Two days ago, I wrote about Voqio's next chapter and why we were moving its best ideas into Swurl. At the time, Swurl was coming online quickly. Since then, it has taken another important step: it is no longer just a larger plan for what Voqio could become. It is a working product people can sign into and use.
We have spent the time since launch connecting the parts that make an AI workspace feel dependable instead of experimental. Some of that work is visible immediately. Some of it only becomes noticeable when a request takes longer than expected, a page is refreshed, a provider has a bad day, or someone wants to understand exactly what they paid for.
Image generation is now part of the workspace
The biggest visible addition is Image Studio. Swurl can now generate images through OpenAI, Google, xAI, Black Forest Labs' FLUX, and Stability AI from one place.
A person can choose one provider for a focused result or select several and compare their interpretations side by side. The images are saved securely to the workspace, can be reopened later, and can be downloaded or reported from the same interface. Each provider is priced and accounted for independently, so one failed generation does not erase the others or quietly charge for an image that was never delivered.
This is the same principle behind the rest of Swurl: the providers remain distinct. We are not trying to make every model look identical. We are making it easier to see what each one does well without managing five separate products.
Research now starts with the live web
Swurl's Search area has also grown into a more complete research workspace. Quick Research is designed for a focused overview. Deep Research explores a subject more thoroughly, tries multiple searches, removes duplicate sources, and gives stronger preference to current and authoritative material.
Research reports keep their sources attached. Titles, links, publication dates, and relevant excerpts remain available so a reader can inspect the evidence rather than receiving an answer that only sounds certain. Past research runs are saved, can be reopened after a refresh, and can be exported when the work needs to move somewhere else.
AI can still be wrong, including when it has sources. Swurl now says that plainly throughout the product and encourages people to independently verify anything important.
Conversations finally stay connected
Threads and messages now persist in the live workspace. People can begin a conversation with one model, return later, switch models for another opinion, and keep the resolved model attached to the history. Side-by-side comparisons and Roundtables use the same durable foundation, so a refresh no longer has to mean starting over.
Auto-Route is also live as a real routing system. It looks at the kind of work being requested, the context required, provider availability, the workspace's remaining credits, and the user's call limit. It then chooses the least expensive capable model before any provider is called. The selection and a short reason are shown in the response instead of happening invisibly.
Clearer accounts, credits, and control
We rebuilt the account and subscription experience around real workspace data. Users can manage their profile and subscription, buy one-time credits without being forced into a monthly plan, and see what remains in the workspace ledger. Administrators have a dedicated area for managing users, credits, subscriptions, and reported content.
Behind the interface, requests reserve their estimated credits before work begins and settle against the actual result afterward. Duplicate submissions, interrupted streams, cancelled work, timeouts, and partial multi-provider results all have explicit outcomes. That may not be the most exciting sentence in this update, but it is some of the most important work we have done.
A better product on smaller screens
We have continued refining the mobile workspace: full-width conversations, safer spacing around the keyboard, a composer that stays where it belongs, menus that fit the screen, and clearer access to Roundtable, Side-by-side, Search, Images, and the rest of the toolset.
The public demo and the live application are now intentionally separate. The demo uses sample data and simulated activity so someone can understand Swurl without creating an account or spending credits. The live workspace is where signed-in users do real work with real providers.
We have also prepared the safety and account-deletion flows needed for native distribution, including content reporting, an administrator review queue, a published response policy, clearer provider safety messages, and privacy disclosures. The iOS and Android work is still moving through developer-account and store requirements, but the product underneath those submissions is becoming much more complete.
What comes next
There is still plenty to improve. We will keep testing every provider, simplifying labels that make sense to builders but not yet to customers, and polishing the places where a powerful tool can still feel too complicated.
What has changed since the last post is confidence. Swurl now has persistent work, live research, multi-provider images, real billing controls, stronger recovery, and a much clearer path from the public site into the product. Those pieces are not isolated experiments anymore. They are beginning to feel like one connected workspace.
Voqio will remain available during the transition. When Swurl is ready to fully take its place, we will move Voqio's remaining traffic carefully with permanent redirects rather than leaving broken links behind.
Thank you for following the work, trying the product, and continuing to point out the rough edges. We are moving quickly, but we are building Swurl to last.
- Chris