diff --git a/src/invidious/jobs.cr b/src/invidious/jobs.cr
index b6b673f7..57d62816 100644
--- a/src/invidious/jobs.cr
+++ b/src/invidious/jobs.cr
@@ -1,40 +1,62 @@
 module Invidious::Jobs
+  # This line defines an empty array named JOBS to hold objects of type BaseJob.
   JOBS = [] of BaseJob
 
-  # Automatically generate a structure that wraps the various
-  # jobs' configs, so that the following YAML config can be used:
-  #
-  # jobs:
-  #   job_name:
-  #     enabled: true
-  #     some_property: "value"
-  #
-  macro finished
-    struct JobsConfig
-      include YAML::Serializable
+  # SEMAPHORE is a Channel that allows up to 5 items (jobs) to be processed concurrently.
+  # This is a way to limit the maximum number of jobs running at the same time to 5.
+  SEMAPHORE = ::Channel(Nil).new(5)
 
+  # The `macro finished` block is executed once the module definition is complete.
+  macro finished
+    # Define a struct named JobsConfig inside the module. 
+    # This struct is for storing configuration for different jobs.
+    struct JobsConfig
+      include YAML::Serializable  # Allows serialization and deserialization from YAML.
+
+      # This loop iterates over all subclasses of BaseJob.
       {% for sc in BaseJob.subclasses %}
-        # Voodoo macro to transform `Some::Module::CustomJob` to `custom`
+        # Generate a getter method for each job. The job's class name is transformed
+        # from something like `Some::Module::CustomJob` to a simpler form `custom`.
         {% class_name = sc.id.split("::").last.id.gsub(/Job$/, "").underscore %}
 
         getter {{ class_name }} = {{ sc.name }}::Config.new
       {% end %}
 
+      # Define an empty initializer.
       def initialize
       end
     end
   end
 
+  # This class method allows registration of a job to the JOBS array.
   def self.register(job : BaseJob)
     JOBS << job
   end
 
+  # This class method starts all registered jobs.
   def self.start_all
+    # Iterate over each job in the JOBS array.
     JOBS.each do |job|
-      # Don't run the main rountine if the job is disabled by config
-      next if job.disabled?
+      # Send a nil value to the SEMAPHORE channel. This is like acquiring a "slot".
+      # If the SEMAPHORE is full (5 jobs running), this line will block until a slot is free.
+      SEMAPHORE.send(nil)
 
-      spawn { job.begin }
+      # Start a new concurrent fiber (lightweight thread) for the job.
+      spawn do
+        begin
+          # If the job is disabled in its configuration, skip the execution.
+          next if job.disabled?
+
+          # Start the job by calling its 'begin' method.
+          job.begin
+        rescue ex
+          # If an exception occurs, log the error.
+          Log.error { "Job failed: #{ex.message}" }
+        ensure
+          # After the job is finished or if an error occurred, release the "slot" in the semaphore.
+          SEMAPHORE.receive
+        end
+      end
     end
   end
 end