I've found it very difficult to work with multiple blocks/tasks using AI, and even moreso with custom prompts.

The issues I've mainly faced are these:

* When multiple blocks (e.g. tasks) are selected, the only AI interaction available is AI Assistant. But that doesn't support custom prompts, so only the inbuilt AI functions are available. (Completed: Custom Commands for AI Agents)

* Using dark theme, mouse-highlighted block selection colour is difficult to see (not enough contrast), making it hard to tell what blocks have been selected.

* When output is generated by AI Chat, using the Copy function copies in backticks that aren't visible. This means visibly ok output such as a newline-separated series of URLs must be highlighted then Cmd+C/Ctrl+C copied, to exclude the backticks. For me this makes the Copy button there unusable since it's harder not easy to capture the visible output.

* Running background agents has to be done on one block at a time as far as I can tell - it would be nice to be able to multi-select blocks then invoke the same agent on all of them. That would really open up the power of background agents, from my perspective, as would the ability to "live filter" a project then do something with that filtered selection. (Completed: Bulk Select and Assign to AI Agents from Toolbar)

* I've made various attempts to ask Taskade AI to work on multiple blocks of my choosing - for example by copy-pasting the block link to AI chat then asking it to summarise all tasks "under that block" or "nested in that block". Whether I use the full pasted link or just the node id at the end, the AI cannot work with such input. Similarly, pressing the plus on a parent node and invoking either AI Agent or AI Assistant then asking it to e.g. "summarise tasks here for pasting into Things 3", it has no awareness of the tasks in/under that parent block, and instead generates the output based on tasks it generates itself. (Completed: AI Agents use subnodes/subtasks as context when executing commands)

* In a similar vein, Taskade AI seems to have poor context-aware intelligence. For example, I create a short & simple Project with a block somewhere in the middle titled "✅ Today's Tasks", and a task list nested under thta. When I ask AI Chat to "summarise all Today's Tasks", it isn't able to identify what I'm referencing. (Completed: This should be resolved by chatting with AI Agents from the sidebar in Taskade)

You can hopefully see how many ways I've tried to apply AI to a chosen series of blocks and the difficulties the app currently produces in doing so - in fact it's almost impossible from what I can tell, or the time investment is far too great to make it usable.

Currently the lack of multi-select custom AI prompting and difficulty in otherwise referencing multiple blocks or children of a block or running multiple instances of a custom prompt swiftly is, for me, making the app more of a fancy todo list with promising AI developments that don't yet live up to that promise. In a separate issue yesterday I've reported how, the first time I've tried to use the Background Agents - which you've been actively promoting - I encountered what seems to be two bugs that make those currently unusable, for me at least.

I'm a paid subscriber who's a software expert, so will happily elaborate on this and provide more feedback in the future as well. I really like what you've made so far, hence subscribing, but in practical terms I don't feel confident to fully adopt/switch to the app due to these difficulties in generating useful AI output for multiple blocks or part(s) of a project.