diff --git a/Cargo.toml b/Cargo.toml
index 156bb23..2ccbd60 100644
--- a/Cargo.toml
+++ b/Cargo.toml
@@ -1,7 +1,7 @@
 [package]
 name = "slowcat"
 version = "1.0.0"
-edition = "2021"
+edition = "2024"
 
 [dependencies]
 
diff --git a/src/main.rs b/src/main.rs
index 8f569c4..fe89a12 100644
--- a/src/main.rs
+++ b/src/main.rs
@@ -1,20 +1,70 @@
 use std::{
     env,
-    io::{stdin, stdout, Read, StdoutLock, Write},
-    time::Duration,
+    io::{Read, StdoutLock, Write, stdin, stdout},
+    time::{Duration, Instant},
 };
 
 static DEFAULT_HZ: u32 = 100;
 
 fn output_slow(data: &[u8], hz: u32, out: &mut StdoutLock) {
-    // write 1 byte at a time
-    data.iter().for_each(|&b| {
-        out.write_all(&[b]).unwrap();
-        out.flush().unwrap();
+    let interval = Duration::from_secs_f64(1.0 / hz as f64);
 
-        // sleep according to given frequency
-        std::thread::sleep(Duration::from_secs_f64(1.0 / hz as f64));
-    });
+    // For very high frequencies, use adaptive batching
+    if hz > 1000 {
+        adaptive_batch_output(data, interval, out);
+    } else {
+        // Original behavior for lower frequencies
+        let mut next_time = Instant::now();
+        for &b in data {
+            out.write_all(&[b]).unwrap();
+            out.flush().unwrap();
+
+            next_time += interval;
+            let now = Instant::now();
+            if next_time > now {
+                std::thread::sleep(next_time - now);
+            }
+        }
+    }
+}
+
+fn adaptive_batch_output(data: &[u8], interval: Duration, out: &mut StdoutLock) {
+    let frame_duration = Duration::from_millis(20); // 20ms target window
+    let mut batch_size = 1usize;
+    let mut i = 0;
+    let mut last_batch_start = Instant::now();
+    
+    while i < data.len() {
+        let batch_start = Instant::now();
+        let end = (i + batch_size).min(data.len());
+        let actual_bytes = end - i;
+        
+        // Output the batch
+        out.write_all(&data[i..end]).unwrap();
+        out.flush().unwrap();
+        
+        let batch_duration = batch_start.elapsed();
+        i = end;
+        
+        // Calculate how many bytes we should output per 20ms window
+        let target_bytes_per_frame = (frame_duration.as_secs_f64() / interval.as_secs_f64()) as usize;
+        
+        // Calculate optimal batch size based on measured throughput
+        if batch_duration.as_secs_f64() > 0.0 {
+            let measured_bytes_per_sec = actual_bytes as f64 / batch_duration.as_secs_f64();
+            let optimal_batch_size = (measured_bytes_per_sec * frame_duration.as_secs_f64()) as usize;
+            batch_size = optimal_batch_size.max(1).min(target_bytes_per_frame);
+        }
+        
+        // Sleep for the expected time for this batch
+        let expected_sleep = interval * actual_bytes as u32;
+        let elapsed_since_batch_start = last_batch_start.elapsed();
+        if expected_sleep > elapsed_since_batch_start {
+            std::thread::sleep(expected_sleep - elapsed_since_batch_start);
+        }
+        
+        last_batch_start = batch_start;
+    }
 }
 
 fn main() {