Solve-ent

Every release

Newest first. Written for the people who use this, not for programmers: if a release changed something you can see or do, it says so in the words you would use; if it changed something under the surface, it says that plainly.

Every release also says what is still missing, because a list of only fixes is an advertisement.

• 1.0.0

  • This is version one. Nothing changed in it. The number moved because everything that was meant to be here is here, has been used on a real device, and has been fixed where using it showed a problem.
  • What that means in practice: seven topics, each one a move you cannot do a chemistry calculation without. Every question worked one step at a time, with what you have already written kept in front of you. A wrong answer that gets told back to you as which mistake produces that exact number, or says plainly that it cannot tell — never a guess. A calculator and a way to report a problem in the bar, on every screen. Difficulty chosen by you and never changed on your behalf. It works with no internet after the first visit, keeps nothing about you, and asks for nothing.
  • What it still will not do, on purpose. It will not solve the question. It will not tell you how you did overall, give you a score, count a streak, or say anything about you as a person — only about the move in front of you. And it will not guess between two mistakes that would produce answers you could not tell apart; it says so instead, and that is counted rather than hidden.

  Still missing: the calculator keeps only the line you are on. Reports have to be copied and pasted, because nothing here reaches the internet. And the check that makes sure things are on the screen on a small phone works from a written-down list of moments — a new control has to be added to it by hand.

• 0.16.0

  • The bar stays with you now. The calculator and the report used to scroll off the top of the screen the moment anything moved — and getting a step wrong scrolls the page on purpose, to put the reason in front of you. So the two things you most want after a mistake were gone at exactly the moment you wanted them. They stay at the top of the screen now, wherever you are on the page.
  • And the reason no longer hides behind that bar. Making the bar stay put covered the top of the explanation with it, on a small screen. It sits clear of it now.
  • Letter spacing stops widening the things that are read one character at a time. If you have the wider spacing switched on, it still applies to sentences, where it helps — but not to a rearrangement you are choosing between, a number you are typing, a value you have already written down, or the problem report. Those are read character by character, and spreading them out makes them harder, which is the setting working backwards.
  • Under the surface: there is a new check that opens the app at the size of a small phone with the keyboard up, and fails if anything you need is off the screen or behind the bar. Both of the last two faults reported here were things that existed, worked, and could not be seen on the device this is used on. Every other check ran on a taller screen where they were all perfectly visible.

  Still missing: the check knows about a fixed list of moments. A new control has to be added to that list, or it is not measured — the list is written down where it can be seen rather than worked out automatically.

• 0.15.0

  • The calculator is in the bar at the top now, on every screen. It was there before, but it sat underneath the answer box below the button that checks the step — which on a tablet, with the keyboard up, is off the bottom of the screen. Somebody working through a question read the whole screen and reported that there was no calculator, and they were right: a control you cannot see is not a control you have.
  • Tap the keypad icon next to the ⓘ and it opens over whatever you are doing. It is in the same place during practice, during an assigned set and while you are drilling one move on its own, so it is not something to find again each time.
  • It still puts the number straight into the answer box, and it closes itself when it does, because the next thing you want is the box it just filled. On a step that asks you to choose between rearrangements there is nowhere to put a number, so that button is not offered and the panel says why — the keys still work if you want to check something before choosing.
  • It still never rounds and it still never sees the question. It multiplies the numbers you chose in the order you chose them. Working out how many figures an answer is entitled to is one of the seven things this is for, so that part is left to you.
  • Reporting a problem is in the bar too, next to it. It was five headings down inside the ⓘ panel before, which is where you look when you are curious about the app — not when it has just done something wrong and you want to say so. Tap the flag, pick whichever line is closest to what happened, and copy the report.
  • There is nowhere in that panel to type anything, and that is the point. It means there is nothing in the report about you. It carries the version, the size of the screen, the kind of device and whether anything is stored — and it does not carry your settings. If you have text size, spacing, one-step-at-a-time or read-aloud switched on, none of that goes in, so reporting a fault never tells anybody how you have the app set up. The report says so in its own last line rather than asking you to take it on trust.
  • Under the surface: the check that keeps score-and-praise language out of this app had only ever been reading the code, never the screen. It reads both now, which is most of the words you actually see.

  Still missing: the calculator does not keep a history of what you worked out — only the line you are on. The working log above the question is what keeps your finished steps. And the report cannot be sent from inside the app; it has to be copied and pasted, because nothing here reaches the internet.

• 0.14.3

  • The version you are running is on screen now, quietly, next to the name at the top. It was only inside the ⓘ panel before, which is no use for the thing a version is actually for: when you send a screenshot of something going wrong, the number has to be IN the screenshot. Without it there is no way to tell a problem that still exists from one already fixed, or from an old copy the device had kept.
  • It is small, it is selectable so it can be copied rather than typed out, and it is the real running number rather than one written down twice.

  Still missing: nothing new.

• 0.14.2

  • Three more of the same, two of them found by reading how MoleBridge — the sibling app, same room, same students — had already solved them.
  • The keyboard was coming back up over the reason. Get a step wrong and the app put the cursor back in the answer box, which on a tablet raises the keyboard and scrolls the box into view — pushing the explanation of what went wrong off the bottom of the screen. So you got the "wrong" and not the "why", which is the whole point of this app. Nothing takes the cursor back now: it goes to the explanation instead, the keyboard drops, and retyping costs the one tap you were about to make anyway.
  • There was no way out of a set. Once you had started, the only ways out were answering every question or reloading the page — no help at all if you picked the wrong topic or just wanted to go and read something. There is a Stop this set button now. In practice that is one tap, because practice records nothing. In a set somebody gave you it asks twice, and the second time it says what leaving costs: Stop — you will not get a code.
  • The scratch line only existed on one screen. Practising one move on its own had no way to do the arithmetic, which is absurd when one of the moves you can practise IS doing the arithmetic. It follows you onto that screen now, and stays away on steps that ask you to choose rather than to work something out.
  • Also: when a mistake keeps happening, the note during a run and the summary at the end used to say what the mistake was CALLED — "what fixes it is undoing an operation on both sides". They now say the same thing the panel says, which is what to actually do about it.

  Still missing: nothing new.

• 0.14.1

  • Four things found by using it on a phone, which is where it will actually be used.
  • A question is several steps, and the screen never said so. You saw a box with a chemistry question in it and then, underneath, a question about rearranging letters — with nothing to say the second was a piece of the first. On a phone the box scrolls away and they are simply two questions. Each step now says where it sits: Step 1 of 2 in this question — choosing the move.
  • When a step went wrong it read like a shout. A panel appeared with "What happened at this step" at the top of it, which is a verdict in the loudest place on the screen at the worst moment. It says Have another look now.
  • And it did not say what to do. It named the mistake and then named the topic the mistake belonged to — "the move underneath this step: undoing an operation on both sides" — which tells you which chapter you are in and leaves you holding the same wrong answer. It now says what to actually do: to get a letter on its own, undo what is being done to it, and do the same to both sides — a letter that is multiplying the one you want is a letter you divide by. And it ends by saying you get to try again, which it never did except when it could not work out what happened at all.
  • The scratch line was offered where there was nothing to work out. On a step asking which of three rearrangements is right, a keypad and a "put this in the answer" button say the step wants a number, so you go looking for numbers that are not there. It only appears on steps that ask for one now.

  Still missing: nothing new. The icon and the deploy landed in the two releases before this.

• 0.14.0

  • A set you were given is still there if you close it. Shut the tab, let the device sleep, come back later — the app offers to pick it up at the step it stopped on, with everything already done still counted.
  • Only a set somebody gave you is remembered. Practice is not, because practice records nothing, and there is nothing in a practice run to lose.
  • It says whose set it is. The offer names the number, which matters most on a shared device: if it is not your number, you can see that before carrying on and finishing somebody else's work under it. Hiding the number would have been the same disclosure with a wrong record on the end of it.
  • It does not sit there forever. An unfinished set is forgotten after twelve hours — long enough for a lesson, a break, a bus and a battery, short enough that yesterday's set does not surface for whoever picks the device up tomorrow. It is also forgotten the moment a set finishes and its code exists, and one button forgets it on the spot. Forgotten means gone rather than blank.
  • A set that was left across an update is not picked up. The questions come from the key, the topic and the difficulty, so a release that changes how a question is built changes what sits at that place in the set — and carrying on would mean counting answers to questions that no longer exist.
  • One thing that was quietly broken and is fixed: pressing through the opening screen went straight to the topics without checking for an unfinished set. The one route a person who had lost their place is most likely to take was the one route that never offered it back. Everything goes through the same check now.

  Still missing: nothing is deployed yet, so this all runs from a device that already has it.

• 0.13.0

  • There is a real icon now. A conical flask, with the liquid in it drawn as an equals sign.
  • The flask is the chemistry — it is the one piece of glassware everybody recognises at the size an icon is actually seen. The equals is the maths, and it is the right sign for this in particular: what this teaches is rearranging around an equals sign, using a ratio, moving powers of ten. A plus would have named the one operation the seven topics barely touch, and a flask with a cross on it looks like a first-aid app.
  • It is a solid shape with the sign cut out of it rather than an outline with a sign inside, because at 32 pixels a solid shape reads and thin lines inside one do not. There is a second version for the phones and tablets that crop an icon to a circle, with the flask kept well inside what survives the crop.
  • The colours are read from the same file everything else here is coloured from. An icon is the one part of an app that the readability checks cannot reach — it is a picture, not a page — so it was the easiest place for a colour nobody had ever measured to sit unnoticed. It is generated from the palette now, and a check refuses to let the two drift apart.

  Still missing: a set closed part-way through is still lost, since the code only exists at the end. Nothing is deployed yet.

• 0.12.0

  • Codes are read a stack at a time. Paste the whole lot into the box — one per line — and press once. Each gets its own block, in the order they were pasted, headed by the number so it can be found against a list on paper.
  • A line that does not read stays where it is rather than being dropped, and says what was wrong with it and what was on the line. Underneath there are two counts: how many read and how many did not. Two numbers rather than one over the other, because one over the other reads like a mark out of something.
  • A number that turns up twice is pointed out and not decided about. It happens when a code is written down twice and it happens when two people were given the same number, and this cannot tell those apart, so it says which line the number appeared on before and leaves it there.
  • And the moves are added up across the whole stack, ordered by how much went wrong. That is the part worth taking back to the next lesson: it says what to teach again, not who to talk to.

  Still missing: the icon is still a simple drawing rather than a proper one, and nothing is deployed yet, so this runs from a device that already has it.

• 0.11.0

  • A set can be handed in now. If somebody gave you a key and a number, there is a way in from the front screen, and at the end you get a code — sixteen characters in four groups — to give back.
  • The code says which number you are, how many steps you did, how many were right first time, which kinds of move went wrong and how many times, and roughly how long it took. That is everything in it. No name, nothing you typed, and nothing about the settings you were using: text size, spacing, one step at a time and read-aloud stay on the device, and there is nowhere in a code for them to go. Nobody should have to tell a room something about themselves just by handing work back.
  • And there is a page for reading them. Whoever set the work opens /teacher, types in the set as they gave it out, and pastes the code. It reads on the device, with no connection and nothing to sign in to, because the page doing the reading is the same code that did the writing.
  • The characters are chosen for handwriting: there is no I, L, O or U in a code, and if you write O for zero or l for one it still reads. A code from a different key, a different topic or an easier difficulty than the one that was set does not read at all, and the page says which way it is wrong rather than just refusing.
  • One thing said plainly, because it would be easy to imply otherwise: the check on a code catches a character copied wrong, or last week's code, or the wrong set. It cannot stop somebody who reads the source of the page and writes their own — there is no server here, so there is no secret to keep one out. The page says so itself.
  • A drop-down list on the new page turned out to be painted by the browser rather than by this app: nineteen pixels tall, in system greys that had never been checked for readability in either light or dark. Both are fixed, and the sweep that checks every screen now covers twenty of them rather than fifteen.

  Still missing: the icon is still a simple drawing rather than a proper one. Codes are read one at a time, so a stack of thirty is thirty pastes. Nothing is deployed yet, so this all runs from a device that already has it.

• 0.10.0

  • You can do the arithmetic without leaving the question. Every step that asks for a number now has a scratch line under the answer box, with a keypad, and a button that puts what it worked out straight into the answer.
  • The keypad is there because this gets used on a tablet by finger, and × and ÷ are two keyboard layers away on most of them — far enough that you would reach for a phone instead, which is exactly the thing worth avoiding. There is an EE key for the powers of ten, since a mole is 6.022 × 10²³ and nobody types twenty-three zeros. It takes typed input too if you have a keyboard.
  • It does not round, and that is deliberate. Working out how many figures an answer is entitled to is one of the seven things this teaches. A scratch line that quietly handed back 90.3 would be doing that step for you at the exact moment you were being asked to do it, so it hands back the whole number and says to round it yourself.
  • It also never sees the question. It multiplies the numbers you chose in the order you chose them — deciding what to multiply, and which way up, is the whole of what this is for and the scratch line has no part in it. The description in the ⓘ panel used to say this is not a calculator; it now says what is actually true, which is that it does not solve the question for you.

  Still missing: nothing can be handed in and there is still no page for whoever set the work. The icon is still a simple drawing rather than a proper one.

• 0.9.0

  • What you have already worked out stays on the screen. Each step you finish is kept in front of you, in the words you wrote it in, until the question is over.
  • This matters most where one step feeds the next. A proportion asks how many times the recipe you have, and then asks you to use that number. A rate given upside down asks you to turn it over, and then asks you to divide by what you turned over. Both of those used to ask you to remember a number you could no longer see, or to write it on something else — and needing anything other than this app in front of you is friction that has nothing to do with the algebra.
  • It is your own work, not the app's: a line appears only once you have got that step right, and it shows what you typed rather than a number worked out for you. It clears when the next question starts, because working from one question sitting beside another question's numbers is worse than showing nothing.
  • And the closing screen had never been checked for readability. The accessibility sweep is supposed to walk every screen in both light and dark and measure the contrast, the tap targets and the focus outlines. The part of it that drives a set to the end could not actually finish one, so it sat on the question screen and reported everything it measured under the name of the closing screen — for every release there has been. It reaches the end properly now, and the closing screen passed the first time it was really looked at.

  Still missing: there is no way to do the arithmetic inside the app yet, so a calculator is still a second thing to have in front of you. Nothing can be handed in and there is no page for whoever set the work. The icon is still a simple drawing rather than a proper one.

• 0.8.0

  • You can choose how the questions are set now. Pick a topic and it asks which kind of question you want before it starts.
  • They have names rather than numbers, and the name says what the question does: cancelling units offers two steps, three steps or four steps; powers offers squares, then cubes and roots, then past a cube; significant figures offers one rule, either rule, then two rules in order. Every one of them is there the first time you open it. Nothing is locked, nothing has to be earned, and nothing switches you to a different one because of how a run went.
  • Two topics were offering a choice that changed nothing. This is worth being plain about, because the setting has existed since the first release and every question you have ever been given came at the first one.
  • When each was measured against the one below it, six of the fourteen steps produced questions nobody could tell apart — bigger numbers, and nothing else. So the topics now offer what they actually have. Proportions has one kind of question. Fractions has two. The other five have three each.
  • Fractions can hand you the rate upside down. Instead of "the concentration is 4.00 mol/L" you get "the volume one mole takes up is 0.250 L/mol", and the first thing it asks is to turn that over. This is what the topic is named for and it had never once been asked.
  • Cancelling units goes up to four conversions in a row, which is where the long chemistry questions live — a volume of liquid weighed, turned into moles, and then into particles or into the volume it fills as a gas.
  • Two things were found while measuring all of this and both are fixed. A proportion question could ask you to work out a number that was already printed in the question, when a recipe took one mole of the first substance. And on a long conversion, turning one factor over and turning the whole chain over could come to the same answer, so being told which one happened would have been a guess.

  Still missing: nothing can be handed in and there is still no page for whoever set the work. Proportions has one kind of question because a second one means asking about two reactions in a row, and that is a question this has never posed — it is for whoever is teaching to ask for, not for this to decide. The icon is still a simple drawing rather than a proper one.

• 0.7.0

  • You can practise one move on its own now, instead of working whole questions and walking past the step you want four times to reach it once.
  • Whole questions are what makes a move stick once you have it. A move you do not have yet is built the other way round: by doing that one thing again. So there is a second way in — pick one of six moves and get that move, over and over, with the question around it for context but only the one step to answer.
  • The six are choosing the move, isolating the unknown, scaling by a ratio, doing the arithmetic, carrying and cancelling units, and significant figures. Stop whenever you like; it tells you what happened and there is no number anywhere.
  • Nothing about a practice run is stored. There is no session, nothing to hand in, and nothing written to the device.
  • Three questions were asking for letters that were not in the equation. This is the one worth knowing about, and it had been there since the first release.
  • A question would show you PV = nRT and then ask you to rearrange it for ng. Another showed n × M = m and offered you n = m ÷ (Mm). A third showed T(K) = T(°C) + 273.15 and called them TK and TC. Those were names used inside the app, and they were never meant to reach anybody. All of them now show the letters that are actually in the equation.
  • And the answer choices could be read two ways. One option came out as V₁ = (P₁) ÷ P₂ × V₂, which could mean two different things depending on where you put the brackets — and neither of them was the mistake the option was supposed to represent. Brackets go where they change the meaning now, and nowhere else.
  • Both of those were found by building the practice screen, because it is the first thing that ever showed one of those steps on its own.

  Still missing: every question still arrives at the same difficulty when you work a whole set. Nothing can be handed in and there is no page for whoever set the work. The icon is still a simple drawing rather than a proper one.

• 0.6.0

  • It works with no connection now, which is the one that matters most for whoever is doing this on a bus, in a car, or somewhere the signal gives out halfway through.
  • Open it once with a connection and it keeps a copy on the device. After that it opens whether or not there is any network at all — the questions, the marking, the explanations and the list of what changed. Nothing is ever fetched while you work, because nothing ever was: the questions are worked out on the device.
  • And when a newer version arrives, it waits for you. This is the part most apps get wrong, so it is worth saying what happens instead.
  • A newer version does not replace what you have while you are using it. You get a line at the top of the screen saying one is ready, with a button to switch to it and a button to leave it for later. Nothing reloads on its own, nothing is lost part-way through a question, and if you ignore it entirely you get the new one next time you open the app.
  • If you have never opened this before, you are told none of that. There is nothing you missed.
  • You can keep it on your home screen, and the information button now lists how to do that on each kind of device.
  • Also: the panel that tells you which copy of the app is on the device now says so exactly, which matters because the version number on screen is reported by the code that is running — on an out-of-date copy, that is the old code reporting itself perfectly accurately.

  Still missing: there is still no way to practise a single move over and over, which is the thing most worth having when one particular step is the one going wrong. Every question still arrives at the same difficulty. Nothing can be handed in and there is no page for whoever set the work. And the icon is a simple drawing rather than a proper one.

• 0.5.0

  • There is something to open. Until now everything in here worked and none of it could be seen.
  • You get a short page saying what this is and what it is not, and then you pick one of the seven kinds of question and work through five of them, one step at a time. Get a step wrong and it tells you which mistake produces that exact number — and the step stays where it is until you get it, because a step you can walk past is just another question.
  • It can be made easier to read, and those settings never leave your device. Bigger text, more space between letters and lines, showing one thing at a time, and reading the question aloud where the device can do that. None of them is written into anything you hand in and none of them is sent anywhere. That is deliberate rather than an oversight: what somebody needs in order to read something is nobody else's business, and a tool that quietly passed it on would make them disclose it every time they used it.
  • Day and night colours, following the device unless you say otherwise. The colours were measured against a contrast standard before anything was drawn, in both, on every screen — including the ones that only appear after a wrong answer, which is where this kind of thing usually goes unchecked.
  • An information button in the corner, with what this is, how to keep it on your home screen on each kind of device, where the questions come from, what it stores, what changed, the accessibility statement and the licence. The page you see first is moved in there when you press Start, so it is still reachable rather than gone.
  • And a page listing every release, which the panel above links to. The panel itself never shows more than five, and says how many it is not showing.

  Still missing: it does not work offline yet — close the connection and it will not open, which is worst for exactly the people who need it on a bus. There is still no way to practise a single move over and over, which is the thing most worth having when one particular step is the one going wrong. Nothing can be handed in and there is no page for whoever set the work. Every question arrives at the same difficulty for now. And there is no way to be told when a newer version is waiting.

• 0.4.0

  • When a new version arrives, this will be able to tell you what changed — and it will not be an endless list.
  • Every app should say what is different since last time, in the place you already are, rather than sending you off to a website meant for programmers. This release builds the part that holds those words: the newest release plus four before it, and nothing older than that. Everything further back lives on a page inside the app itself.
  • Five is a decision and not a number that happened. A list that grows by accumulation eventually becomes longer than the thing it is describing, and nobody opening a maths trainer wants to scroll through two years of it. The app also knows how many releases it is NOT showing, so it can say so, instead of leaving an impression that five is all there has ever been.
  • The words are now written in one place and only one place. Until now, what a reader would see and what was kept alongside the working parts were two separate lists — and two lists drift, so an app ends up describing a change that is not in the version somebody is actually running. There is one set of words now, and a check refuses any change that lets the two disagree.

  Still missing: there is still no screen, so the panel that would appear on a new version, and the page it would open, do not exist yet. What exists is the list of words and the rule about how long it is allowed to be. Nothing here can be opened and read.

• 0.3.0

  • Significant-figures questions can now mix adding and multiplying, which is the kind people get wrong most.
  • Two rules govern how many digits an answer is allowed to keep, and they are different rules. Adding is limited by decimal places. Multiplying is limited by significant figures. A question that does both — add two measurements, then multiply by a third — needs both rules, in order, and the usual mistake is to skip the first one and round on the fewest digits in sight.
  • Until now this app could not ask that question at all. It could only ask questions that used one rule or the other, so the hardest case in the topic was missing. It is there now, from the second difficulty level up, and it asks about the middle step on its own: how many significant figures is that sum entitled to? — without asking anyone to round it, because rounding partway through is the mistake, not the method.
  • This was found by reading published research on what people actually get wrong, rather than by testing the app against itself. The app's own checks could never have found it: they only ever look at questions the app is able to ask.

  Still missing: there is no screen. Nothing here can be opened and used yet. Everything above lives in the part that works out the questions and marks the answers, and there is nothing to look at.

• 0.2.1

  • Nothing this app says will ever suggest the problem is you.
  • An app that fails to teach someone should leave them thinking that app did not work for me, not I cannot be taught. So none of the words in it talk about effort, or about what someone ought already to know, or call anything easy. Calling something easy leaves only one explanation for why it did not go well, and that explanation is the person reading it.
  • This is now checked automatically every time the app changes, the same way the no-scores rule is, so it cannot quietly stop being true.

  Still missing: there is no screen, so none of these words have reached anybody yet. The check is in place for when they do.

• 0.2.0

  • No scores, no streaks, no targets, and nothing congratulating anybody.
  • A streak teaches you to protect a number, and it makes stopping feel like losing something. That is the opposite of what is useful for the person who most needs to sit and do twenty of these.
  • What you get instead is what changed:
  • A mistake made once is not a pattern, and nothing is said about it.
  • Made twice, it turns up at the end.
  • Made three times, it is named while you are still working — once, with what fixes it — and then not raised again.
  • If you were making the same mistake and then you stopped, it says so. Only if that actually happened, and one right answer after a run of wrong ones is not enough to count.
  • And if a whole set went wrong, it does not tell you how many. Four questions and none right is true and is no use to anybody. It says what the mistake was and that it is one thing to fix, not a pile of questions to do again.
  • Also fixed: a multiple-choice question could mark the wrong option correct if it was one the marking had not been told about. On a rearranging question that meant the upside-down answer would have been accepted. Nobody could have run into this yet — there is no screen — but it would have been waiting.

  Still missing: there is no screen, so nothing above can be seen. And there is still no way to practise a single skill on its own, which is the thing most worth having if one particular move is the one going wrong.

• 0.1.0

  • The first release. It works out questions and marks answers; there is nothing to look at yet.
  • What it covers. Seven kinds of algebra, and only these seven, because each one is something you cannot do a chemistry calculation without: rearranging a formula, proportions and cross-multiplying, scientific notation, powers and roots, fractions and rates, cancelling units through a chain, and significant figures.
  • What makes it different from a calculator. It does not just mark an answer wrong. It works out which mistake produces that exact number — the ratio used upside down, the exponent multiplied instead of raised, a conversion factor turned the wrong way — and says so.
  • And it refuses to guess. If two different mistakes would produce answers you could not tell apart, the app will not ask that question at all. It would rather stay quiet than tell somebody they have a misunderstanding they may not have. When it genuinely cannot tell what happened, it says that too, rather than picking something plausible.

  Still missing: no screen, nothing to open, no way to practise a single skill on its own, and nothing to hand to anybody at the end.