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.
-
Running out of questions used to be silence. We found it by using the app on a phone
**Ask your sixth question of the month and nothing happened.** No answer, no explanation, no Try again — your question sitting on its own above an empty space. Ask another and the same nothing. **Anybody who did not know a limit existed would keep typing into a box that had stopped working.** **The infuriating part is that the app knew.** It had the sentence written and ready — *that is all the questions on the free plan this month* — and it wrote it down correctly the moment the limit was hit. **Then a background sync wiped it a second later.** The sync's job is to agree with our server about which messages exist, and it treated "the server has your question" as proof the question had been answered. Your question WAS delivered. That is why it came back. **What failed was the reply**, and a sync that reconciles delivery has no way of knowing that. **We fixed it as the rule rather than the case:** the sync owns what it knows, and it does not know about the answer. Anything else that can fail after a message is delivered would have had the same bug. **Second thing, from the same session.** Asked "when is my first frost again?", the chat said it had no note of your frost dates — while the home screen one tap away was showing them. **Then it explained the gap by inventing a setup step you never did**, and pointed at a settings screen that does not exist. Until you give a ZIP those dates are the typical ones for your zone, they live on your phone, and the chat was simply never told. Now it is sent the same line the home screen shows, plus where the dates came from. It gives you the date, says it is the typical one rather than yours, and tells you a ZIP gets you your own. **Both were found by walking the app on a phone-sized screen, signed in, like a stranger** — which nobody had ever done. Every number in every test before that came from a machine talking to a machine.
-
Setup moved above your jobs, and stopped being a checklist once it is nearly done
**The setup list sits above the day\u2019s jobs now, because it feeds them.** A garden with no ZIP code gets the typical dates for its zone, so every deadline underneath is an average wearing a date\u2019s clothes. Fixing what the app knows about you outranks working through what it worked out. **It only earns that spot because it shrinks.** With one thing left it is a single line — the step, the reason, the count — instead of four rows where three are ticked and carry a paragraph each explaining why they were worth doing. **Nobody needs the reason for something they have already done.** **And the two cases are genuinely different screens.** A brand-new garden gets the full list, because then the list IS the thing to do. A nearly-finished one gets a line. **The date and frost countdown were also redrawn** — same height, but the days-to-frost figure is set apart, because that is the half that decides what is still worth planting. The date is context and now looks like context.
-
The add-a-plant list used to open on African marigold. Now it opens on what you can plant
Tap **Add a plant**, do not type anything, and you got all 151 crops in alphabetical order. On a phone that is a 260-pixel window over eight thousand pixels of list, so the answer to **"what can I actually put in today"** was about thirty scrolls away — and the first thing offered was African marigold. **The search was never the problem. The default was.** **Now the list opens with what is plantable this week in your garden**, worked out from your own last frost date rather than a national average — the same calculation the crop page uses, so the two can never disagree. Below that, a divider, and then everything else we hold. **Nothing is hidden.** A browse list exists to show you the catalog is bigger than you thought; filtering it would teach the opposite. All 151 are still there, in one scroll, just in an order that answers the question you opened the screen with. **And the ones about to close come first.** If a sowing window has a week left, that crop sits at the top with **"about 5 days left"** beside it, because a window nearly shut is the most urgent thing on the screen rather than the least. Where it is the final week, it says **"last chance this year"** instead of counting down to zero. **How much it changes depends on the day**, which is the point: in late April a zone 6 garden sees 84 crops ready, in mid-July 18, in late October 5.
-
The add-a-plant screen was giving two different answers to one question
On the add-a-plant screen there are two ways to ask **"what can I plant now?"** — the button at the top, and the crops highlighted at the top of the list beneath it. **Yesterday, on the same day, in the same garden, they said 34 and 11.** **Both were honest. That is what made it bad.** The button asked the real question — *will this crop actually finish before your first frost* — and the list underneath asked a shallower one, *is the sowing window technically open*. Two reasonable questions, one screen, and no way for you to know which you were reading. **They now run the same calculation, so they cannot disagree.** Not "we checked and fixed the gap" — there is one piece of code answering the question and both parts of the screen call it. The list can no longer drift away from the button, because there is nothing left to drift. **Crops that will "only just" make it count as plantable in both.** If a source says a crop will finish but without much room, that is a yes with a caveat — and the list must not be stricter than the answer sitting above it. A test now checks the two numbers match on four dates spread across the year, because agreeing today is not the same as being unable to disagree tomorrow.
-
Hardening off now names the tray you are about to pick up
**A card that said "Put 6 plants out" made you work out which six.** If those six are on one named tray, it now says **Put the windowsill tray out** — the object your hands are actually going to lift. Same for bringing them in. **Two different trays are never called one tray,** even on the same day of the plan. Sending you out for one and leaving the other on the sill overnight is the exact failure this whole feature exists to prevent, so a mixed group goes back to counting plants. **And here is what we did not do.** The ask was to run hardening off by tray, so we went and read 34 extension and seed-company sources first. The guidance is not written that way — it describes the plant, by about 17 to 1, and the tray language belongs to seed catalogs rather than the horticulture. So the ten-day plan is still tracked per plant. **What is genuinely a tray thing is the carrying,** and that is the part we changed.
-
Two universities disagree about what eats dill, and now we say so
**Ask us about hornworms on your dill and until today you got a description written about tomatoes**, with nothing telling you that is where it came from. **Here is the honest tangle underneath.** University of Minnesota's hornworm page names the hosts plainly: "Tomato is the host of choice, but they have also been found on potato, eggplant and pepper." No dill. **But NC State's dill page says the opposite** — "Dill is susceptible to aphids, tomato horn worms, and powdery mildew." **Both are real extension services and neither is wrong.** One is a page about the insect, listing the crops it was studied on. The other is a page about the plant, listing what troubles it. They are answering different questions, and we had quietly stacked one on top of the other. **We nearly deleted the dill link to tidy it up. That would have thrown away a sourced fact** because a different page did not repeat it — which is exactly the mistake we spend most of our time trying not to make. **So nothing was removed. Something was added:** when the app answers about hornworms on a plant Minnesota did not study, it now tells you that page is about tomato. **You get the disclosure instead of the deletion**, and you can decide what it is worth on your dill. **One thing genuinely worth knowing while you look:** the big green caterpillar on dill, parsley, fennel or carrot is usually a black swallowtail — we hold a separate page for it. Plenty of gardeners leave that one alone on purpose.
-
The home screen went from four and a bit screens to one
**Measured on a real garden of eleven plants: nineteen separate things asking for attention before you scrolled, and the largest type on the page said 23 when one job was actually due today.** The other twenty-two were this week and no rush. **A pile of bills, not a morning.** Our own research file has the most damning review in this category, about a rival: a planning app that *"just made me constantly feel behind."* **Now the number is today’s, and the job is named.** *One thing today — the skin has set, it is ready to come in.* One button. The rest of the week is a single line underneath. **Nothing was deleted.** The season charts, the will-not-ripen warning, the tree jobs and the pollination note moved to your week screen, whose whole rule is that doing nothing there is a successful visit. Your plants moved to the Garden tab, which already listed every one of them grouped by bed. The setup checklist stayed, because it is what tells you what the app still needs from you. **Four and a bit screens became one.** Sixty-three things to look at became sixteen.
-
The card is a header now, and Done means done
**Your jobs were inside the orange card, so the card grew with the list and every job sat on a block of color.** The orange is a header now — the count and one way to be walked through everything — and the jobs sit below it on the page, each with its own button. **And "Done" now finishes the job where you are standing.** If there is something left to fill in, it opens on the same screen: log that you watered and the how-much question appears right there. **The word "Open" is gone** — it was an extra tap that started nothing, and the few jobs that genuinely move you somewhere (a photograph of a leaf, permission to send you an alarm) now say what they will do instead. **One of those buttons was appearing where it should not have.** Our own rule said a job covering more than one plant could not be finished in a tap — but that is the rule for the "how much" question, not for the job. The app had always logged all of them happily and simply not asked. **The guard was stricter than the thing it was guarding**, and the cost was a pointless tap on a job that needed none.
-
The season chart moved off the home screen, and nothing was deleted
**Your home screen had nineteen separate things asking for attention before you scrolled.** That is measured on a real garden of eleven plants, not guessed — and the season chart was two of the biggest. **It is a thing you look at, and it was sitting above the one thing you opened the app to do.** So it moved to your week — tap the job count on Today and it is underneath, unchanged. **That screen was already built for looking.** Its own rule is that doing nothing there is a successful visit: no progress bar, no percentage, nothing that treats an unticked job as a debt. The week and the season are the same question at two zoom levels, so they belong together. **Nothing was removed and nothing was hidden** — one tap, and it is the same chart with the same two bands. This is the first of two changes to the home screen; the second is the diet.
-
The tomato virus you spread yourself — and the bit about smoking that everyone gets wrong
**Yesterday we added the tomato virus an insect brings you.** Today, the one you bring yourself. Penn State: "The most important way that TMV can be spread from plant to plant is on workers' hands, clothing, or on tools." **The mechanism is the useful part, because it tells you exactly when it happens.** "When plants are handled, the tiny leaf hairs and some of the outer cells inevitably are damaged slightly and leak sap onto tools, hands, and clothing. If the sap contains TMV, it can be introduced into other plants when those come in contact with this sap." **Touching a sick plant and then a healthy one is the event.** Not a season of bad luck — a minute. **And it waits.** The virus "can also survive outside the plant in sap that has dried on tools and other surfaces", and in dead tissue it stays infectious "for years after the infected plant part dies". **Now the part almost everyone has backwards.** Tobacco products "particularly those containing air-cured tobacco, may carry TMV", and hands pick it up from handling them or keeping them in a pocket. **But Penn State says plainly: "TMV is not spread in the smoke of burning tobacco."** The packet is the risk. The smoke is not. **One limit we have written into the record rather than glossed:** that page describes the virus across plants generally rather than on tomato specifically, and it warns its own symptom list can also be caused by heat, insects, herbicides and mineral problems. Mottled leaves alone do not settle it. That takes us to **95 problem descriptions**.
-
The way in had three names, and none of them was the one you were told
**The Talk screen tells you: "sign in on Profile to ask."** You go to Profile, which offers **"back it up"**, and tapping that lands you on a screen headed **"back up your garden."** **Three names for one door — and the word you were sent looking for was on none of them.** Nothing was broken. You simply had to work out for yourself that all three were the same thing, at the exact moment you had already been stopped once. **Profile now says "sign in to back it up" and the screen says "sign in to back up your garden."** The benefit is still the better half of the sentence, so both keep it. **And the screen now says what you actually came for.** It explained that signing in means your garden survives a lost phone — true, and not why you were there. It also says that signing in is what lets you ask questions, which is the thing that sent you.
-
Today's jobs are on the home screen now, each with its own button
**You used to get a count and a Start button.** Now you get the jobs themselves, listed, and a button on each one — so a job you have already done takes one tap and no screens. **Not every job gets a button, and that is on purpose.** Anything that needs a photograph, an amount or permission to send you an alarm opens properly instead. **A one-tap button that skipped the second step would record half an answer and look like the whole one.** Everything you do tap can be undone from the same line. **The list leads with what cannot wait.** A tray that has to come in tonight sits at the top with a mark — hardening off is the one routine where being late costs you the plant, and it used to be a line at the very bottom of the screen. **And the two kinds of task now say what they are.** Garden tasks are what you do outside. App tasks are what you tell us — which variety, what kind of bed, your ZIP — and they sit above, because with no ZIP every date underneath is your zone\u2019s average rather than your garden\u2019s. **One more fix:** a failed backup was announced in the quietest text on the page, tucked behind a button. It is the only thing on that line that will not fix itself, so it is the only one that now looks like a problem.
-
Two names, one fungus — and a squash that rots on the shelf was already sick
**Most of what we have added this month has been one name hiding several different organisms** — blight, mildew, damping off, cabbage worms. This one runs the other way. University of Minnesota: "Gummy stem blight describes the leaf and stem infection stages of the disease. Black rot describes the fruit rot stage. Both are caused by the fungus _Didymella bryoniae_." **Two names, one fungus.** So a rot in the fruit and a blight on the vine are not two things that went wrong — the one you can see tells you about the one you cannot. **The stem sign is the one to learn.** "Tan lesions develop on infected stems, often close to the crown of the plant. These lesions crack open and exude gummy, amber-colored fluid." Amber goo weeping from a split near the base of the plant. **And on the leaves it depends which crop you are looking at**, which is the kind of thing a general answer gets wrong. On cucumber and summer squash the browning starts at the leaf edge and becomes "a brown wedge extending into the leaf tissue". On melons it is "brown discoloration between veins or into circular tan to brown spots" instead. Do not go hunting for a wedge on a melon. **The part worth knowing if you store winter squash:** Minnesota says "fruit infections may develop in the field or in storage depending on the crop". A squash that blackens on the shelf was carrying this, not catching something in your house. That takes us to **91 problem descriptions**, covering a bit over three-quarters of the times a plant in our database names a problem.
-
Some blanks on our plant pages are deliberate. Now they say so
Ask us how much to feed a crop and you will find no number. **That is a decision, not a gap**, and until today the page looked identical either way — an empty row tells you nothing about whether anyone looked. **Here is the reason, which is now printed on the page.** Extension services do publish fertilizer figures, and they are pounds of nitrogen **per acre** for a production field. Converting one to your raised bed needs your soil test and your fertilizer's analysis, and guessing errs in the direction that burns roots. **It is also why every feeding chart you find online disagrees with the next one.** We would rather tell you that than invent a number. **Same for watering.** We give a depth per week rather than gallons per plant, because that is how irrigation is actually measured — a per-plant figure depends on your spacing, your soil and whether it is in a pot, so it would look more precise than it is. **And where a source does clear the bar, you get the number.** Blackberries have a real per-plant watering figure and grapefruit have one too, so those pages show it and skip the explanation. The refusal is a default, not a rule — thirteen crops beat it. **Behind this:** a field can now be one of exactly four things — it has a value, it does not apply, we looked and nobody published it, or we decline to say and here is why. **There is no fifth state and no silent blank**, which is what let 285 of these decisions sit in internal notes where no reader could ever see them.
-
Two onion rots that look identical from outside. Cut it top to bottom.
**You store onions or garlic, they look fine going in, and six weeks later some of them are rotten.** There are two common causes and from the outside they present the same way — a bulb that was sound at lifting and is soft now. **"It went bad in storage" does not tell you which one you have. Which END of the bulb does.** **Fusarium basal rot comes up from the bottom.** Utah State describes "red-brown discoloration and rot" forming "along the root-basal plate margin" — the flat plate the roots grow out of — with the tissue inside turning brown. It starts where the roots are and works upward. **Botrytis neck rot comes down from the top.** It starts "in the neck area during storage", the neck goes sunken, the scales turn "gray to dark brown", and pulling the scales apart shows "white to gray mycelium" growing between them. **So the test is one cut, lengthways.** Rot at the root plate is one disease, rot at the neck is the other, and you can tell in a second what you could not tell from the outside at all. **Both of them hide until they are in the bag**, which is the frustrating part and worth knowing: Utah State says basal rot symptoms "may not be visible in the field, as they tend to develop during storage", and puts neck rot at one to two months in. That takes us to **93 problem descriptions**, covering a bit over three-quarters of the times a plant in our database names a problem.
-
You do not need to know the name of the thing eating your kale
**"Something is eating my kale" used to save as a plain note.** So did "bugs on my beans", "holes in the leaves" and "something ate my lettuce". Four sentences that are unmistakably a pest sighting, and the journal counted none of them. **The reason is embarrassing and worth saying.** The app was looking for the NAME of the pest — aphid, slug, caterpillar, beetle — **which is the one thing you do not have when you are asking what is wrong.** Somebody who already knew it was aphids got a proper entry. A beginner describing exactly what they could see got a note. It did not even recognize "bugs", because the word it was looking for was "bug" and the plural fell off the end. **Now the damage counts, named or not.** An unnamed something doing the eating, a generic word like bugs or critters, or just the holes themselves — all of them record a pest sighting on that plant, with your words kept exactly as you wrote them. **And it will not invent one.** "We ate the first tomato" is not a pest report, and neither is digging holes for the tomatoes — the difference is whether something unnamed is doing the eating, which is what the app now reads. A wrong entry in a journal built to be trusted is worse than the miss it replaces, so both halves are in the tests. **And the same day, the sentence at the other end of the season.** "Got our first zucchini" recorded nothing, and "got the first tomato" recorded the plant SETTING fruit rather than you picking it. Nobody says pick, harvest or dig about their first one — they say got. It counts now, along with "got a few beans off the plants". Not for the two things you also get your first of that are not a harvest: a frost, and a flower. **Where that leaves the wider number:** of 56 real sentences a gardener might say, the app now understands 80%, up from 57% a week ago and 43% before that. Anything it still does not follow is saved in your own words. Nothing is ever lost; it just does not count toward a total yet.
-
Your name for a bed now answers the question about it
**We were asking "what kind of space is this?" under a bed somebody had named "raised bed."** The name says it. Now the app reads it — **raised bed**, **deck pots**, **the orchard**, **greenhouse** all stop the question. **It refuses more often than it answers, and that is the design.** Call a bed "Back bed" and it still asks, because a bed is raised about as often as it is in the ground and the word settles nothing. Name two things at once — "greenhouse pots" — and it asks rather than picking one. **A plant name is not a place**, so "Orange tree" asks too. **The reason it is careful is that the kind decides watering.** A pot holds a fraction of what a bed does and dries from every side, which is why pots are the commonest thing people lose. Reading "bed" as "pot" would water somebody’s garden wrong on the strength of one word. **And nothing read off a name is stored as if you had chosen it.** What you pick always wins, and rename the bed and the reading changes with it. **Fixed alongside: a chart that quietly dropped a plant.** Two of the same crop planted the same day shared one identity, so one of them vanished from your season chart with nothing on screen saying so. Two tomatoes in on the same afternoon is an ordinary thing to do.
-
"Nothing is wrong" is now something the answer can actually say
**Roughly half of every honest answer about a plant is that nothing is wrong** — and until now there was nowhere in the answer to put that. The reply had a field for the likely cause, one for what it could also be, one for what to do. **No field for whether anything was actually wrong.** **So everything downstream had to guess.** A healthy answer would come back saying the speckling on an apple was normal lenticel dotting, and the card would headline the cause and leave you to work out that this meant you were fine. Our own checking tool guessed by looking for disease words in the text — and marked a healthy tree as a confident mistake because the word "aphids" appeared in a sentence saying there were none. **Now the answer says it outright.** *Nothing is wrong with it*, and underneath, what the marks actually are. Verdict first, then the reason — because being told you are fine without being told why is not reassuring, you just go and look again tomorrow. **And there is a third answer: it can say it cannot tell.** That is different from not knowing which problem it is. It means take a closer photo, rather than assume either way. **The lesson we are keeping:** we had patched that guessing twice in two days. The second patch is the point where you should stop patching and ask why anyone is guessing at all.
-
"Picked the first tomato" was recorded as the plant fruiting, not as a harvest
**The first tomato is the moment of the season, and we were writing it down as the wrong thing.** "Picked the first tomato" got filed as the plant setting fruit rather than as you picking one. **"Picked the first strawberry" worked perfectly**, which is the part that gives the cause away. We had a short list of fruit words — fruit, tomato, pod, bean — that meant "this plant is fruiting", and that list outweighed the word "picked" sitting right beside it. Strawberry was fine because it was not on the list. **The cases that worked were the ones nobody had written down.** **Now the verb decides.** Picked, pulled, harvested — any of those makes it a harvest, whatever the fruit is called, including fruits we have never listed. With no picking word, "first tomato" still means the plant is setting fruit, which is its own thing worth keeping. **We found it by driving the app rather than waiting for it to be reported.** Nobody had hit it yet. It was sitting on one of the first sentences a new gardener types.
-
Three citrus numbers now say what they depend on — and twelve deliberately say nothing
Some of the numbers we publish come from a source writing about a whole family rather than the one plant. **"Most citrus trees"** is not a measurement of your mandarin. We went looking for every number like that, found seventeen across the five citrus, and then read them — because the interesting question is not which ones are general, it is **which ones could mislead you.** **Twelve could not, so they were left alone.** The chill-hours figure is the clearest case: citrus needs essentially no winter cold, and that is true of a lemon and a satsuma equally. Stamping *"this is a general citrus figure"* on it would be technically accurate and completely useless — and every pointless note makes the next one easier to skip. **Three earned a note, and none of them for being general.** They earned it for the same real reason: **a citrus tree's size is set by the roots it was grafted onto, not by the fruit on top.** A semi-dwarf mandarin wants about eight feet; a standard one wants fifteen. Same fruit, nearly double the space. So the spacing rows now tell you to check what yours is grafted onto, and the pot size says fifteen gallons is where a young or dwarf tree starts, not where a mature one ends up. **The rule we settled on: a note has to change what you would do.** If it only tells you our sourcing was broad, it is our problem to fix rather than your problem to read.
-
We taught the app the harmless curl first. Today it learned the serious one.
Yesterday GardenTrack learned that curling tomato leaves usually mean nothing is wrong. **That left it half-taught, and the wrong half.** It could reassure you and it could not warn you — so the one case where you needed it to speak up was the one it had nothing written down for. Now it holds tomato yellow leaf curl virus too, from NC State Extension, and the two are meant to be read against each other. **Do not go by which way the leaves roll.** That is the test everywhere else on the internet, and NC State says this virus curls leaves **both** ways. Two things separate them instead. **Which leaves:** the harmless roll starts on the older leaves low down and the plant otherwise looks well, while the virus takes the youngest leaves at the top, yellowing them between the veins, and the plant stunts and goes bushy with it. **And the pattern across your row:** a feeding problem is, in their words, "likely to be consistent across the entire crop", where infected plants "will be patchy across the field and not uniform" — because it arrives one whitefly at a time. That takes us to **87 problem descriptions**, covering a bit over three-quarters of the times a plant in our database names a problem. **Two things we deliberately did not write**, because that page does not say them: that the virus cannot spread on your hands or tools, and that there is no cure. Both were in our first draft, both sounded obviously true, and neither was on the page we were citing.
-
Two of the ladybugs in your garden are eating it
**Everyone learns that ladybugs are the good guys. That is true right up until it is not.** Clemson Extension: "This is one of two species in the lady beetle family that feed on vegetable crops. The other is the Mexican bean beetle." So there are exactly two ladybugs you should look twice at — the **squash beetle** on your squash and pumpkins, and the **Mexican bean beetle** on your beans. Every other one is working for you. **The Mexican bean beetle is the one people wave through.** It is domed and coppery-brown with eight black spots on each wing cover in three rows, and Clemson describes its damage in a way that identifies it on its own: it feeds on the **undersides** and leaves the upper surface "more or less intact", so the leaf ends up skeletonized and lacy rather than holed. **Its neighbor on the same bean plants is the opposite.** The bean leaf beetle chews "large irregular holes" straight through, and — the part worth knowing — also girdles stems down at the soil line, so a young plant that keels over for no visible reason is worth checking at ground level. Its spots sit in a single row along the inner edge of each wing cover, not in three rows across. **And do not go by color on that one.** Clemson says bean leaf beetles "vary considerably in color and markings", which means every color-led identification guide for it is arguing with the source. Where the spots sit is the test. That takes us to **90 problem descriptions**, covering a bit over three-quarters of the times a plant in our database names a problem.
-
We put quote marks around a paraphrase. Now a machine checks every one.
On our tomato leaf curling page we wrote that herbicide damage comes with **"white, split stems and deformed fruit"** — in quotation marks, credited to Clemson. **Clemson did not write that.** What they wrote is: "In addition, stems may turn white and split, and the fruit may become deformed." **The word we lost, twice, is "may".** Ours read like a checklist for the diagnosis; theirs says these things sometimes happen. A gardener could have ruled out herbicide drift because their stems were not white and split — which the source never asked for. The page now carries the sentence as written, and says plainly that the absence of those signs does not clear it. **Two people had already checked that page for exactly this, on the day it was written.** The difference is that a person compares a quote against their memory of the source, and a machine compares it against the source. So we built the machine: every one of the 229 pages we cite is now fetched and kept, and **every quoted sentence we publish is checked, word for word, against the page it is credited to.** A build fails if one does not match. It is running on 80 quotes today, and it will run on every one we write from now on.
-
We said the shape settled it. It does not — correcting our cucumber page.
Our page on yellowing cucumber leaves told you that angular patches with straight edges, stopping dead at the veins, meant downy mildew — and that **nothing else makes that shape.** **That last part was wrong, and it was ours rather than anything a source told us.** Angular leaf spot is a bacterial disease named for doing exactly that: Minnesota describes its spots as "typically confined within leaf veins". Same tiling, different organism. **Where the mistake came from is worth saying,** because it is an easy one. Minnesota's cucumber diagnostic lists nine causes and only one of them has that shape — so "nothing else does this" felt like reading the list. It was not. A list of nine causes is a list of nine causes, not a statement about everything that exists. **The page is better for it, and the fix is a thing you can actually do:** turn the leaf over. Minnesota describes "dark purplish-grey fuzz" on the underside for downy mildew, and for angular leaf spot "sticky drops of whitish liquid" instead, with its spots later drying and dropping out to leave "irregularly shaped holes". The shape gets you to two; the underside picks between them. On a cucumber specifically, downy mildew is still the better first guess — Minnesota notes angular leaf spot is "less common on cucumber due to the availability of resistant varieties".
-
What actually grows in the shady bed — and what it costs you to put it there
Ask most plant databases about sun and they will tell you almost everything wants **full sun**. Ours did too: **150 of our 151 crops** said full sun and exactly one said partial shade. So if you told us a bed was shady, the honest answer we could give you was chervil. **The information was never missing.** It was sitting in our own sourced notes for **85 crops** — currants, honeyberry, mache, chamomile, spinach, kale — written in a sentence rather than in a field anything could read. It now has a field. **And it carries what shade costs you, which is the part that matters.** *"Tolerates partial shade"* is true of a blackberry. The same source sentence continues *"produce best in full sun."* A yes-or-no answer would have put blackberry at the top of your shady-bed list, which is a confident recommendation to grow fewer blackberries. So there are three answers, not two: **shade is better here** (chervil, arugula — light shade slows bolting in heat), **also takes part shade** with nothing lost, and **takes part shade but crops less**, which names what you give up. **One thing we deliberately did not build.** The plan called for recording how many hours of sun each crop needs, and 114 of our files contain the phrase *"6 or more hours of direct sunlight a day"* — which looks exactly like that number. It is not. It is one university's **definition of the words "full sun"**, printed identically beside every plant they list. Filling a column from it would have put a made-up figure on 114 crops. **Real per-crop hour figures exist for about eight**, so the column waits until there is something true to put in it.
-
You sowed it four days ago. The app now tells you when to expect something
The week after you sow seed has exactly one question in it: **is anything going to happen?** We have had the answer for **116 crops** since the database was built — how many days that crop usually takes to come up — and no screen had ever shown it. Now the plant's own page says it, and only while it matters: **"Sown 4 days ago. Expect the first sprouts in about 2–8 more days."** **The version that earns its place is the last one.** Once you are past the window, it says so — *"this usually comes up within 10 days, so it is worth a look"* — because that is the point where you have a decision to make about re-sowing, and nobody wants to make it on a hunch three weeks in. **It does not tell you to re-sow.** We cannot see your bed; it could be cold soil, old seed, or a bird. It tells you the window has passed and leaves the looking to you. **And it knows when to say nothing.** A plant you bought as a seedling never germinated in your garden, so it gets no sprout line at all. Thirty-five crops have no trustworthy figure and stay quiet rather than guess. And the whole thing retires itself two weeks after the window closes — a plant that came up last month does not need a card about germination. **How we found it:** not by noticing. A check we built the day before, which walks every fact in the database and asks whether anything on screen actually reads it, flagged this one on its first run.
-
A lime gets three kinds of mite. We had it down as "mites"
The University of Florida names them individually on limes: **citrus red mite, citrus rust mite, and broad mite** — and says which one russets the peel, which is the whole reason you would want to know. NC State, writing more briefly, just says "mites." **We recorded "mites," and the note explains why: somebody took the part both sources agreed on.** That sounds careful. It is the same move as averaging a precise measurement with a rough one and calling the result agreed. **A vaguer statement is not competing evidence against a specific one — it is the same claim at lower resolution**, and the only thing the intersection does is throw away the resolution we already had. **Lime now carries citrus rust mite by name, with the generic entry kept beside it** — because Florida names two more species we hold no code for, and dropping to a single named pest would imply those two are not on the page. **The part worth telling is what it looks like across the family.** Grapefruit was already specific. Lemon and mandarin say spider mites. Lime and orange said "mites" — and orange is the only one of the five whose source is genuinely generic. **Five crops, three answers to the same question, and four of them decided one crop at a time by whoever was reading that day.** That is the real finding, and it is the kind you only see by looking at a family rather than a row.
-
A photo of damaged leaves landed as a plain note and nothing looked at it
**Take a photo on a plant's page and it now asks whether you want it looked at.** One line, a yes and a no. **Before this it just filed.** Somebody photographed damaged apple leaves and the picture went into the plant’s history as an untyped note — three rows below the pest that had been logged from the same conversation, and the two never met. The app had the sourced descriptions and the ability to look, and did neither, because nothing connected a new photo to either one. **It asks rather than just doing it, and that is deliberate.** Most photos people take of their plants are records, not questions — the tomato getting big, the first flower, what the bed looked like in July. An answer nobody asked for is the app talking when it was not spoken to. **The one exception is a photo we asked you for.** If it said "photograph the underside of that leaf" and you went and did it, the asking already happened, and asking again would be the app forgetting its own question.
-
A sunflower that flowered is not a problem
When a lettuce shoots up and flowers, that is bolting: the leaves turn bitter and the crop is over. The app treated that as bad news, correctly. **It treated it as bad news on every plant in your garden**, which meant telling somebody their sunflower had gone wrong. **Of the 28 event types the app records, this is the only one whose meaning depends on what is planted.** A harvest is good news on every plant. A disease is bad news on every plant. Flowering is the whole point of a sunflower, a cosmos, lavender or chamomile — and on a cover crop it is the signal to cut it down — while on a lettuce it is the end. **And it runs both ways.** If your spinach *flowers*, that is bolting under a different word, and the app now says so rather than congratulating you on a milestone. **Nothing new had to be recorded to fix it.** The app already knew which plants are grown for their flowers — it just was not asking before deciding. 28 of our 151 plants, and the two that were hardest to spot were coneflower and feverfew: filed as herbs grown for their leaves, and nobody on earth grows a coneflower for the leaves.
-
A true description, a real source, and the wrong plants
**Our description of root rot came off a University of California page about peppers.** It was quoted faithfully and credited properly, and we had it attached to 46 different crops. Same with leaf spot, off a page about trees and shrubs, on 31. **Nothing was misquoted, which is exactly why it was hard to see.** An independent read scored the writing 15 out of 16 and still found this, because the problem was not in the sentence or in the source — it was in which plants we hung them on. A source makes a claim look checked, and that one was checked, for peppers. **Now the answer tells you.** Where a source covers something narrower than your plant, it says so — "this is described for peppers" — instead of quietly applying it to your beans. **One was worse than a scope problem and was simply wrong.** Our spotted cucumber beetle entry said the grubs feed on the roots and wilt the plant. That is the *striped* cucumber beetle. The spotted one lays into corn, beans and grasses instead — and every crop we had it on was a squash, cucumber or melon, so it was wrong on all of them, and it sent people digging for grubs that were not there. **And our stink bug description was one species pretending to be all of them.** The banded antennae we called the giveaway are what tells the brown marmorated apart from the native ones — so if you were holding a native stink bug with no banding, we told you it was not a stink bug.