We have published roughly thirty World of Warcraft: Forever items in the last four days. Not one of them came from playing the game, because none of us has a beta client.
That is not a confession of inadequacy. It is a description of a source, and readers are entitled to know exactly what a source is before they rely on it.
Where our Forever knowledge comes from
First-party developer statements. The class deep dives dated 30 September, the Legacy System post dated 23 September, the named Wow Weekly posts, the podcast episodes. These are Blizzard writing down numbers. They are the highest-grade material we publish anywhere and they are the only Forever material that contains values you can argue with: level 10 talents, level 32 Shadow Word: Death, 16 points per character, 65 challenges.
First-party CDN metadata. The wowdev2 manifest, which tells us a build exists, whether it is genuinely new or a re-point, and whether regions are in parity. Infrastructure, not content.
Our own earlier beta observations, dated in the archive: datamined mounts and quest text, addon ecosystem checks, client tooling, player workarounds, the auto-shot bug. All of it recorded while it was true, all of it subject to the same decay as anything a beta changes.
Where it stops
We cannot tell you how the game plays. Not one line of our Forever coverage is a first-hand account of combat, questing, grouping or progression. When we publish a class deep dive's numbers, we are reporting a developer's post, not evaluating the class.
We cannot tell you whether a bug is fixed. We have carried the auto-shot and wand cast bug as open since 23 September. Three builds have shipped since. None of them has notes. It stays open because the honest answer is "we do not know", and it will stay open until someone with a client says otherwise.
We cannot tell you if a date will hold. Forever launches 4 November, 33 days from this post. Blizzard's own deep dives say abilities and talents "may continue to evolve throughout the beta test and are subject to additional changes". Our launch countdown exists because a date is worth publishing, not because we expect it to be accurate.
We are not going to rank the servers in the scene. That one deserves a sentence of its own, because the temptation is real. We track 148 servers and we have never played any of them. Our population bands are the operators' own published widgets, read on a date and labelled as self-reported, and where we cannot find a source we say so and refuse to band the entry at all. That is the whole policy, and it exists because we got it wrong repeatedly this week in both directions.
What we did this week that is worth reading as a lesson
Four days, five published errors, all the same shape:
- ArgusWoW was rated playable and medium. It had been shut down since May 2022 and its page still showed a live-looking player count.
- Warmane was published as possibly dead. It has a forum post from an hour before we looked, and we had failed to open a browser.
- Epsilon was demoted to dev by a data cleanup whose stated goal was "0 playable unknowns". Nobody had opened the site.
- Six entries sat at tiny or small on Discord counts we could no longer refresh. All six are now
unknown.
Every one was a value with no source attached, and every one was found by going and looking rather than by reading the file again. We have added four build gates this week as a result, and the newest one fails if any entry carries a population band with no website behind it.
What that means for trusting us
Trust the dated readings. If we say the Legacy cap is 16, that is Blizzard's number with a date on it and you can check the post.
Distrust our silence on anything experiential. We have no opinion on whether the game is good, and we will not develop one by inference.
Hold the two open claims loosely. The auto-shot bug and the launch date are both live questions, and both are recorded as open rather than resolved.
Ignore our earlier numbers where they conflict with a newer post. When a developer post supersedes a datamine, the developer post wins. We strike the old value rather than quietly replacing it, which is our correction policy and the reason you can audit us.
The uncomfortable summary
We are a tracker and a quote service with a CDN watcher attached, publishing during a beta we cannot enter, and we have been wrong five times this week in public and corrected each of them visibly.
That is a worse position than having played the game and written a review. It is also the position we can actually be in honestly, and given a choice between that and writing confident reviews we did not earn, this is not close.
SOURCES: this site's own published archive, all items cited above with their dates, read and verified on 2 October 2026. The five errors named are documented in our correction posts and the numbers in them are our own data. We have not played World of Warcraft: Forever and have no beta client. Related: the build reports explained, the correction policy, and how we review a private server.