What Shipped This Week: Workspace Usage, Paid-Only Media, and a Shared Projects Fix

The biggest change this week is one you might never notice if everything works: when you reach a monthly limit, Writingmate no longer swaps your chat to a different model without telling you. We also made your usage page show the whole workspace, put image and video generation behind a paid plan, and fixed shared projects that returned a 404 for everyone except their creator.

It was a busy week, and a few things moved back and forth while we worked out what the right behavior was. We say so below where it matters.

What shipped

Usage that covers the whole workspace

Profile → Usage used to answer one question: what did I use? That is the wrong question when several people share a workspace. You can now see messages, tokens and provider usage for the workspace you have selected, added up across its members.

We built this because admins kept asking where a shared allowance went. Now the answer sits in one place instead of requiring each person to read out their own numbers.

To try it, open your profile, go to Usage, and switch workspaces if you belong to more than one. The usage numbers also line up with your subscription period now, so an annual plan keeps its monthly allowance windows on the right anniversary instead of drifting.

One honest note: we briefly showed more internal detail on that page, then pulled it back. We removed internal provider-dollar breakdowns from the customer view and kept the parts that matter to you, such as purchased credits and what you have left. Expect that page to keep getting simpler.

Your model stays the model you picked

Earlier in the week we tried switching chats to a free model automatically once a monthly spend cap was reached. We rolled that back within hours. A silent swap changes the answers you get without you knowing, and that is worse than a clear stop.

Here is how it works now. If a paid model reaches your monthly limit, you get a 'Monthly Limit Reached' message. Free models stay available, and you choose to switch to one. We also removed the remaining places where background features, such as model comparison, message compaction and the task tool, quietly substituted a free model. If we cannot verify access or budget, the operation stops instead of changing models.

We also fixed two cases where the limit was too strict. Previously, each paid text request reserved a worst-case estimate before it ran, which on large-context models could be several dollars for a message that really cost about a cent. That blocked some paying Ultimate customers with most of their month unused. Now a request is allowed while your actual recorded spend is under the cap, its real cost is recorded afterwards, and one request may go slightly over so your last message of the month is not refused. The cap for a workspace is now what the workspace actually pays, not a smaller fraction of it. Annual plans use the same monthly cap as their monthly counterparts.

You can see this in chat: if you are at the limit, the message names the limit instead of a generic error.

Image and video need a paid plan

Image and video generation now requires a paid workspace. Free and canceled workspaces had been generating large volumes of media on our provider keys, and nothing stopped them. We closed that gap, including the edge case where a model with a zero price could reopen it.

If you are on a free plan and try to generate, you see a 'Paid plan required' message with an Upgrade button, not a bare error. If we cannot check your plan at that moment, you get a retryable error instead of a billing message. Chat, speech and the other features are unchanged, and requests that use your own provider key are not affected.

AppSumo video pools changed too. Pools are still counted in baseline seconds, so nobody's allowance shrank, but a second of an expensive model like Seedance 2.5 now uses more of the pool than a second of PixVerse. If a generation is refused, the pool slot is returned, so retries do not burn through your allowance. Try it at /images and /video.

Also worth knowing: Sora 2 and Sora 2 Pro are retired. We published a separate note on what that means for saved videos.

A clearer homepage and pricing

Our homepage used to sell savings. It now leads with what people tend to stay for: getting the newest Claude or ChatGPT model within days of launch, in one account, and writing that does not sound like ChatGPT. The 'newest AIs' list is ordered by release date with one model per well-known lab, so it keeps up with launches.

We removed video from the plan cards, the comparison tables, the homepage price comparison and the FAQ, and dropped old-plan notes from the FAQ since new buyers only see credit plans. The top menu is now AIs, Features, Pricing and Blog. See it at /pricing.

Fixed this week

• Workspace members can open projects created by other members. Before, the sidebar listed them but the page returned a 404. Only the creator can still edit or delete a project.
• People you invite to a workspace no longer receive repeated invitation emails.
• A voice clip with no recognizable speech now returns an empty transcript instead of an error and a retry loop.
• Gemini 3.8 Flash is back in the model picker, and text-to-speech variants no longer show up as chat models.
• Duplicate usage records no longer cause database errors behind the scenes.
• Free tools used without signing in keep working on a fixed free model.

This week was rougher than we like, with a few reversals on how limits behave. We would rather tell you that than pretend it was a clean line. Thank you to everyone who wrote to support with specifics, because several of these fixes came directly from those emails. You can browse every past update on the changelog. See you next Friday, Artem