November 11 2026

We're removing chat hibernation

We are removing mandatory chat hibernation from dogAdvisor Max starting tomorrow, allowing you to resume older conversations on Max with limited to no disruption to your experience. Furthermore, we're making a second change to how conversations work in dogAdvisor Max with updates to the number of available messages in one conversation. We are also announcing changes to Generation 4 usage. Together, these updates are designed to help Max maintain more useful context while making conversations more broadly available.

Removing hibernation and introducing automatic titles

Previously, conversations older than three days would be hibernated automatically by dogAdvisor Max, allowing you to review and read conversations but preventing continued discussion. Hibernation was introduced to address a problem with conversational context where Max's responses to a follow-up required processing enormous amounts of earlier conversational contexts in chats. As a conversation went on for longer, more history had to be carried forward making it more difficult for Max to manage, introducing a significant risk of misunderstanding or degraded responses. Previously, chats older than three days would become read-only rather than continue with context that may no longer be available to Max in a useful form to protect the quality of the conversation, though it did mean owners couldn't return to useful older conversations to pick up directly where they left off.

Following feedback from those using Max Series 5 (our latest, soon-to-be-released model), we are making updates to this process by removing hibernation. Now, when owners return to an older conversation and send a new message, Max condenses earlier discussions into shorter summaries that can be used together with new conversational context to continue older conversations. This gives Max a way to carry forward important or helpful background information without relying on the conversational history in its entire form every time. For example, a summary may preserve the main questions asked and relevant details shared whilst disregarding irrelevant context so owners can follow-up without Max reviewing the entire situation again. We are cautious that summaries are likely to lose context or detail, and we will continue to test and improve our summarisation models to better control this risk. For now, we'll notify users that they may see temporary context loss when resuming an older conversation. If a follow-up question is relevant to a specific part of a question, we are encouraging Preregistration users to use the follow-up capability. We will continue to roll this out to more test users to examine performance and usage before we make this capability available to Max Series 5 on release.

Excitingly, we are also bringing conversational compaction and summaries to owner messages, meaning Max can now better understand the central topics discussed in a conversation, which means we are now able to offer automatic titles to your conversations in Chats. So the next time you ask about dog food, Max can detect this as a central topic thanks to compaction and call your chat "Discovering the best types of dog food" to help you find it easier in chats, rather than relying on remembering the date of your conversation as before.

We will begin pushing these updates tomorrow morning, and all owners using Max Series 5 in Preregistration will see these updates by November 13th.

Restructuring conversations

We are also reducing the number of messages in one Ask conversation from 20 to 14, effective tomorrow morning. We are not making any changes to monthly usage ceilings for any users. The new message limit for Series 5 remains nearly 30% higher than Generation 4 (prior to our recent reduction), and is a significant improvement for the majority of owners who are currently using Generation 4. This change helps keep each individual conversation more focused with more manageable context, while supporting broader access of Max.

As Max carries huge amounts of context from previous messages in a single conversation, our testing saw increased degradation in response accuracy and context management around the 15th message, especially across complex health and care questions, complex medical questions, or questions which invoked Advanced or Ultra reasoning for Series 5 Preregistration. Our statistical analysis of usage patterns for Preregistration users has also shown the majority of owners (closer to 80%) never hit the 14 message mark in a single conversation.

We are aware that this change may disrupt the experience of some Preregistration users, particularly when asking multiple follow-up questions or clarifications over Max's responses. We are therefore improving our conversational ceiling alerts now notifying owners on their 10th-13th message as they approach limits in one conversation. For a limited time, we are also introducing compacting of conversations so owners can resume chats with the same summarisation models discussed above, but now summarising all 14 messages for a new conversation. We will continue to look closely at adoption and usage of this conversational compacting, and will either extend or sunset this capability subject to statistical usage pattern analysis outcomes. These changes are intended to support a more continuous experience without requiring every older message is carried forward in full.

Usage adjustments for Generation 4

Starting Friday, we’re making temporary changes to usage limits for Generation 4 owners. Overall usage limits will be reduced by 50%, and the maximum number of messages in one conversation will be reduced to 40% of current levels. Please note that a previous announcement made by us described message limits in conversation as reduced by 40%, which was a minor factual inaccuracy as limits have now been reduced by 60%. We apologise for this confusion. These changes apply across Generation 4 conversations and users, and will remain in place until December 31.

This means Generation 4 owners will have less usage available during this period, and each individual conversation will end sooner. The overall usage reduction affects how much owners can use Max across conversations. We understand that this is a substantial reduction, and owners may reach their overall usage limit sooner than they do today. Owners can start another conversation within their available usage, but doing so may require restating information from the earlier chat. We recommend including the details Max needs to understand a new question, especially when continuing a topic about a dog’s health. Compute (or the processing capacity needed to sustain Max) can vary dramatically over time and according to how many people use Max at one time. The adjustments we are making today give us a longer period to assess impacts on usage while maintaining access for Generation 4 owners. Additionally, as we allocate substantial compute to Series 5 Preregistration testing, we have been forced to make updates to reduce usage limits by 50% for Generation 4.

We are also responding to activity that we believe may involve attempts to distil our models or access them through automated systems. Model distillation can involve using one model’s outputs to develop or improve another model, which is a serious breach of our Platform Services Agreement. We will share more on the update of this review and the steps we are taking in Series 5 to prevent such attacks soon. but we are temporarily removing reasoning explanations/trails from Generation 4, and will assess the benefit of this capability externally to dog owners in the coming weeks. We are taking these possibilities into account as we assess Generation 4 compute demand and the conditions needed to keep the service available. The temporary limits give us a longer period to observe how changes affect usage and compute demand while maintaining access to Max for Generation 4 owners. We know that reducing usage will affect owners’ experience, and we will use the time through December 31 to evaluate the impact of these changes alongside the broader question of continued Generation 4 access. You can always reach out to someone on the team if you have any additional questions or have been severely impacted by these reductions.

If we continue to extend Generation 4 access into 2027, we will restore existing usage limits and conversation message limits to their current levels on January 1. We’ll share further information about that decision, and the timeline for Series 5 rollout, as it becomes available.

Further updates on our progress with Series 5

Together with the updates we've shared above we are also sharing more updates on new capabilities we're bringing to Preregistration. They take effect this week, on November 15th.

First, we are permanently removing usage limits for Max Voice. Previously, you'd use up both your Ask message usage limit and voice limit together when using Max Voice. Following feedback from owners and additional technical work by dogAdvisor we're excited to announce we are now permanently removing usage limits for Max Voice, so your usage is now only dictated by Chat limits. This means the maximum 60 minute Voice usage is being scrapped, so your usage of Max Voice is unlimited but use of Max itself is subject to standard usage limits. We are also tripling the limit on stored conversations in Chats, and we are permanently removing usage limits for data exports from Health. We will no longer offer export controls for users switching away from Max. These updates mean we now have four usage limits: usage for preferred reasoning, usage for advanced reasoning, usage for ultra reasoning, and chat limits. This helps us simplify our usage limits on all accounts to make it easier to manage your usage with Max. We are also improving and simplifying Max Voice and introducing further customisation over voice speed. For MaxID, we are removing the 90 day limit for chat authentication, and we will now support all chats sent by Series 5 for authentication with MaxID.

We are also introducing significant changes to Chats. With the help of compaction and summarisation announced today, we are now introducing Automatic Grouping to chats, so your new chats with Max are now automatically categorised into five folders: lifestyle, care, health, training, and other. We are also improving search with in-line and in-conversation capabilities, expanded correction, and more powerful semantic search. We are also introducing the ability to manually move chats to one of these five folders, and we will look to introduce custom folders based on feedback from Preregistration. We are also introducing Chat Lock which now allows you to lock private chats behind your account authentication. If you locked a chat but don't remember the password we are offering resets subject to account verification. We are also extending chat deletion to all chats. Furthermore, we are now expanding the number of chats available on 'Save' from three to thirty, a 10x increase. Additionally, we are introducing more controls over sharing allowing you to unshare a chat with Transporter, so recipients no longer have access to saved chats if you made a mistake.

We are also extending advanced trials of several capabilities to everyone in Preregistration starting November 15th. First, we are extending advanced Voice to everyone and we're now introducing voice compaction (permanently removing the 60 second message timer and voice message length limits in favour of dramatically more accommodating time limits) to summarise longer voice messages removing ems and aahhs from speech. The advanced Voice capability now supports live interruptions in beta, so owners can now interrupt with voice as Max is responding. We are also introducing live microphone and output audio selectors inside of voice so users can troubleshoot local issues themselves, and we're finally bringing (upon much request) reasoning customisation options directly into voice. We are also introducing advanced isolation for voice, so you can now speak to Max in more noisy environments. Finally, we quashed all the bugs in Ask and Chats you told us about!

On Friday November 13th we will also introduce our new Settings/Accounts page to everyone on Preregistration which has been designed to be easier to navigate than our previous version. We are also dramatically improving how you're notified of conversational compaction in new chats, redesigning the facts interface design in Ask, easier to access usage limit notifications (without having to navigate to settings), and improved follow-ups.

dogAdvisor is super pleased to announce that we expect to finalise our work on Accounts, Settings, and Health by November 17th and move fully to the final stages of model safety testing and deployment as soon as possible. Our roadmap sets Series 5 release timelines around the end of December - start of January in 2027 and we're excited to share more about our progress as we get closer to the new year. Once we publish Series 5's model card and launch it to the public we will begin sunsetting the previous generation of Max.