Every submission in one place, and a tier above Gold

My Submissions browser

/submissions (in the user menu) is now a split-pane browser for your whole history.

• Filters for search, verdict, language, topic, difficulty and source (practice, Trail Blaze, collab sessions, study groups). They live in the URL, so a filtered view can be bookmarked or shared.
• Two views. The attempts log lists every submission newest first. The by-problem view groups them into one card per problem with a verdict timeline, and a card with more than 100 attempts can load older ones.
• A summary strip over the list: what the current filters add up to.
• A code preview of the selected submission, read-only, with Compare to diff it against another of your attempts on the same problem.
• Keyboard: j and k move through the list.
• Export downloads a zip of your accepted solutions.

On a phone the list keeps plain links to each submission's page.

Platinum

There is now a fourth difficulty tier above Gold. Platinum is for problems that need a contest technique or a proof that isn't obvious: matrix exponentiation, Euler tours with a Fenwick tree, subset-partition DP and the like. A first solve is worth 600 XP (Gold is 300) and 13 study-group points (Gold is 10).

Nine of the hardest Gold problems have moved up. If you already solved one, your XP is topped up to the Platinum price. Solving a Platinum problem earns the new Platinum Standard ribbon.

Platinum has an ice-blue label and a pearl-to-lavender medal, and gets its own column on the leaderboard, its own filter on the problem list and its own slice of your profile stats. It is opt-in: the daily challenge never picks a Platinum problem, and Trail Blaze maps stay Bronze to Gold.

How you solved it, and how everyone else did

Every accepted submission is now labelled with the techniques it uses (say, two pointers and sorting) and its time complexity.

• Labels on your submissions. They appear as chips on your submission history in the workspace, in the My Submissions browser and on each submission's page. Only you see the labels on your own code. A fresh accept picks up its labels a few seconds later.
• "Wrong label?" If a label is off, the button on the submission's page lets you pick the right techniques or complexity. Your corrections go straight into making the labels better.
• Approaches. On a problem's Activity tab, Show approaches reveals how everyone who solved it went about it: the share of solvers who used each technique, and how their solutions break down by complexity. Your own approach is outlined. It stays collapsed until you click, so nothing is spoiled by accident. Click a technique to see the shared solutions that use it.

Test cases you can edit

The test case panel has been reworked:

• Samples are edited in place. Change a sample's input and the edited version runs in its place. Reset cases goes back to the originals.
• Case tabs list every case you can edit, with an x to remove one. + copies the case you're on, as many times as you like.
• One field per parameter, labelled name =. String parameters take plain text, no quotes needed. Source flips the case to plain text, one value per line, for quick copying and pasting.
• No more hand-written expected output. When you run a custom case, the problem's reference solution runs on the same input and its answer becomes the expected value, labelled as coming from the reference. If you edit a case after running it, the old result is flagged as out of date instead of sitting next to the new input.

Trail Blaze

• Open to everyone. Signed out, /trail-blaze shows the pitch, this week's countdown, both leaderboards and the rules, with "Sign up to play" where the Start buttons would be. A shared link now shows something. Playing a run still needs an account.
• Tabbed leaderboards. The lobby's two stacked boards are now Weekly Run and New Runs tabs. Each has a short explainer: what the board is, who is on it, and how it ranks.
• More ways to slice them. The full Weekly board has an All attempts / First attempts only toggle, where first attempts are ranked only against other first attempts. The New Runs board covers the last 7 days, 30 days, year or all time.

Smaller improvements

• Search by words. Problem search used to need the whole query as one unbroken piece of text, so "max trading profit" missed "Maximum Trading Profit With Fee". Each word now matches on its own, in any order, against the title, slug or a category.
• Problems open on Description. In the tile layout, a Lesson or Submissions tab you left open on one problem used to open on the next one too. Each problem now opens on its Description tab, unless a link asks for a specific tab.
• Tougher tests. We checked every problem's tests against known wrong answers and too-slow approaches, and nearly 70 problems let some of them through. Each one now has cases that catch them: checks that only look at part of the input, greedy choices that only work sometimes, 32-bit overflow, and brute force that is fast on small inputs but not on large ones. Submissions you already have keep their verdicts.
• BlazerBot and line numbers. BlazerBot now sees your code with line numbers, so "I changed line 24" points it at the right line. Before, it could miscount and wrongly claim it was looking at an older copy of your code.

Fixes

• Tile widths. On a problem without a Visualize tile, its column's width was spread across every other column, so a narrow sidebar could suddenly take a third of the screen. Each column now keeps its exact size as you move between problems.
• Tiles return where you left them. A tile one problem doesn't offer (Visualize, Lesson) used to come back as a tab at the end of the problem group on the next problem that had it. It now goes back to its own column, its spot beside the editor, or its floating window.
• Adding custom cases. The + button added one custom case and then silently did nothing.
• Stats page. The first visit to /stats after its 15-minute refresh window showed the old numbers, and only a reload showed the new ones. It now waits for the fresh numbers.
• Link previews. The card shown when you share an AlgoBlazer link was still in old colors and said we support 6 languages. It now matches the site and says 8.