What Shipped This Week: Faster Pages, No More $1 Trial, Chat Timeouts Fixed

The biggest change this week isn't a feature — it's how fast the site gets out of your way. We found and fixed four separate causes of a 17-to-25-second homepage load, on top of simplifying checkout, fixing a chat timeout that was cutting off long responses, and bringing back tools that had quietly gone missing.

None of this is flashy. All of it is the kind of thing that either saves you a click or stops you from bouncing off a slow page before you even get to try the product. Here's the rundown.

What shipped this week

Feature	Category	Who it's for
Marketing page speed fix (17-25s → near-instant LCP)	Improved	Anyone visiting writingmate.ai from a link or ad
Checkout skips the $1 trial, charges plan directly	Improved	New signups choosing a paid plan
OpenAI chat timeout raised to 800 seconds	Fixed	Anyone running long reasoning or long-output prompts
Four free tools + Discussion Board restored	Fixed	Free-tool and no-login users
Full-width video embeds in blog posts	Fixed	Blog readers on desktop

Marketing pages load dramatically faster

We measured the homepage on a throttled mobile connection (4x CPU slowdown, Slow 4G) and found first content painting at about 0.9 seconds, but the largest visible element — the hero heading — not finishing until 17 to 25 seconds later. That gap is what search engines and real visitors both penalize hardest.

We traced it to four stacked causes. The hero heading and subtitle were animated in with framer-motion, which server-renders them as invisible (opacity:0, blurred, shifted down 50px) until React hydrates and animates them into view — so the page's most important text was literally not there for the first several seconds. We removed the entrance animation from above-the-fold copy so it renders as plain, immediately paintable text.

The second cause was a render-blocking stylesheet. Around 60 requests — font preloads, the logo, ten external favicon lookups, and dozens of JS chunks — all fired at once and starved the one ~33KB stylesheet that was actually gating the page's first paint, delaying it from 219ms to 773ms. We now inline that critical CSS directly into the HTML, removing the extra round trip entirely.

Put together, the homepage and other marketing pages should now feel meaningfully snappier, especially on slower connections. If you bounced off a slow-loading page before trying Writingmate, it's worth another look.

Checkout is simpler: no more $1 trial

We removed the $1, 3-day trial from new checkouts. New signups now pay their selected recurring plan immediately instead of starting on a refundable trial charge. The old flow added a step and a confusing line item without changing anyone's actual decision to subscribe, so we cut it.

Pricing, onboarding, and the upgrade paywall now say what's actually happening — "Start for free" for the no-card app preview, and "View plans" or "Upgrade" for anyone heading to paid checkout. We also renamed the purchase button on /upgrade from "Continue to checkout" to just "Upgrade," which is shorter and says exactly what it does.

If you're already in a grandfathered trial, none of this changes your current terms — this only affects new checkouts going forward.

Long chat responses no longer time out

We caught two 504 timeout events in production where OpenAI-model chat completions hit our old 300-second server limit while the model was still generating a long response. The request had done nothing wrong — it just needed more time than we were giving it.

We raised the timeout on the OpenAI completions route to 800 seconds, while every other API route keeps the existing 300-second limit. If you regularly ask for long, detailed answers or run multi-step reasoning prompts, this should stop those requests from getting cut off mid-generation.

Free tools and Discussion Board are back

Four historical free-tool URLs had dropped out of our 55-tool catalog at some point and weren't reachable anymore. We restored all four, along with their original canonical URLs and sitemap entries, so old links and bookmarks work again instead of 404ing.

We also exposed the Discussion Board generator — one of our free, no-login tools — through a public iframe route, and added regression tests so both the tool catalog and its embeds stay in sync going forward instead of silently drifting out of the sitemap again.

A cleaner blog reading experience

If you read our blog on desktop, embedded YouTube videos in articles were rendering at their old intrinsic size — a tiny 300 by 315 pixel box in the middle of an otherwise full-width article. The full-width rule for embeds only applied inside a mobile media query, so desktop never picked it up.

Video embeds now render block-level, full article width, at a proper 16:9 aspect ratio, with rounded corners and spacing — across every post, not just the one it was first noticed on.

Fixed this week

• Chat completions to OpenAI models no longer 504 on long responses — timeout raised from 300 to 800 seconds
• Blog video embeds now render full width on desktop instead of a small fixed box
• Four free-tool URLs and the Discussion Board generator, previously missing from the catalog, are reachable again
• Homepage and other marketing pages no longer hide hero text behind an invisible entrance animation
• A render-blocking stylesheet on marketing pages is now inlined instead of loaded as a separate, contended request
• Stale "3-day free trial" CTAs across pricing, onboarding, and the upgrade paywall now say what actually happens at checkout

Thanks for reading this far — if you hit any of the bugs above before this week, sorry about that, and they should be gone now. For the full history of what's changed, week over week, the changelog has every entry.

— Artem