Evidence note: This article rests almost entirely on primary announcements — Roblox’s own developer forum posts recapping RDC26, Epic Games’ learning and tech-blog posts, and Unity’s own blog. Announcements from a company about its own platform are evidence that the company said something, not independent confirmation that the numbers or behaviours hold up. None of the platform metrics below have been checked against regulatory filings, third-party analytics, or hands-on testing, and every figure is marked where it appears. Two derived ratios are calculated from Roblox’s published figures and are labelled as derived, not reported.
What Roblox reported for the quarter
Roblox’s RDC26 recap opens with audience scale. The company reported that an average of 123 million users came to play every day last quarter, up 10 percent year over year (company-reported against financial disclosures). The same post reports game visits up 55 percent from last year (company-reported).
Two figures, two very different growth rates, and the gap between them is the most interesting thing in the announcement — more on that below.
A caution that applies to this whole section: a developer-conference recap is a marketing document. It selects which metrics to name and which to leave out. "Users who came to play every day" is not defined in the post, and sources do not confirm whether it matches the daily active users figure the company reports to investors, or how either is counted.
Creator payouts: about $1.7 billion through DevEx
On the money side, the same recap reports approximately $1.7 billion earned by creators through the Developer Exchange programme, up about 50 percent since last year (company-reported).
If both the total and the growth rate are taken at face value, the implied prior-year figure is roughly $1.13 billion (derived by dividing $1.7 billion by 1.50; not a number Roblox published). The arithmetic is only as good as the rounding in "approximately" and "about" — a range of $1.65–1.75 billion at 45–55 percent growth puts last year anywhere between roughly $1.06 billion and $1.21 billion (derived).
Worth noting for anyone modelling their own share: DevEx growth of about 50 percent sits close to the 55 percent visit growth, and well above the 10 percent audience growth (comparison derived from the three reported figures).
Nearly 13,000 first-time DevEx qualifiers
Roblox also reports that nearly 13,000 creators qualified for DevEx for the first time this year, up 38 percent year over year (company-reported). That implies roughly 9,400 first-time qualifiers the year before (derived from 13,000 ÷ 1.38; not published by Roblox).
What the number does not say matters as much as what it does. Qualifying for DevEx is a threshold event, not an income statement. The recap does not break down what those first-time qualifiers earned, how many crossed the threshold once versus repeatedly, or what the median payout looks like — and sources do not confirm any of it. A rising count of qualifiers is consistent with a broadening middle class of creators; it is equally consistent with a very large number of people clearing a low bar once. The announcement does not distinguish between those two worlds.
The safety changes behind the numbers
Roblox frames a large part of the year around trust and safety. The recap states that the company introduced age checks in April and launched Roblox Select and Roblox Kids in May, and that together these reduced harmful content reaching its youngest users by 99.5 percent (company-reported).
This section carries the single heaviest claim in the article, so it gets its own reservation. A 99.5 percent reduction is a figure with no published methodology attached in the source material: sources do not confirm what counts as "harmful content", what the measurement baseline was, whether detection changed alongside the policy, or who audited the result. Improvements in an internal moderation metric can reflect a genuine change in what children encounter, or a change in what the classifier flags, and nothing in the announcement separates the two. The claim should be treated as the company’s own account of its own safety systems until an outside party publishes a check.
For developers, the practical consequence is structural rather than statistical: age checks and the two new curated audiences — Select and Kids — change which players a given experience can legally and technically reach.
Visits grew more than five times faster than the audience did
Here is the ratio the recap does not state. Taking Roblox’s two reported figures at face value, visits grew 55 percent while the daily audience grew 10 percent — a factor of 5.5 (derived from Roblox’s own reported figures; not a number the company published).
A second derived figure follows from the same two numbers: visits per daily player rose roughly 41 percent (1.55 ÷ 1.10 = 1.409; derived). Both derivations assume the two figures cover the same period and the same population, which the recap does not explicitly state.
If that holds, the platform’s growth last quarter came far more from existing players launching more sessions than from new players arriving. For a creator, those are different businesses. Growth in audience means a larger pool of people who have never seen an experience — a discovery problem. Growth in visits per player means the same people are opening more sessions — a retention and re-entry problem, where session length, return hooks, and update cadence do the work.
One limit on the interpretation: a "visit" is a session start, not a minute played. Nothing in the recap ties visit growth to engagement time, and a rise in visits is compatible with shorter sessions. An earlier piece on this site examining what Roblox build and run records actually prove about play makes the same general point in a smaller setting: a recorded event is not a played session.
A network simulator arrives in Studio beta
The most immediately usable tool in the announcement set is a testing one. Roblox’s developer forum post describes a rebuilt Device Simulator organised around a new toolbar, alongside a Network Simulator that simulates latency, jitter, and packet loss during playtests, both in Studio beta (announced by Roblox; feature behaviour not independently tested here).
If it works as described, this closes a real gap. Roblox experiences are played on a wide spread of hardware and connections, and bugs that only appear at 200 ms with intermittent loss are notoriously hard to reproduce on a developer’s own wired machine. Simulated impairment is a standard tool in other engines; having it inside the editor rather than as an OS-level shim would lower the cost of running it routinely.
What the post does not establish: how faithfully the simulation models real mobile networks, whether it affects server-side replication or only the local client view, and whether beta behaviour will survive to release. Those are open questions, not criticisms.
Publishing gets cheaper and faster — with details unstated
Roblox published a separate post on simplifying publishing, describing a faster path to reach 16+ audiences, pre-launch feedback from friends, and a lower cost to reach Roblox Kids and Select (announced by Roblox; specifics not independently verified).
Note the phrasing in the source material is directional — "faster", "lower cost" — without published timelines or price points. A shorter review path to 16+ and a cheaper route into the curated child audiences would both matter commercially, because they change how long capital sits idle between finishing a build and reaching players. But "faster than what, by how much" is not answered, and treating an unquantified improvement as a planning input is how schedules slip.
The pre-launch friends feedback mechanism is the piece most likely to change day-to-day practice: it moves a build in front of real players before the public launch commit, which is a different signal from internal playtests.
Licensing per item instead of per game
The licensing change is the structural one. Roblox announced in-game sales licensing, which lets creators sell game passes and developer products built with licensed intellectual property, with revenue share applied only to those specific items rather than to an entire game (announced by Roblox; terms not independently verified). The same post announces three new Kodansha franchises available to license through the Creator Hub (announced by Roblox; the specific titles are not enumerated here because the post’s list has not been independently confirmed).
Per-item licensing is a meaningfully different economic shape from per-game licensing. Under a whole-game licence, every unit of revenue in the experience carries the rights holder’s share, which pushes developers toward either fully licensed games or fully unlicensed ones. Under per-item licensing, a mostly original game can carry a licensed cosmetic or pass without the licence reaching the rest of the catalogue. That should, in principle, widen the set of games where a licence is worth doing at all.
The rates are the part that decides it, and the announcement does not publish them in a form that can be checked here.
"Dozens" of announcements — what a headline count tells a developer
Roblox’s recap characterises the conference output as dozens of new creator tools across publishing, payments, engine, and avatar (company’s own framing; the individual items have not been enumerated or verified here).
A count is not a roadmap. Announcement volume at a developer conference mixes shipped features, betas, and stated intentions, and those three carry completely different planning risk. Of the items covered above, the network and device simulators are described as Studio beta — testable now, subject to change. The publishing improvements are largely forward-looking. The licensing change reads as available, but its commercial terms are unpublished here.
The useful filter is not "how many things were announced" but "how many can be opened in Studio this week". On the evidence in these posts, that set is small and concentrated in testing tools.
The same season elsewhere: Unreal 5.8 samples and a unified Unity CLI
Roblox was not alone this cycle. Epic published posts describing an update to the City Sample project for Unreal Engine 5.8 and an update to the Game Animation Sample Project for the same version. The contents attributed to those updates — new levels and procedural worldbuilding toolsets in the City Sample; physics, motion matching, pose searching, and look-at features in the animation sample — are described in Epic’s posts and have not been independently verified or opened and tested here; treat the feature lists as claimed rather than confirmed. A previous article on this site covering how developers treat the Game Animation Sample as a starting template goes into what that sample is generally used for.
On the Unity side, the sourcing is cleaner because the claim is the company’s own description of its own product. Unity announced a new command-line interface providing a unified way to manage Unity installations, run tests, and execute builds across CI providers, replacing custom scripts (Unity’s own announcement). Unattributed performance or reliability gains from it are not claimed here — only that Unity announced it and described that scope. For context on Unity in production use, see an earlier piece on how a studio rebuilt a classic in Unity 3D.
The common thread across all three platforms this season is unglamorous: testing infrastructure, build automation, and reference content. Not one of the headline items is a rendering or gameplay capability.
Developer takeaway, and the question the announcements cannot answer
Three things are actionable now, assuming the announcements describe what ships:
- Test under degraded network conditions before the next release. If the Network Simulator behaves as Roblox describes, a latency-and-loss pass becomes a cheap routine step rather than a special project.
- Plan for retention, not just discovery. Visits growing about 5.5 times faster than the daily audience (derived from Roblox’s two reported figures) points to existing players returning more often as the dominant driver — if the two figures are comparable.
- Re-examine licensing deals that were uneconomic as whole-game licences. Per-item revenue share changes the break-even, though the actual rates are what decide it and they are not published in the source post.
The question none of this answers: whether more visits means more revenue for the individual developer. Roblox published aggregate DevEx growth of about 50 percent and a rising count of first-time qualifiers, but no distribution — no median, no breakdown by game size, no indication whether the additional visits concentrated in the top experiences or spread across the catalogue. Platform-level growth and per-developer outcomes can diverge sharply, and the recap does not contain the data to tell them apart.
That is the figure worth watching for in the next quarterly disclosure, and the one a conference recap was never going to provide.