ASCN.ai Update: AI Agents Are Now More Stable and Better at Working With Images

We received complaints about the stability of ASCN.ai. An AI agent could freeze, stop responding or lose an active task when the connection was interrupted or platform services were updated.

We have fixed these issues and released a major platform update.

Active tasks can now survive temporary failures and service restarts. The agent continues working from where it stopped instead of losing the entire process. Simply put, the agent no longer “dies” in the middle of a task because of a temporary connection problem.

More Stable AI Agent Performance

If the connection is temporarily interrupted or an ASCN.ai service is restarted, the agent preserves its progress and continues the task.

We have also improved the handling of long responses and fixed several causes of interrupted or frozen messages. This is particularly important for complex workflows, file creation, image generation and other time-consuming tasks.

Improved Image Understanding and Editing

The ASCN AI agent now understands images attached to a conversation more accurately.

It can analyze and edit uploaded images, as well as continue working with images it has generated itself. Updated instructions also help the agent understand visual content and choose the correct actions without unnecessary steps.

Background Tasks and Content Generation

Long-running tasks can now be completed in the background. For example, generating a large image or video no longer needs to block the entire conversation.

The agent starts the process, continues working asynchronously and returns with the result when it is ready. This architecture also creates a foundation for more advanced automation, orchestration and collaboration between multiple AI agents.

More Accurate AI Memory

We have improved how the agent processes and stores information in its memory:

• Contradictory facts are excluded.
• Similar but different facts remain separate.
• The same information is no longer stored multiple times in different wording.
• The agent can better distinguish its identity from information remembered about the user.

These changes make long-term interactions with an AI agent more consistent and reliable.

Improved Integrations and Telegram Experience

The agent can now find the correct applications and connect external services more efficiently.

We also fixed the integration buttons in Telegram. Instead of receiving broken links, users should now receive a working connection button.

Another fix prevents parts of an agent’s response from being duplicated while the message is being streamed in Telegram.

More Accurate Generation Costs

The cost of generating images and videos is now calculated according to the selected AI model and generation parameters.

We have also improved billing for cached parts of prompts. This makes credit usage more accurate and can reduce costs in scenarios where part of a request can be reused.

ASCN.ai Is More Reliable for Real-World Tasks

This update improves AI agent stability, image processing, memory, integrations and the execution of long-running workflows.

We will continue developing ASCN.ai as a platform where AI agents do more than answer questions. They can work with files, create content, connect to external services, run tasks in the background and return with completed results.

👉 Try the updated ASCN.ai platform