Agent Ping Pong, a Print Ad, and a 2-Out-of-10 Triage

What people were saying

Quiet day for shipping, loud day for opinions, most of them from Linear's own people.

The interesting thread was internal disagreement about code review, which is notable given the company just put its name on a diff viewer. @adjohu posted four words ("human code review is a bottleneck") and left them there. Eight hours later @moofeez said he really doesn't like code review turning into agent ping pong, which is either the opposite position or the same position at a later stage of grief. Replies split accordingly: one camp argued agent-to-agent review is fine as long as it happens locally before the PR exists, another argued the real job now is checking whether the agent followed your conventions, and one reply went for the throat, saying nobody wants to review code the author didn't write. Someone also noted that a sleek diff viewer doesn't much help with a 10,000-line PR, which feels like a product insight disguised as a complaint.

The day's most-loved artifact was non-digital: @pdotcv showed off a full print ad in Arena magazine. Design X immediately flooded in with old Volkswagen comparisons, applause for the body copy, and a lengthy tangent about justified text and a Super Metroid guide whose author allegedly picked words to make the monospace grid line up. One quote-poster grumbled that the ad never says what the product is or does, which is both true and, in a magazine that costs money to hold, arguably the point.

Elsewhere on the team, @emieljanson went off-brand-but-on-brand, publicly asking Apple to hold quality Wednesdays after macOS search failed to find things. It was the day's highest-liked post about anyone from Linear, which says something about the market for Apple complaints.

The pushback

The sharpest outside data point came from @jessemchen, whose team got two months of unlimited tokens from a frontier lab and graded the results. Refactors and computer-use QA scored 10/10. Agentic Linear triage scored 2/10, with the diagnosis that either the tickets were too vague or the right action was too ambiguous. His conclusion (tokens aren't the bottleneck, judgment is) lands awkwardly close to the triage-and-prioritize part of the product. In the same neighborhood, @gkhalin asked outright whether Linear or any AI-first tracker actually lowers the cost of deciding what to work on next, having done the ritual evening sort through twenty saved Slack messages.

Notable anecdotes

@praveenvnktsh open-sourced a personal build agent triggered by dragging a card to Todo, which then plans, ships, redeploys and comments back on the issue. @nikolay_dp called working GitHub and Linear issues this way his favorite workflow, with caveats. And a widely-read essay by @aporia9n about company names being cheap asymmetric bets listed Linear alongside Ramp, Stripe and Cursor, prompting one reply to point out that Cursor's actual company name is Anysphere.