts on **Pick your plan**. You choose, then you make the account, then setup asks its questions. (Pricing has since changed. See /pricing.)
It used to run the other way round. The plan question was the **sixth** screen of setup, so you told us where you garden, your frost dates, your first crop and where it lives — and had already made an account to get that far — before anyone mentioned what it cost. That is backwards, and it is fixed.
Nothing is charged, and no card is asked for. There is still no checkout: you can see what you would be choosing, which is rather the point of showing you.
-
Add a pile of photos and get a garden
There is a new screen, **Photos to plants**. Add a batch of pictures from your camera roll all at once — or take one there and then — and we read each one, name the crop where we can, and put the photos that look like the same plant together. Drag one card onto another to join them, or tap one and then tap the other if dragging is awkward with one hand.
Nothing is written into your garden until you tap a card that lists every plant it is about to make, by name, with the number of photos on each.
The part worth knowing is what it refuses to do. It never asks the AI whether two pictures show the same plant — we tested that and found it answering from the fence and the mulch in the background rather than the plant itself. So it groups on two things it can actually be sure of: the words on a nursery label, and the clock. Anything less certain than that is left as two cards for you to join. Showing you two cards for one plant costs you one drag. Quietly merging two of your apple trees into one record costs you a tree you never find out you lost.
-
A seed that is not up yet is not forming heads
A broccoli sown on 23 May could read *Forming heads in 3 days* on the day it went in. The app already refused to say a seed was ready to pick before it could have come up; it did not refuse the stages in between — heads, bulbs, swelling roots, flowers. Now, until a seed has had long enough to come up, the line under it says *Sprouting* and nothing later.
Checked across every crop we hold, sown on every day of the year: 96 wrong lines gone, and not one plant flipped to something else wrong in its place. Plants that do not start from seed — a potato, a sweet-potato slip — are untouched, because leaves on day one is right for those. A shallot put in last October also stopped being mistaken for a seed sown this spring.
-
Every change Rafe can make, tried on a real garden — and two that failed, fixed
Fifteen things Rafe can do to your garden — add a zone, a bed, a plant; correct a date, a name, a bed’s sun; move a plant; take an empty bed away; undo an entry; remember and forget something about how you garden; set a reminder; hand you the link when only you can delete something — were each asked for in plain words, three times, on a real garden, and checked against what was actually written down. All fifteen hold.
Two did not, on the first pass. **“Actually, scrap that”** after logging a watering failed twice in three: Rafe found the entry, named it the way the app shows it, and then could not match his own name for it — and blamed “something on my end”. Fixed. And asked to delete a tree, Rafe sometimes asked *finished, or a mistake?* before handing over the link; the page the link opens holds both choices, so he now hands over the link straight away.
-
Rafe never removes anything without your tap, and says where he is guessing
Two things Rafe used to do on his own now come to you as lines on a card with one button.
**Removing an empty bed or an empty zone.** Ask Rafe to clear out an empty growing space and he writes the line — *remove the empty growing space “Spare pot”* — and waits. Nothing is removed until you press the button, and if a plant is still standing in that bed when you press it, the bed stays. On 11 September he removed a bed a gardener had not named, on a sentence that was about a different bed. That is the day this became a button.
**Guessing where something goes.** Say *put the pots on the front porch* when there is no zone called that, and Rafe has to pick somewhere. He used to pick quietly. Now the line on the card says where he would put them, so a wrong guess is something you read and correct — just tell him *the front porch, not the backyard* — rather than something you find later.
Two smaller fixes rode along. When Rafe lists your places he now says which names are zones and which beds sit in no zone at all, so two zones with similar names no longer get read back as one. And a bed whose kind had been stored as “1.0” — a bug, not a kind — reads as a pot again.
-
Plan your season by dragging — or just tell Rafe the day
Your season is now something you plan on, not just look at. Every plant you’re growing draws as a bar across your own frost dates, grouped by zone and bed so a big garden doesn’t turn into a wall of rows. Press and hold a planned one and drag it to a different day — the crop’s own planting window lights up under your thumb and a line tells you straight: “Oct 6 — in the window. Picks from about Nov 20,” or “Dec 1 — too late this year.” Let go and it’s saved.
A Sow again button shows up wherever a crop is one you’d normally plant more than once — tap it and you get the next round, one interval on, with an undo.
And you don’t have to touch the ribbon to use it. Tell Rafe “put lettuce in the porch pot on October 6” and he plans that planting for you — nothing goes in the ground yet, but it shows up on the ribbon outlined, right where you’d have dragged it yourself. Ask later what’s growing and Rafe knows the difference: a planned plant is read back as planned, with a countdown, never as one already in.
And when you want the same question answered for your whole garden at once, tap **What still fits** at the top of the ribbon. One screen, every bed: how many more would fit, the best crop for that space and a couple of backups, each with its own pick date — “6 would fit · picks from about Nov 5.” Tap **Plan it** and they’re in, dated today, the same as if you’d dragged them there yourself, with an undo. A bed with no size on file just says so, with a link straight to setting it. Or just ask: “what could I still plant in the bean pot?” — Rafe reads you the same answer.
-
When Rafe cannot do it, he hands you the link
Ask Rafe to delete a plant and he will tell you he cannot — that one is yours to do, because deleting takes the whole record with it. Until now he described where to go. Now he gives you a **Take me there** link that opens that plant or growing space and drops you straight at the delete control.
He still names both choices, because they are not the same: *It is done* keeps everything you recorded and simply takes the plant off Today, and *Delete this plant* is for something you added by mistake. If it really grew, you want the first one.
The link only shows up when we are certain which plant you meant. Two things by the same name and Rafe says so, rather than sending you to the wrong one.
-
When Rafe wants to change things, you get a button that lists them
Sorting out a muddled garden takes a few messages. At the end of one, Rafe knows exactly what needs to change — but every change he makes has to point at your own words asking for it, and after a long back-and-forth there often is no single sentence that does.
**Now he shows you a card instead.** It lists every change in plain words — *rename the garden zone “backyard orchard” to “front yard”*, *move “Bean Pot” into backyard* — and one button makes them. Pressing it is the permission; there is nothing to mistake.
**Or just keep talking.** If a line is wrong, say so instead of pressing the button and Rafe will adjust.
⚠ **Every line is a real change to a real bed or plant.** If Rafe cannot say precisely what he would do — because two of your beds share a name, say — that one stays off the card rather than hiding inside an “and 2 others”.
-
Every menu row has its own picture
The menu carries a small picture on every row now, drawn in the same line as the five buttons along the bottom — so after a few days you stop reading the words and go for the shape. No two rows share one: a shape you learn and then get wrong is worse than no shape at all.
-
Fill a bed with a salsa bed Partly live
A bed with room in it now offers ready-made beds — a salsa bed, a salad bed, and one for children — worked out to the space you actually have. It is a meal rather than a plant list: most people would rather start from "I want salsa" than from a search box and 153 crops. It tells you how much of the bed each part takes, gives a plant count only where the number is useful, and says what is left over. If part of it has missed its season it says which part, and whether that matters — a salad bed whose radishes have missed is still a salad bed, so it drops them and gives the room to the rest; a salsa bed without tomatoes waits for spring. Nothing is saved until you tap a row, which takes you to the ordinary add screen with that crop chosen.
Not yet: Three ready-made beds so far, and it shows the plan rather than saving it — each row is a door into the add screen. There is no save-the-whole-bed button yet.
-
A new growing space now opens with what to put in it
Make a bed and its page used to open on a photo slot and a settings form, with the ready-made bed suggestions in small text near the bottom. An empty bed is the moment you most want ideas, so that is now the first thing on it: a big **Add plants** button, and the ready-made beds that go in **today where you are**, each with a picture of what you get out of it.
There is a whole page of them now too, and it is searchable. Every ready-made bed is there all year — a salsa bed, a pizza bed, three sisters, a bed that keeps through winter, one for the bees. **The search covers the crops as well as the names**, so typing "corn" finds the three sisters plot even though the word is nowhere in its name.
Open one from a growing space and it tells you how it stands in *that* bed — whether it fits, how much room is left over, or what it is waiting for. Open it on its own and it tells you what the bed is and what goes in it, and no measurements, because there is nothing to measure against yet.
Once something is planted, the page goes back to leading with your plants. By then the question has changed to what to do with the rest of the bed, and that answer is still there where it was.
-
Ten ready-made beds, and a new bed that asks how big it is Partly live
A brand-new growing space showed no planting suggestions at all, which is worst on exactly the bed you most want ideas for. It now asks instead of going quiet: four common bed sizes as buttons, and one tap fills in both sections. The library went from three ready-made beds to ten — salsa, salad, pizza, stir-fry, herbs, a bed that keeps through winter, one for the bees, one for children, a first bed for anyone who has never grown anything, and a three sisters plot. The ones that can go in today are the buttons you see; the rest sit one tap behind a line saying how many are waiting, and each will tell you what it is waiting for. A bed only sees the ones that suit it.
Not yet: Beds and row plots only — how many plants fit a pot is about root volume rather than floor space, and we will not convert one into the other by inventing a depth. It still shows the plan rather than saving it; each row is a door into the add screen.
-
Tapping Try again during the pause no longer asks twice
There is a second or two between sending a question and Rafe’s first word. Tapping **Try again** in that pause used to ask the whole question a second time — two turns, two answers, and twice the usage off your allowance.
The turn now counts as open from the moment you send it, so *Try again* stays out of reach until the answer has finished.
One thing deliberately unchanged: **an answer that comes back empty still gets its Try again.** That is exactly when you want it.
-
Your beds now suggest what goes well with what is already in them
When something is growing in a bed, its page now suggests crops that go well beside it — and **every suggestion says why, and how well established the claim is**. Some of these are measured in trials, some are a known effect, and plenty are long-standing practice with no study behind them at all. We tell you which, because "basil goes with tomatoes" and "basil cuts thrips numbers on tomatoes" are different sentences and only one of them was measured.
It only offers things you could actually plant this year, so a companion that has missed its window is left out rather than dangled in front of you. **And it will not suggest a plant that clashes with something already standing there** — a companion for your tomatoes that suppresses the beans next to them makes the bed worse. If two things already in a bed do not get on, it says so either way.
One thing it deliberately does not do: tell you what other gardeners plant together. That is a good question and we have built the plumbing for it, but we will not answer it until there are enough real gardens to count. A number worked out from a handful of them would not be a fact about gardeners.
-
Rafe can act on what you told him earlier in the conversation
Sorting out a muddled garden takes a few messages — you say what you have, Rafe reads it back, you correct a bit, he asks which of two you meant. **At the end of that, nothing was getting written down.**
Every change Rafe makes has to point at the words you used to ask for it. He was only looking at your *last* message, so after a long back-and-forth he could not find them — and instead of saying so, he tidied your words into a summary, which counts as putting words in your mouth and was refused.
He now quotes you properly, from **whichever message you actually said it in**. A correction you work out over several messages gets recorded.
⚠ **One case still needs you to say it in one go:** if a single change depends on facts spread across three separate messages — the variety in one, where it is in another, when you planted it in a third — Rafe quotes the closest one and may still ask you to put it in a sentence.
-
Rafe can rename a part of your garden, and clear out an empty bed
Until now, when Rafe got something wrong about how your garden is arranged, the only thing he could do was **add another one beside it**. Ask him to fix the name and you ended up with two.
He can now rename a part of your garden — "call the backyard orchard just backyard" — and he renames only that. Every bed inside it, and everything growing in them, stays exactly where it is.
He can also get rid of a bed or pot **that has nothing in it**, for the ones you made by accident. If anything is still growing there, he will not — that bed holds the whole record of those plants, and losing it is not a correction. Move the plants first and he will clear the bed after.
Deleting a **plant** is still yours to do, on the plant’s own page, where the app shows you what it is about to remove and asks you to confirm.
-
Rafe can undo an entry you logged by mistake
Logged a watering twice, or on the wrong plant? Ask Rafe to undo it and he will — where before he told you he could not, and then described a screen.
**He has to look the entry up first.** Rafe can only undo an entry he has actually been shown in that conversation, so he reads your journal, finds the one you mean, and removes that. He cannot act on one he half-remembers from a chat last week.
He takes the entry and nothing else. The plant it was about, and every other entry on it, stay exactly as they were.
-
Every plant picture is new, and lit by the same sky as the rest of the app
All 153 plant pictures have been redrawn. The old ones were built out of hard angular facets, which looked sharp at full size and turned to mush at the size a plant actually shows in a list. The new ones are smooth and solid, and you can tell a garlic from an onion at a glance.
The bigger change is underneath. Each picture now arrives with **no background at all** — just the plant. The app puts the ground behind it, using the same colors that shift across your screens as the day goes: a warm glow from the horizon at dusk, cool and quiet at night, bright at midday. So your tomatoes sit in the same light as everything else you are looking at, and they change with it through the day.
If you have not set up a garden yet, or the weather has not loaded, the pictures simply sit on a plain quiet background. We would rather show you nothing than guess what the sky is doing where you are.
-
A bed with room in it now tells you what to put there
Open a growing space with space left and it names one crop — the one with the most days to spare before your own frost — with up to three more that would also make it and a count of what cannot finish this year. One answer, not a list: a gap in a bed has one right answer and a deadline. The line under the crop is why it is the answer, and for something like garlic that is not racing the frost it says what is true instead — it goes in now and is lifted next summer. It tells you how many would fit only when the bed is what limits you. In October it will say there is room and nothing to put in it, which is the honest answer for the middle of autumn.