Linear on X

Praise keeps naming the same things: speed, a clean interface, small touches — "super fast, great UI/UX", a backlog that fits on two screens, a build that runs well on a Duo. A group just as large values Linear as the layer agents work through: a model started from a mention on a ticket in under 90 seconds, loops on GitHub and Slack, agent session APIs wiring issues into Claude Code and Codex, and one person keeping a private team for personal productivity agents. One explains the draw as design for humans — triage, blocking, parent and sub-tasks — that agents can then use.

Code review is a distinct third theme at 9 people: "guided code review", Diffs that group a large change so it stays readable, and asking about a line before commenting.

5 people like it with a qualifier — one hated it as a project management tool but cannot live without it for agent collaboration, another calls it the best UI with weaker branding. 3 are plainly unhappy, one calling it annoying for a small team.

Ask a quick question about code and the agent spins up a sandbox — "asked a quick question, got a sandbox and a coffee break" — so people give up before an answer arrives. Agents also open issues faster than anyone closes them; one says agent-created issues overtook human ones in spring, and another wants to know how much of that growth ever leaves the backlog, or how to separate a good agent issue from noise. GitHub's fix-it flow throws a merge commit into Linear's git history and, for one person, hid the open-on-GitHub link.

The rest is quieter. Some want a human kept in the loop before agents decide what is urgent or ship a fix; one asks who picks whose customer requests the agent reads. A few cannot get environments working — manual setup, an agent picking the wrong environment, and docs that do not mention the thing that is blocking them. Two name price or issue limits as why they left.

Jira is the benchmark, usually as something people escaped from: one credits Linear with a request that would once have taken stakeholder calls, 10+ Jira tickets and three months of backlog waiting, and another is trying to move their work team off Jira. Claude and Claude Code are weighed differently — one keeps all planning in Linear so Claude has the context and updates issues and raises pull requests itself, while another asks why use Linear for this when Anthropic does it better, and a third measures Linear against simply asking their own Claude Code.

Plane comes up twice as the cheaper move: two people switched over price, and one misses Linear's fast interface afterwards. Two would rather work in GitHub — one waiting only for diffs on mobile, another arguing specs belong in the repo and tickets are for incidents and triage. Two use a tool they built themselves, one after hitting issue limits. Then single voices: moving company docs off Notion, still reaching for Confluence to nest documents, calling the guided reviews a fork of Graphite's code tours, and asking how hard it would be to migrate tasks off Linear.

Under posts about 400% growth or fixes landing faster, people ask what is being counted: is that chart cycle time or throughput, did p95 drop because fixes got faster or because triage did, do issues opened by humans and by agents break out separately, and how many were closed. One states the worry that once agents drive the counts, issue numbers measure how chatty the agents are rather than how much work exists.

The next group asks whether the trail survives. Does the agent take the "why we built it" from a decision doc or does someone still write that by hand, can it show which customer request led to each shipped change, and how does that link hold when agents split or rewrite issues along the way. The smaller questions are practical: is it worth trying, what took Linear beyond issue tracking, does using agents need an API key or does OAuth do, and where to find a shortcut or the command bar on a foldable screen.

What they describe is small and concrete: highlight a few lines of a pull request and ask "was this necessary and if so why?", with the answer arriving as an inline thread only they can see. The sandbox is what gets in the way — one waits minutes while it sets up an environment and a coding agent — and another says a fast agent with the diff and the linked issue preloaded would cover most "why is this here" questions, leaving the full sandbox for when the answer needs the code run.

Docs are the next theme at 2 people: folders and nested folders in Team Documents, the one thing still sending someone to Confluence; docs that live inside a doc the way sub-issues live in an issue; a doc site for the whole workspace; and callouts that stand out. 2 want Diffs led by what changed for the user, with the current line kept anchored.

Beyond that, wishes are single voices: a native Linux app from paying users, agent identities, a calendar view, marking an issue blocked while creating it, an official CLI, personal workspaces, and external doc sharing.