How to manage aggregates in DDD to not be too coupled?

I want to know a way (or ways) to not couple too much my aggregate, there's a concrete explication.

For example i have an order that can not have more than 100 items inside.

So i create an aggregate like

Order {
    List<Item> Items { get; private set; }

    private int _maxItemCapacity = 100;

    public Item AddItem(string itemName) {
        if (Items.Count + 1 > _maxItemCapacity) 
        {
            throw new Exception("You have reached the maximum limit of items");
        }
    var item = new Item(itemName);
    Items.Add(item);
    return item;
  }

    public Item GetItem(Guid itemId) {
        var item = Items.Where(item => item.Id == itemId).First();
        return item;
    }
}

Item {
    public Guid Id { get; private set; }

    public string Name { get; private set; }

    public Item(string name) {
        Id = Guid.NewGuid();
        Name = name;
    }
}

So now when i want to add some item to my Order i will pass through my invariant and that's perfect but...

If i want to get like only 1 of these items inside of my order by item Id, first i need to get my order with all he's items and after find my specific item.

And what if i had like 1000 items or more, i don't want to get all these items with my order, i want to pass through my validation (invariant) when i add some but without getting all items every time with the order.

How could i have an order with these validations without all items inside? (I don't want to have these validations in Application Layer).

Please don't forget that aggregates are optimized for writes, not for reads. So if you just need to read specific data you are not planning to change within that specific action it is totally fine to bypass the aggregate and read from your data storage in whatever way is most efficient for reading.

For instance, if just need to get the data of a specific order item to display it on your user Interface than just read the data from your data store.

If you, on the other hand also need to write to items inside your aggregate as a separate transaction where this aggregate is not needed to guarantee the consistency across all items, then you could think of splitting these items out as separate aggregates and reference those by id in your current aggregate.