Linear's landing is a conviction bet.

The hero asks for nothing.

h1 8 words · subhead 11 words · CTA buttons above the fold 0 · /signup links on the whole page 2 · first in-flow CTA at 93% scroll

The hero holds an 8-word h1, an 11-word subhead, one campaign link — and no CTA. The only conversion control above the fold is the "Sign up" pill in the nav corner. The whole 10,781-pixel page contains two links to /signup; the first inside the page flow appears at 93% of the scroll. On mobile, where the campaign link is dropped, the hero body has no clickable element at all.

For pre-convinced visitors the nav suffices — and Linear's traffic may well skew that way; we can't see their mix. But a landing page earns its keep on the undecided, and the undecided get twelve screens of product without one contextual ask — only the same nav pill they've been scrolling past since the fold.

RecommendationOne "Get started" beneath the subhead, mirroring the nav pill. The hero keeps its air; the page stops outsourcing its only ask to a corner.

The hero's one link points off the page.

The single link inside the hero body is the campaign banner — a pulse dot, a few words, and a destination off the landing page. At our 2026-06-11 capture it read "Issue tracking is dead," pointing to linear.app/next (boxed top-right in FIG. 03); three days later the slot had rotated to "Coding Sessions," pointing to /coding-sessions. That rotation is the deeper tell: whatever the week's message, the hero's one clickable element is leased to the campaign of the moment — never to the product's own ask. It sits right where a hero CTA usually lives, so the one thing a visitor can click in the hero exits the conversion surface. A launch window justifies any single instance — it's a campaign with its own microsite, and they may be optimizing for reach, not signups. Worth noting: their own mobile build hides the banner, which says it's understood as optional.

RecommendationMove the campaign to a slim ribbon above the nav — the standard placement keeps the launch story sitewide — and return the hero's action slot to the product.

Eight logos, then 7,600 pixels of silence.

The spine is right: claim, proof, demonstration, testimony, ask. The proportions aren't. Between 13% and 84% the page offers no customer evidence and no ask; every mid-run link points deeper into product. A visitor who exits anywhere in the five-chapter run — 8.4 desktop screens of it — leaves with logos but no number, no story, no ask.

RecommendationOne proof beat per chapter: a single customer line — name plus one operational number — under each section paragraph. Five lines close a 70% evidence gap without touching the layout.

The brightest pixels carry the lightest claims.

A row-brightness scan of the full-page capture (median row luminance 10 of 255 — this is a dark page) finds exactly one sustained bright band: the two testimonial cards, 84–89% of the scroll. Card one — Gabriel Peal, OpenAI: "You just have to use it and you will see, you will just feel it." Fifteen words, nothing a reader can verify. Card two — Nik Koblov, Ramp: eleven words on speed, no number. The page's only quantified proof — "Linear powers over 33,000 product teams" — sits beneath them in body-text gray.

By luminance, these cards are the loudest moment on the page, and they spend it on feel. The OpenAI mark does real work; the sentence next to it adds nothing a founder can check.

RecommendationKeep the cards, upgrade the cargo. Swap one quote for a line carrying an operational number, and promote the 33,000 figure out of gray body text into the bright tier.

In the lab, the LCP image is lazy.

Lighthouse 12.8.2, median of 3 runs, re-confirmed against the live page 2026-06-14 (Chromium 148.0.7778.96). It pins the LCP element to the hero frame image, and Lighthouse's own "Largest Contentful Paint image was lazily loaded" audit fails on every run; in the captured DOM, all three 1920×1080 frame images ship loading="lazy", and one — a grain texture rendered at 8% opacity — is requested at quality 95 and double DPR.

These are throttled lab numbers, not field data; Linear's real-user metrics may look different, and we can't see them. But lazy-loading the LCP image costs every visitor something, and the contrast with CLS 0 is the tell: this team measurably sweats layout stability — loading priority hasn't had the same pass.

Recommendationfetchpriority="high" and eager loading on the hero frame image; serve the 8%-opacity grain at a compressed tier or generate it in CSS; re-run the lab.

Twenty-three mentions of agents, no on-ramp.

"agent / agents" rendered on the page ×23 (×24 in the markup — the 24th is the h1's screen-reader copy) · sentences addressed to non-agent teams 0

The page renders "agent" or "agents" 23 times — 24 in the source, where the h1 ships a duplicate copy for screen readers. The h1 commits to them, the subhead names "the AI era," two chapter headings reference agent workflows. What the page never includes: a sentence addressed to a team that hasn't adopted agent workflows yet.

This is a deliberate, species-level bet — the h1 commits to it outright. Linear presumably knows what share of its 33,000 teams already run agents; we don't. From outside, the exposure is visible: a 2026 buyer who runs product development without agents reads a page describing someone else's workflow.

RecommendationOne bridging line under the manifesto — the feature set on this same page (Issues, Cycles, Customer Requests) predates the agent era and works without one. Test whether it lifts signups from non-agent traffic without diluting the bet.