A review you had already finished no longer asks you to continue it. If you marked even one claim "unsure", reloading the argument put the Review panel back to "Continue review" — your work was all still there, and opening the review corrected it, but the panel was reporting the wrong thing on arrival. It now reads "View Results" as soon as you have reached the results step.

The grade beside an argument's conclusion now agrees with your review's verdict. If you rejected one of the connections between premises, the review said the argument did not hold up while the grade beside the conclusion still read "Sound" — because the grade was ignoring your decisions about the connections and assuming every one of them held. It now shows the grade from your own evaluation, the same one the Review panel reports.

Signing up is no longer hidden behind a page that says "Log in". There is only ever one door — signing in with a provider for the first time creates your account — but nothing said so, so a first-time visitor who followed "State Your Case" from the home page landed on a login page that greeted them with "You need to log in to perform that action". That page now introduces itself as "Sign in or sign up" and says plainly that any option on it will either sign you in or create your account, the buttons read "Continue with…", and the button in the top bar reads "Log in / Sign up".

"Explore Arguments" on the home page now takes you to the arguments. It used to scroll you a little further down the same page.

A review now tells you two things instead of one, and they no longer contradict each other. Where the conclusion used to carry a single word — "Sound", "Indeterminate" — you now see what the conclusion came out at (true, false or unknown, and whether that was your own answer or something the argument reached on its own) beside what the argument itself did (it reaches its conclusion, it doesn't reach it and why, or its premises contradict each other). They are separate readings because they answer separate questions: a conclusion can be true while the argument that offered it did no work. The same two readings now appear on the conclusion row, in the Review panel and in the text you share.

Rejecting a step no longer invents a fact. Turning down one of an argument's connections used to be read as "this consequence is false", which was never what the control meant — it now takes that step out of the reckoning and leaves the argument to be judged on what remains. Premises you turned down are counted beside the outcome rather than folded into it, so "Reaches its conclusion" and "1 premise rejected" can honestly appear together.

If your own answers contradict each other, the review stops and shows you why. When you grant a step and then assign values that break it, the results page now names the premise, spells out what granting it commits you to, shows where any value it depends on came from if you did not set it yourself, and offers you both ways out: reject that step as not holding in this case, or change one of the answers. It re-checks after every change, and the review cannot be finished or shared until it comes out coherent. Nothing there says you made a mistake — often the honest reading is that the step generally holds but not here, and there is now a way to say exactly that.

You are told at the start when an argument's premises cannot all hold at once. Nothing can be worked out from a set like that, so previously you could spend a whole review on an outcome no answer of yours could have changed. The notice appears before the first claim, never blocks you, and costs you no progress — and if you turn down the step responsible, the rest is assessed normally.

Publishing is blocked while your own premises cannot all hold together, and you are told while you are writing rather than when you press Publish.

You can see how other readers decided a step. Once enough people have judged a premise, the review shows whether they generally grant it, generally refuse it, or are split — while you are still undecided, and never after you have answered.

The review panel's counts now read "claim steps" and "premise steps". They were labelled as the argument's claims and premises, which they are not: a review skips the parts there is nothing meaningful to judge, so the two numbers can differ.