diff --git a/crates/ml/src/cuda_pipeline/gpu_dqn_trainer.rs b/crates/ml/src/cuda_pipeline/gpu_dqn_trainer.rs
index 3b52fdb41..0eda19606 100644
--- a/crates/ml/src/cuda_pipeline/gpu_dqn_trainer.rs
+++ b/crates/ml/src/cuda_pipeline/gpu_dqn_trainer.rs
@@ -2992,7 +2992,7 @@ impl GpuDqnTrainer {
     pub fn replay_adam_and_readback(&mut self) -> Result<FusedTrainScalars, MLError> {
         static ADAM_CTR: std::sync::atomic::AtomicUsize = std::sync::atomic::AtomicUsize::new(0);
         let adam_step = ADAM_CTR.fetch_add(1, std::sync::atomic::Ordering::Relaxed);
-        if adam_step < 3 { tracing::info!(adam_step, "adam_readback: enter"); }
+        if adam_step < 3 { eprintln!("adam_readback: step {adam_step} — enter"); }
 
         // 1. Collect previous step's scalars (if pending)
         let prev_scalars = if self.readback_pending {
@@ -3016,11 +3016,11 @@ impl GpuDqnTrainer {
         };
 
         // 2. Launch current step: grad_norm + adam
-        if adam_step < 3 { tracing::info!(adam_step, "adam_readback: compute_grad_norm"); }
+        if adam_step < 3 { eprintln!("adam_readback: step {adam_step} — compute_grad_norm"); }
         self.compute_grad_norm_outside_graph()?;
-        if adam_step < 3 { tracing::info!(adam_step, "adam_readback: replay_adam"); }
+        if adam_step < 3 { eprintln!("adam_readback: step {adam_step} — replay_adam"); }
         self.replay_adam()?;
-        if adam_step < 3 { tracing::info!(adam_step, "adam_readback: async readback"); }
+        if adam_step < 3 { eprintln!("adam_readback: step {adam_step} — async readback start"); }
 
         // 3. Async DtoH for current step's scalars (non-blocking)
         unsafe {
@@ -3047,6 +3047,8 @@ impl GpuDqnTrainer {
             );
         }
 
+        if adam_step < 3 { eprintln!("adam_readback: step {adam_step} — async readback done, recording event"); }
+
         // 4. Record event — marks the point after the async copies
         let event = self.stream.record_event(None).map_err(|e|
             MLError::ModelError(format!("readback event record: {e}"))
@@ -3054,6 +3056,8 @@ impl GpuDqnTrainer {
         self.readback_event = Some(event);
         self.readback_pending = true;
 
+        if adam_step < 3 { eprintln!("adam_readback: step {adam_step} — returning"); }
+
         // Return previous step's values
         self.scalars_readback_host = [prev_scalars.total_loss, prev_scalars.grad_norm];
         Ok(prev_scalars)