We noticed something odd in our referral logs last spring: a small, steady stream of readers arriving from a domain we had never heard of, staying for eleven, twelve, sometimes nineteen minutes, and then leaving without bouncing. A reader shared the source in our community Slack — a weekly journal called Zq686, which publishes one long-form explainer on technology, design, or decision-making every week, each one distilled from primary sources and closed with a single question. No infinite feed, no breaking-news churn. Just one piece, once a week. We followed a project built around that reading habit for fourteen weeks to see whether the format could survive contact with real audiences.
Week 0: The Brief
The project was run by a two-person editorial team we will call North Desk, working with a sponsor who wanted to test whether slow, primary-source reporting could outperform the usual syndicated roundup. The constraint was deliberate: no topic shorter than 2,500 words, no piece published without at least three cited primary documents, and every essay ending with one open question for the reader. The team had a budget of roughly six weeks of full-time work and a distribution list of 4,000 subscribers inherited from a defunct newsletter.
Their first decision point came before writing anything. They chose to distribute through RSS and email rather than social feeds, on the theory that a slow-reading journal needs a slow channel. That choice would later define the entire measurement plan.
Weeks 1–4: Building the Habit
The first four issues covered a municipal broadband rollout, a typographic redesign of a transit map, and two decision-making post-mortems from inside small companies. Open rates started at 38 percent and drifted down to 31 percent by week four — normal decay, but the team had expected better. The obstacle was not the writing. It was timing. Readers were opening the email on Tuesday morning and finishing the essay on Thursday night, which made a weekly cadence feel both too fast and too slow.
North Desk responded by moving publication to Thursday and adding a short editor's note at the top of each issue explaining what the reader would get and roughly how long it would take. That single change lifted completion rates — tracked via a simple scroll-depth pixel — from 44 percent to 61 percent over the following three issues.
Weeks 5–9: The Compression Problem
Around week five, a different obstacle surfaced. The design analysis essays were image-heavy: archival posters, type specimens, annotated screenshots. On mobile connections, page weight was killing the reading experience. We watched one 3,000-word piece take 8.4 seconds to become fully interactive on a mid-range Android phone. Readers who started the piece on desktop and finished on mobile were dropping out at the point where the first large image loaded.
This is where the journal's own back catalog became a test case. The team pulled the ten published essays, ran every inline image through an automated pipeline that converted them to WebP, stripped redundant metadata, and served them from an edge CDN with responsive breakpoints. They kept the visual quality identical at reading size. Median page weight across the archive fell from 2.9 MB to 810 KB. Mobile completion rose to 74 percent within two issues, and the bounce rate on the two heaviest design essays halved.
Weeks 10–14: What the Numbers Said
By week fourteen the journal had published fourteen explainers, averaged 6,100 unique readers per issue, and — more interesting to us — recorded a 68 percent completion rate on pieces over 3,000 words. That is unusual. Most long-form publishers we track celebrate anything above 40 percent. The team attributed it to three things: a fixed weekly rhythm, a closing question that invited replies, and a reading experience that never made the reader wait.
Perhaps the most striking result was the reply rate. Roughly 9 percent of completing readers answered the closing question by email, and 2 percent of those replies turned into essay tips that became future issues. The slow format was generating its own editorial pipeline.
We asked the editor what they would change. The answer was blunt: start the image pipeline on day one. "We spent four weeks learning that a good essay on a slow page is still a bad essay," they said. "The words were never the bottleneck."
What We Took From It
Three lessons stand out for anyone running a long-form product.
- Cadence beats volume. One piece a week, published on the same day, outperformed three pieces a week published whenever they were ready.
- Reading time is a design problem. The team's biggest single gain came not from better writing but from cutting page weight and letting images load progressively.
- End with a question. A single closing prompt turned passive readers into correspondents, and correspondents into sources.
The full archive and the journal's stated editorial method are documented on its own site, including a short page describing how each explainer is sourced before publication — the same page the North Desk team used as their style guide. You can read that method here: how Zq686 sources and structures each weekly explainer.
Fourteen weeks is not a long experiment. But it is long enough to suggest that the appetite for slow, sourced, single-topic reading has not disappeared — it has just been buried under fast pages and faster feeds. A weekly journal that respects both the reader's attention and their bandwidth can still hold it. The North Desk team is now planning a second season, with the image pipeline already in place and the cadence locked to Thursday.