Thank you @benwilson512 !

The code below worked:

 def init(_params) do   
   send(self(), :broadcast)
   {:ok, nil}
 end

def handle_info(:broadcast, state) do

Then looking at the Genserver documentation, handle_continue did the job too:

  @impl true
  def init(_params) do
    {:ok, nil, {:continue, "continue"}}
  end
  
  @impl true
  def handle_continue("continue", state) do

You call start in handle_event. Start is not in another process, it’s in the liveview process. init is in another process, but start waits on init to return before it returns, that’s part of the standard genserver contract see:

iex(4)> defmodule Foo do
...(4)>   use GenServer
...(4)>
...(4)>   def start() do
...(4)>     GenServer.start(__MODULE__, nil)
...(4)>   end
...(4)>
...(4)>
...(4)>   def init(_) do
...(4)>     Process.sleep(1000)
...(4)>     {:ok, nil}
...(4)>   end
...(4)> end
iex(7)> :timer.tc(fn -> Foo.start() end)
{1000836, {:ok, #PID<0.145.0>}}

EDIT: Further reading

https://github.com/erlang/otp/blob/master/lib/stdlib/src/gen_server.erl#L74-L91

Oh, I see, that makes total sense! Thanks for the well-rounded explanation @benwilson512. I completely forgot that the GenServer process everything linearly like you mentioned (because it’s basically just a loop waiting for its time to process something). This definitely is what’s going on here.