New Runs, open leaderboards, and a tile layout

Trail Blaze: New Runs

The Trail Blaze lobby now has two buttons: Start Weekly Run and Start New Run.

• Weekly Run is the map you already know: the same map for everyone, all week, where every attempt reveals more of it.
• New Run draws a fresh random map just for you, for that one attempt. Nobody else gets it and you can never replay it, so every New Run is a cold read, like a real interview.
• Same rules either way: six problems, 90 minutes, Submit only, two chances.
• New Runs pay node and clear bonus XP too, up to one full clear's worth (365 XP) per day.

New Runs have their own leaderboard at Trail Blaze → New Run leaderboard. It shows each player's best New Run, this week or all time, ranked by problems solved, then fewest wrong submissions, then time. The lobby shows both boards side by side.

Trail Blaze: see how others did it

Every row on the leaderboard now has a View link that opens that attempt read-only: who played it, how it ended, the map as they left it, and every stop on their route with its time and each submission they made there, wrong ones included. Code sits behind a toggle and stays hidden if the player marked it private.

To keep the weekly map fair, you can open other players' attempts on it once you have finished an attempt of your own on that map, or once the week is over. Attempts still in progress never open. Any finished New Run can be opened, since nobody else will ever play that map.

Trail Blaze: a fairer weekly leaderboard

Every attempt on the weekly map reveals the problems it visits, and your later attempts see them. A clear on your seventh try had most of the map scouted, so it should not beat someone's clear on their first. The weekly board now ranks by:

1. Problems solved, most first.
2. Which attempt it was, earliest first.
3. Wrong submissions, fewest first.
4. Time to the last solve, fastest first.

The lobby explains the order beside the board.

More Trail Blaze

• Drawing tool. The drawing overlay from practice and rooms now works in a run. Peeking at the map keeps your drawing; moving to the next problem clears it.
• Following. When someone you follow is mid-run, the Following band says "On a Trail Blaze" and links to the lobby. The problem itself is never named, so it can't spoil the weekly map.
• Dashboard card. Fixed the missing space in "6 attempts on this week's map".

Tile layout (beta)

On desktop, a Tile layout / Classic layout switch beside Random on the problem page lets you try a dockable workspace. Every panel (Description, Editor, Test cases, Submissions, Solutions, Live feed, People, Room chat, Global chat, and Lesson or Visualize where a problem has them) is a tile you can drag, split, stack and close, and your layout is remembered on this device.

• Chat tiles show unread counts on their tabs.
• Run and Submit bring the Test cases tile back into view (reopening it if you closed it), so a result never lands out of sight.
• Picking a past submission brings the Editor to the front.
• Classic layout is unchanged and stays the default.

Smaller changes

• "Activity" is now two names. On the problem page, the discussion board with shared solutions is now Solutions, and the live submission stream is Live feed. Old links still work.
• View notifications at milestones. Instead of a notification for every person who opens your shared solution, you now get one when it reaches 5, 20, 100, 500, 1,000 views and so on. Your older view notifications stay in your inbox.
• 3.0 vs 3. When a Python solution returns 3.0 where 3 is expected, the two look the same but the answer is still wrong. The failed case now carries a note explaining why, with a tip for returning an integer.

Fixes

• Stopwatch. It no longer resets to 0:00 when you spectate someone or open a past submission and come back. An Accepted submission now pauses it; typing in the editor again resumes it.
• Numbers in results. Output is shown exactly as your program printed it, so 3.0 no longer appears as 3, and very large integers are no longer rounded.
• Problem fixes. A couple of problems moved to a better-fitting tier, and one had its heaviest tests trimmed so the intended solution fits the time limit in every language.