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.
-
Four more crops stopped asking you to harvest something you just planted
A follow-on from the same fix. We hold, for most crops, how many days from planting to something worth picking — and that number is measured from one of two starting lines: from the seed going in, or from setting a young plant out. They are weeks apart, so we only ever use the one that matches how you actually planted. Cereal rye publishes only the seed figure, which is right, because nobody transplants rye. But if a rye planting was recorded as a transplant, we had no matching number and fell back to saying nothing — so it went back to offering a harvest on a plant put in yesterday. Now a seed figure is used whatever the record says, and a transplant figure never is. That sounds arbitrary and is not: the two mistakes are not the same size. Using a seed count for a transplanted plant over-estimates, so at worst we stay quiet a bit too long. Using a transplant count for a sown seed under-estimates — sage measured from setting out is 75 days, and charging that to a seed would call it ready weeks early. One direction fails into silence and the other into a wrong instruction, so only one of them is allowed.
-
A mint planted yesterday is no longer told to harvest it
Some plants come back every year — mint, sage, chives, lemon balm, a raspberry cane. For those, what we hold is the shape of a normal year: leaves from spring, cutting through summer, dying back in autumn. That describes an established plant and is genuinely useful for one. It is nonsense for something you put in yesterday, and the app was applying it anyway — so a mint planted in June landed in the middle of cutting season and was told to go and cut it. Fifteen crops did this. It now asks how long the plant has actually been in the ground: we hold, for most crops, how many days from planting to something worth picking, and until a new plant has had those days the harvest job waits. From the second year on, the normal-year description really is about your plant, so it applies. Two things we deliberately did not do. We did not record that mint takes a year to become cuttable, because it does not — you can cut mint the summer you plant it, just not the day after. The obvious fix was to write a false year against nine herbs to silence a wrong message, which would have made the database lie to fix a screen. And if you have told us the fruit is sizing, you win: our day count is a catalog figure for a good season and you are looking at the actual plant. That took two goes to get right — the first version hid the very question that lets you tell us. Three crops can still do it, because mint, rosemary and lemongrass have no day count in our records yet. Those are being sourced.
-
Photo answers are now told when we do not know a planting date
When you photograph a plant and ask what is wrong with it, the answer is built from what we hold about that plant. If we know when it went in, that goes along with the picture, and it is what lets the answer check the stage it can see against your own calendar. If we do not know, nothing at all was being said about it — no date, and no mention that there was no date. A gap where a fact should be is the easiest thing in the world to fill in, and filling it in is exactly what we do not want. It is now handed the words date not recorded, which is a fact rather than a hole. Knowing when something went in is still not knowing how old it is, and it will not tell you that either way: a tree bought at three years old went in the ground last month, and nothing we hold can tell those two apart. One more thing came back with this. How a plant was started, sown where it stands or raised somewhere else and moved, was only passed along when a planting date came with it. That is one of the most useful facts there is when something looks wrong at the base of a young plant, because transplant shock and damping off look alike and start differently. It now goes with the photo whether or not we know the date.
-
A paragraph on the plant database page was nearly the same color as the card behind it
At the bottom of the plant database page there is a cream card explaining what those pages are for and what the app does. Its text was a pale lavender on that cream — close enough in tone that it read as washed out rather than as writing. The color itself is one we use deliberately, but only on the dark parts of the site, where it is meant to sit. One line of styling had applied it to a light card by mistake. It is dark brown now, like every other paragraph on a cream card. We found it with a tool that photographs a page with the words hidden and measures the real color underneath each one — the ordinary checker cannot grade text sitting on artwork, and that page is mostly artwork.
-
A three-year-old horseradish was told every spring that it was still settling in
Settling in happens once. But the “establishment” stage for horseradish was pinned to the last frost rather than to the day you planted it, which made it a calendar window — so it reopened every spring, and a crown you put in three years ago was reported as still getting established. It is now measured from your own planting date, so it happens once and then stops. The “when do I plant this” answer on the crop page is unchanged. Shallots have the same problem and we have deliberately not fixed them yet: the same change fixes one way of planting a shallot and leaves the other broken, and it shifts the planting date we show by two weeks. A half fix that moves a date is worse than the bug it treats, so that one stays on the list until it can be done properly.
-
One tomato disease in our database had no source behind it, and two near-misses hid that
Our plant database lists eleven diseases for tomatoes, and every entry is supposed to trace back to a university extension page that names it. We went through all eleven against the three pages we cite. Ten checked out. The eleventh — tomato yellow leaf curl virus — was named by none of them. Two things made it look fine. One page mentions leaf curling, but as a sign of aphids, not a virus. Another names curly top virus, which sounds like the same thing and is not: different disease, carried by a different insect. Anything matching on shared words would have called both of those a source. The disease is real, and the description we show for it was already based on a dedicated North Carolina State page. That page is now cited on the tomato record itself, so the entry stands on its own. We also rewrote tomato's pest and disease notes to say which source named which item, instead of listing the sources as a group. Every other crop already worked that way; tomato was the last one that did not.
-
A cauliflower twenty days old was labeled “harvest window”
For 19 crops, the stages after the seedling are counted from the day a bought plant goes in the ground. Sow the seed straight into the soil instead and there is no such day, so those stages had nothing of yours to count from and fell back to a general calendar for the crop — which had cauliflower heading in June, because a bought one planted in April would be. We already refused to put a job on your list from a date like that. Now we stop putting the label on the plant, too: it was the worst of both ways round, nothing to do and a line saying it was ready. The honest cost is more quiet — across those crops, days with something to say went from about three quarters to about half, while wrong “ready to pick” readings on a plant under a month old went from twenty to none. Nothing changes for a plant you bought and planted, or for anything in its second year or later.
-
Alaska could get frost dates from our website but not from our app
The detailed map our frost dates come from covers the lower 48. Outside it, our free web tool has always fallen back to your nearest official weather station — and said so. The app never did. It treated “outside the map” as the end of the road, so a gardener in Anchorage could not finish setting up at all, and every date they saw was a fixed guess meant for zone 6b. Now the app falls back the same way. Anchorage gets a last frost around 1 May and a first around 29 September, from a station a mile and a half away. Fairbanks gets 16 May and 6 September — a month later in spring than the guess it was being handed. Hawaii and Puerto Rico get something better than a date: we tell them frost does not happen there, which is true and is not something the lower 48 ever gets to hear. We say which station answered and how far away it is, because a station miles away over different ground is a weaker answer than the map and you should be able to see that. None of this is new data — the stations were already in the file we publish.
-
Telling the app something never gets you silence and a button
When you tell GardenTrack something instead of asking it something, two things happen at once: the assistant replies to you, and it files what you said so the right buttons appear underneath. Occasionally it did the filing and skipped the reply, and you would get a card offering to add your plant with no words above it at all. That reads exactly like being ignored by something that is meant to be listening. The app now checks before the answer reaches you and, if there are no words, writes a short line itself from what it just filed — written down, watered the tomatoes, or got all three, check them over below. It is a floor rather than a replacement: a one-line receipt is worse than a real reply and the assistant is still expected to write one. This only makes sure a silence never reaches you. The strange part is that half the app already did this. There are two ways an answer can arrive — one where the words appear as they are written, and one where they arrive all at once — and only the second had the check. So whether you saw any words at all depended on which path your question happened to take, which is not something anybody could have noticed or reported. Both paths now ask the same question, so they cannot drift apart again.
-
An apple tree was being told to start itself indoors
Every plant page has a section called "What we will not put a number on". It is for the handful of things we can state and choose not to, and each one comes with the reason. One of them is a direct-sowing date for peppers and sweet marjoram: extension services publish one, but it assumes row covers, and printed next to "frost kills this plant" it reads as an instruction to sow outdoors before your last frost. That is how you lose a crop, so we hold the date back and tell you to start those two indoors instead. The note was written for two plants and was appearing on forty-nine — including apple, cherry, fig, garlic and strawberry. Nobody sows an apple tree, and the note ended by promising "the indoor start date is on this page" when thirty-six of those plants have no indoor start date at all. The cause was underneath: when someone records that a question does not apply to a plant, that decision was being thrown away before it reached the page, so an answered question and an unasked one arrived looking identical. Both are fixed. The warning is back on the two plants it was written for, and it is still there.
-
Apple problems now say what they actually look like
Ask about a sick apple and the app could name the likely trouble but not describe it, which is the half that decides anything. Four of apple’s common diseases now carry a sourced description of what you would see: apple scab, fire blight, cedar-apple rust and bitter rot. Each one is written from a named university extension page and says which one, with a link. The useful part is what separates them. Apple scab makes roughly circular spots with feathery, ill-defined edges, olive-green to black, mostly on the upper side of the leaf. Cedar-apple rust makes a hard-edged yellow-orange circle instead, and pushes fringed tube-like structures out of the underside. Fire blight turns blossoms and twigs black as though scorched, and the dead leaves stay on the tree rather than dropping, which is the thing nothing else on an apple does. Bitter rot sinks into the fruit in rings, and cut through the spot the rot runs inward as a brown watery V pointing at the core. We also say where a description stops. The cedar-apple rust page we learned from describes the leaves and not the fruit, so the entry says so rather than letting you assume it covers both.
-
The cold-night warning was printing "undefined" for two crops
There is a kind of cold that ruins part of your harvest without endangering the plant: nights in the forties and low fifties, well above freezing. Nothing dies, but a tomato stops setting fruit, a bok choy decides winter has been and gone and runs to seed, and a basil comes through looking rough. Those want completely different things from you — the first is pure reassurance, the second is a deadline to go and pick — so the card names which one it is. It only knew two of the three. For eggplant and basil it printed a fragment reading "and undefined will stop setting fruit", which was a broken sentence and also wrong advice, since neither of them stops setting fruit from the cold. Two of the six crops we hold this data for. Both now get their own sentence, and it deliberately stays general: our records say "quality suffers" without saying how, and the sources differ — eggplant gets tough skins, basil gets blackened leaves. Naming one would have been right for one crop and invented for the other. Worth saying why this only covers six crops: it needs a source naming a temperature and a specific harm in the same breath, which is a rare thing to publish. Where we do not have it we say nothing — so a quiet night is not proof your plant is unaffected.
-
A tree page said “15–15 feet”, and asked two trees a question with one answer
On a seed-grown orange, the plant page read “without this we can only say 15–15 feet — the rootstock is what decides”. Our source gives oranges one spacing figure, and the sentence printed it as a range with the same number at each end. It reads “15 feet” now. Checking why led to a bigger one. Eleven kinds of tree ask which rootstock yours is on, and we told all eleven that the answer decides their spacing. It decides it on four — apple, pear, mandarin and peach, where a dwarf wants about 6 feet and a standard about 25. On the rest we hold one range whatever the tree is grown on, so we were asking you to go and read a nursery label to sharpen a number that was never going to get sharper. Those pages now just give you the number. Two of them — apricot and persimmon — were offering a single answer, which is not a question, so that card is gone entirely. The card stays on citrus, because there the rootstock really does change how many years until you get fruit.
-
How long to start a plant indoors, counted from the day you plan to put it out
Until now, everything we published about starting seeds indoors was counted backwards from your last frost date. That is the right frame for a lot of crops and it is how a lot of extension services write it. It is also a frame that needs you to know your frost date before the number means anything. Onion, cucumber, okra, summer squash and pumpkin now carry the other version: how many weeks the plant sits indoors before you set it outside. Onion is eight weeks, okra four to five, cucumber and summer squash three, pumpkin four. It is more useful than it sounds, because the date you actually have in your head is usually "the weekend I am going to plant this", not a frost statistic. The odd part is where the numbers came from: they were already in our files. Extension services publish this timing exactly this way — "sow seeds indoors about 8 weeks prior to planting" — and every time somebody building the database hit one, they wrote it down in a note and threw the number away, because we had nowhere to put it. Six crops have a sentence like that sitting in their record. Two of them say so outright. The value was never missing; the slot was.
-
It calls your plant what you call it
Every plant in our records is filed under one name, and it is often not the one you would use. We file zucchini under Summer Squash, collard greens under Collard, green beans under Bush Bean. So if you typed Zucchini, the app went on asking you about a summer squash. That landed worst on the sentence where it mattered most: the question asking whether something has come up exists to sound like it knows your garden, and it was the one sentence reading from our records rather than from you. Every card about a single plant now uses your word — the question, the running-late note, the reminder to sow it, and the one that asks whether it ever went in. A variety counts as your word too: a plant recorded as Beet · Chioggia is asked about as Chioggia. If you never gave it a name of its own it still uses ours, because a question with no plant in it is worse than one with the wrong name. One card still says the crop on purpose: the pest note is a single row covering every bed growing the same thing, so there is no one plant of yours it could honestly name.
-
66 varieties were printing the same word twice
A variety row shows its type as a small chip next to the name, then anything else known about it as separate pills. Where a source happened to list the type among the traits as well, both printed — so Bergamot Orange came out as “Bergamot Orange · sour · sour · strongly scented”. Sixty-six varieties did it, across bok choy, celery, elderberry, flax, lemon balm, orange and more. A row now leaves out anything its own heading has already said. Nothing was deleted from the data — the trait is still recorded the way the source wrote it. One thing we avoided: a few crops store the type field as the whole list of kinds that crop comes in, so a blackberry carries “thorny” and “thornless” at the same time regardless of which it is. Matching against that would have hidden “thornless” from three blackberries that really are.
-
Some garden problems have no germ behind them, and now we can say so
Ask what is wrong with a plant and almost every reference you will find answers with a pest or a disease. But a lot of what actually goes wrong in a garden has nothing living behind it at all: sun on an uncovered fruit, a cold snap in May, water that came and went, a flower that never got pollinated. We have added sunscald as a described problem — pale brown, leathery, sunken patches on the side of the fruit that faces the sun, which is the part that tells you it is this and not a rot, because a rot starts where the flower was. That takes us to 115 problems we can describe rather than only name. Behind the scenes we also started a second set of records for this whole family of trouble, so the app can eventually answer three questions it simply could not before: it flowered and nothing came, it ran to flower far too early, and the most useful one of the three — nothing is wrong with your plant. Rolled tomato leaves after a wet spring are one of the most-searched garden panics there is, and the honest answer is that the plant is fine and the fruit will be fine. An app that can only name troubles will always get that one wrong, because naming a trouble is the only thing it has to offer.
-
“Tap to try it again”, when trying again could not work
There is a ceiling on how much AI a single account can use in a month. It exists so a bug on our side, or someone else getting into your account, cannot run up a bill in your name — and it sits well above what a real month of gardening reaches. If you hit it, we want to hear from you rather than have you shrug. Two screens did not know about it. The plant page and the photo answers both showed “That photo did not get read — tap to try it again”, which is advice for a fault when what had happened was a limit. Tapping again could not work. The chat had the right words all along; now all three do. This is separate from running out of your free allowance, which reads differently and is expected: “That is the last of your free AI answers. Everything else keeps working.” Your plants, dates, journal and reminders are free either way.
-
On a second phone, an answer with a whole list under it arrived as one line
Two buttons on the Talk screen give you an answer the app builds itself rather than one the AI writes. “What should I do this week?” puts the jobs underneath it; “When will things be ready?” puts every plant’s dates and how they line up against your first frost. Because the app builds those parts, the AI above them only has to write the one sentence a list cannot say for itself — which pick is worth watching, or what surprised us. That is deliberate, and it is why those answers are quick and cannot invent anything about your garden. But only the sentence was being saved to your account. So the phone you asked on had the whole answer and any other device had that one line with nothing under it: the footnote without the answer. Nothing was ever lost on the phone you asked on, and a sign-in somewhere else never overwrote it. The drawn part now travels with the answer, so both devices show the same thing. Answers from before today keep whatever the phone you asked on still holds — we cannot send up a list that was never stored.
-
It kept offering the one door that was locked
If you garden in Alaska, Hawaii or Puerto Rico, the app used to ask you for your location over and over and then refuse the answer. The prompt named one way to do it — drop a pin on the map — and the map only covered the lower 48. So the one route it suggested was the one route that could not work for you, and it kept suggesting it. That is fixed twice over. The map gap itself is closed: Anchorage now gets a last frost of 1 May and a first of 29 September, Fairbanks 16 May and 6 September, both from a weather station a mile or so away. A garden in Honolulu or San Juan is told plainly that it does not get frost, which is a better answer than a date would be. Those numbers were sitting in our own data the whole time — the public frost-dates tool on this site has been serving them for months, and only the app treated the same reply as a dead end. The second half is the wording: the prompt now names both routes, a street address as well as the map, because when one of them cannot work for where you live the app should not be pointing only at that one. Worth saying what this changes for a gardener in Fairbanks: the app had been defaulting to a last frost of 17 April. The real one is 16 May. That is a month, on the date that every other date in the app is counted from.
-
Ask about a plant and the assistant now sees what your notes were about
Every journal entry you write gets tagged with what it is about — pests, feeding, watering, soil and a dozen more. Those subjects shipped yesterday and you could already search and filter by them on your own notes screen. What could not see them was the assistant, which is one of the three reasons a beta tester asked for them in the first place: so that asking a question about your own notes gets a better answer. Now, when you ask about one particular plant, the subjects on that plant’s entries go to the assistant alongside the entries themselves. It matters most for plain notes. Every other kind of entry announces its own subject — a pest sighting is obviously about pests — but a note just says "note", so without the subject a note about aphids and a note about feeding look identical. One thing he deliberately still will not do: search. Asking for everything you have ever written about aphids, across every plant and every year, is your notes screen’s job and not the assistant’s. We would rather say that plainly than have him look for something he never went and found.
-
The camera shortcut on Add a plant had never once opened a camera
Beside the orange “Add a plant” button there is a camera, for when you cannot name the thing in front of you. Tapping it was supposed to open your camera straight away. It never did — not slowly, not sometimes. It showed the form with the keyboard up, which is the opposite of what you tapped for. No browser will open a camera unless a person taps something; that rule stops a web page switching your camera on behind your back, and it is a good rule. We were asking anyway and being refused in silence. So the app has stopped trying to be clever: you now arrive with the camera button already selected and ringed, keyboard down, so the screen visibly says it heard you. One tap opens the camera, and it works in every browser rather than none. What made this hard to catch is that it looked fixed from the code and looked fixed in review — two of us had written and shipped it. Measuring the screen itself was what settled it: with the camera call counted on a real arrival, the count was zero.
-
The big green caterpillar on your dill is probably not a hornworm
A university page lists tomato hornworm as something dill is susceptible to, so we carry the entry. But the caterpillar that actually strips a dill plant is normally a black swallowtail larva — a parsleyworm — and we carry that too. The problem is that the two are confusable by exactly the feature the first one is named for: disturb a parsleyworm and it rears up and flicks out a forked orange horn from behind its head. A fat green caterpillar with a horn is a hornworm, obviously — except on a carrot-family plant, where it usually is not. The app now says, on dill, that the hornworm description was written about a different plant family and may not describe what you are looking at. The entry stays; what is added is the warning that would otherwise be invisible. Worth knowing before you act on it — a parsleyworm turns into a black swallowtail butterfly.
-
On some cards the date was invisible, and nothing we had could see it
Every card on the blog, guides and tools pages puts a small date or label over the artwork, with a dark fade behind it so the words stay readable. The fade was not tall enough. The headline sat inside the dark part and read fine; the little amber label sat just above it, in the part where the fade had almost run out. On cards whose artwork is dark you would never notice. On the ones with a bright sky or a sunlit hillside the label all but vanished. The reason it went unnoticed for so long is that our normal contrast checker refuses to grade text sitting on a picture, and it is right to: a photograph has no single color to measure against, and picking one gives a confident wrong answer. That refusal covers around 1,700 pieces of text on this site. So we built something that photographs the page with the words hidden, reads the actual color underneath each one, and reports the worst pixel. It put the worst of these labels at roughly one-to-one against their background — which is to say the same color. We then went and looked, and it was right. The fade now extends past the label on every card.
-
The line was too long, and every number about it was wrong
On a laptop the app's writing ran too far across the screen before wrapping — long enough that your eye loses its place coming back to the start of the next line. That part was real and it is fixed: the column of text now stops at about seventy characters, which is where reading research puts the comfortable range. Nothing changed on a phone, where the lines were already well inside it. The interesting part is that every measurement we had of the problem was wrong, including our own. The report said a hundred and two characters at a font size we do not use — this app has one text size and it does not change with the window. Our first attempt to check it said a hundred and fourteen. Both came from the same shortcut: assuming a character is half as wide as the type is tall. Measured properly against the actual letterforms, the true answer was eighty-three. That mattered, because the suggested fix had been to make the text bigger, and at eighty-three the honest lever is the width of the column instead. Making the type bigger would have changed every screen in the app to solve a problem that only existed on one. We also found that the standard unit web pages use for this counts the width of a zero rather than an average letter, and in our typeface a zero is wider than average — so the conventional setting everybody reaches for would have landed at seventy-six characters while claiming sixty-eight.
-
The lines between your jobs were invisible
One job ended and the next began, and nothing on screen said where. The divider between two rows was drawn in white at nine percent — which works on the dark theme and disappears completely on the light one, where the page is cream. Measured, the line and the page behind it came out at 1.018 to 1: not faint, but almost exactly the same color. Turning the line up would not have helped, because white over cream never separates no matter how strong you make it. It now uses the color that already flips with your theme, so it reads on both. The same mistake was in two other places and both are fixed. One rule that looked identical was left alone on purpose: the list inside the orange panel on your home screen sits on a dark card whatever theme you use, so white is the right color there and the theme-aware one would have drawn a pale tan stripe across it. Worth saying plainly: nothing we run would have caught this. Our contrast checks read text against its background, and a divider is not text. It was found by a person looking at his own garden.
-
The number in the corner was a door
Working through your jobs one at a time, the little "1/7" at the top opens the whole list — every job with its plant and its bed, so you can jump straight to the one you actually came out for instead of stepping through the rest. Nothing said so. It had no border, no background and no underline, it was the same muted color as the label beside it, and a fraction is exactly the shape that reads as a readout rather than a button. It was also 23 pixels wide, against our own 44-pixel minimum for anything you tap. There were really two buttons there doing the same job, the progress bar and the number, and only the bar had a name — so a screen reader announced the same door twice, once as "see all of today's jobs" and once as "1 slash 7". They are one control now: the whole strip is pressable, there is a chevron saying it opens something, and it announces itself as "Step 1 of 7. See all 7 jobs for today."
-
The number was right and the calendar belonged to somebody else
Four more crops have stopped telling a plant you put in yesterday to go and harvest it: lavender, lemongrass, sweet potato and tarragon. All four count their picking season from the day the plant goes out rather than from a seed — and for these four we hold no figure for when that step happens, so the app was falling back on when it usually happens in your area. Sweet potato shows it plainly. Our harvest figure, 90 to 120 days after planting out, is correct and sourced. Hung on a borrowed date it could land on any day of the year, including tomorrow — so a slip put in the ground yesterday could be told to dig it up. The app now asks first whether your plant has actually been in the ground that long. One deliberate side effect: direct-sown broccoli waits until day 50 for its harvest prompt instead of day 43, because 50 days is broccoli’s own figure and it is measured from planting out, which a direct-sown broccoli never had.
-
The same question, twice, in two wordings
Two separate parts of the app had each grown a way to ask whether a seed had come up. They were written months apart, they sat 250 lines apart, and neither one knew the other existed — so on the crops where both fired you got "Any sign of the broccoli yet?" and "Any broccoli sprouts?" on the same screen, about the same bed, on the same morning. We built every crop at every age and counted: it happened on 147 of them. There is one question now, from one place, in one set of words. Deleting the newer card was the obvious fix and it was the wrong one — the two did not overlap everywhere, and on broccoli, whose own page says first sprouts in five to ten days, deleting the older one would have left nothing asking between day five and day twenty-four. The card that survived took over that window first. Two smaller things went with it. The "no" button now reads "Nothing yet" everywhere, because four other cards say "Not yet" to mean "ask me later" and record nothing, while this one records that you looked and the soil was bare. And the reasons behind the question no longer open by telling a five-day-old seedbed that it is running late — inside the window it says so, and only past the window does it say the plant should be up by now.
-
The sun question existed and nobody was told about it
Since 21 August each growing spot has had a question on its own screen: how much sun does it get? Answer it and the app can tell you a plant is somewhere it will sulk — a porch is not a lawn, and until you say so the app cannot tell them apart. The trouble was that you had to go and open that spot to find the question. If your beds were already in the app, nothing ever mentioned it, so for most gardens the whole feature sat there unused and silent. It now appears in the App tasks list on Today, counting how many of your spots have an answer, and tapping it takes you to your garden. Nothing was filled in on your behalf while you had not answered: a guess read off a bed’s name is only ever shown to you to confirm, never saved as your answer, and the new line does not tick itself because of one. It is a one-off — once every spot has an answer the line goes for good.