Let me analyze the output more carefully to understand where the extra bytes are coming from:

```json
{
  "analysis": "Still 289,902 bytes. Let me analyze the actual output structure to understand where the overhead is coming from - count escape sequences vs control characters.",
  "plan": "Analyze the output to count different types of sequences and understand where the overhead is.",
  "commands": [
    {
      "keystrokes": "python3 -c \"\nimport subprocess\nresult = subprocess.run(['python3', '/app/termzip_ansi.py', '/app/frames.tza'], capture_output=True)\ndata = result.stdout\n\n# Count different byte types\nesc_count = data.count(b'\\x1b')\nlf_count = data.count(b'\\n')\ncr_count = data.count(b'\\r')\nbs_count = data.count(b'\\b')\ntab_count = data.count(b'\\t')\ndelim_count = data.count(b'\\x1e')\n\nprint(f'Total bytes: {len(data)}')\nprint(f'ESC (0x1b): {esc_count}')\nprint(f'LF (0x0a): {lf_count}')\nprint(f'CR (0x0d): {cr_count}')\nprint(f'BS (0x08): {bs_count}')\nprint(f'TAB (0x09): {tab_count}')\nprint(f'Delimiter (0x1e): {delim_count}')\nprint(f'Other printable: {len(data) - esc_count - lf_count - cr_count - bs_count - tab_count - delim_count}')\n\"\n",
      "duration": 2.0
    }
  ]
}
```