In times like this, it's good to know that what protestors are doing on the ground is having an effect on ICE. In particular:

The Twin Cities Model: Don’t Copy It, Learn from It

What sets apart the Twin Cities rapid response network and its surrounding ecosystem is not strict adherence to a particular structure. It is a clear analysis of their conditions, a willingness to adapt, and the courage to fight back as the violence increases.

I think this article is also useful - setting off car alarms, whistling, filming, and yelling at ICE does have an effect, especially when protesters outnumber ICE. In particular:

In Minneapolis, we’re already seeing the success of ICE Watch. There are countless videos from ICE Watch observers that show agents leaving without detaining anyone after they encounter resistance. There are currently about 3,000 ICE agents in Minnesota, but the Department of Homeland Security claims they have only arrested 2,400 people — less than one person per agent.

other discussions
 

Each chunk of the city (Southside, Uptown, Whittier, and so on) has rotating shifts of dispatchers, who admin a running Signal call throughout operational hours. Sometimes, multiple dispatchers overlap to split up the extra tasks of watching the chat, relaying reports to other channels, and checking license plates. Dispatch also helps people evenly distribute patrols across an area, takes notes, and assists people through confrontations. All patrollers in cars and on foot and stay on the call throughout their patrol. There is a constant flow of information, allowing other cars to decide whether they are well-positioned to join in, take over tailing the car, or continue searching for additional vehicles.

Since the structure has divided up into more granular neighborhood-based zones, people in many areas have also developed a daily chat system, with chats that are re-made and deleted each day to keep them clear and not maxed out of participants (as the maximum number of members of a Signal group is capped at 1000). Various areas of the cities and the suburbs have replicated the basic structure of this system but with slightly different models, chat structures, vetting systems, and data collection.

A data collection team collects anonymized data submitted from Whipple Watch and many of the local rapid response chats, aggregating them into consumable formats, such as interactive maps of hotspots. This team also admins the searchable database of license plates sorted by “confirmed ICE,” “suspected ICE,” “confirmed not ICE,” and other categories.

Additional place-based chats have emerged around school systems, faith communities, mutual aid grocery deliveries, and the like. Another development was the Neighborhood Networks intake chat, which acts as a clearinghouse for incoming volunteers. New people from anywhere in the city—or anywhere in the state of Minnesota—can be added and oriented to a list of chat options, and admins will add them to the open chats or connect them to the vetting and training processes for the more closed chats.

Most recently, dispatchers have experimented with a relay system in which patrollers who tail vehicles to the edge of their zone can communicate through dispatch across chats to pass off the vehicle to a patroller in the next region. This allows the patrollers to remain in tighter and tighter routes, which they can swiftly come to know intimately well in order to navigate them better than any ICE agents.

Finally, Spanish language relayers copy ICE alerts from dispatch calls and local chats, translate them, then send to large Spanish-language Signal and WhatsApp networks.