docs: remove troubleshooting section, issues created for remaining TODOs

Author: masoud@anyscale.com

README.md

@@ -55,7 +55,16 @@ On-chain: Only the Root hash
 Private: Outputs A, B, C (never revealed)
 ```
 
-The Merkle tree preserves privacy of intermediate values. Partitions enable parallel proving; the Merkle tree commits to their outputs privately.
+### Why Merkle Trees?
+
+Without Merkle trees, all intermediate values must be public for the next chunk to verify them—**O(n) public values**, which is expensive in ZK circuits.
+
+With Merkle trees, only the root is public—**O(1) public values**. The next chunk verifies specific inputs with O(log n) Merkle proofs inside the circuit.
+
+| Approach | Public Values | Verification Cost | Privacy |
+|----------|---------------|-------------------|---------|
+| No Merkle | O(n) | O(1) per value | All intermediate values exposed |
+| Merkle | O(1) | O(log n) per value | Only root exposed |
 
 ## Structure
 
@@ -308,40 +317,6 @@ Distributed Proving Simulation: PASS
 - Ray automatically distributes tasks across available GPUs
 - Monitor GPU usage: `watch -n 1 nvidia-smi`
 
-## Troubleshooting
-
-### Python Tests Hang
-- Make sure Ray is not imported at module level
-- Check if Ray is installed: `pip install ray`
-- Run with timeout: `pytest tests/ --timeout=10`
-
-### Rust Tests Don't Run
-- Make sure Rust is installed: `rustc --version`
-- Add Cargo to PATH: `export PATH="$HOME/.cargo/bin:$PATH"`
-- Use nightly Rust: `rustup override set nightly` (in zkml directory)
-
-### Tests Fail
-- Check dependencies are installed
-- Verify environment variables are set (for AWS tests)
-- Check GPU availability (for GPU tests)
-
-### AWS GPU Testing Issues
-
-#### "Missing AWS credentials"
-- Ensure all three environment variables are set
-- Check credentials are valid: `aws sts get-caller-identity`
-
-#### "nvidia-smi not found"
-- GPU instance may not have CUDA drivers installed
-- Install NVIDIA drivers: `sudo apt-get install nvidia-driver-535`
-
-#### "Ray initialization failed"
-- Check GPU availability: `nvidia-smi`
-- Verify Ray installation: `pip install --upgrade ray`
-
-#### "PyTorch CUDA not available"
-- Install PyTorch with CUDA: `pip install torch --index-url https://download.pytorch.org/whl/cu118`
-
 ## CI
 
 Lightweight CI runs on every PR to `main` and `dev`:
@@ -352,9 +327,9 @@ Lightweight CI runs on every PR to `main` and `dev`:
 ## TODO: Next Steps
 
 1. ~~**Make Merkle root public**: Add root to public values so next chunk can verify it~~ ✅ Done
-2. **Complete proof generation**: Connect chunk execution to actual proof generation
-3. **Ray-Rust integration**: Connect Python Ray workers to Rust proof generation
-4. **GPU acceleration**: Current implementation is CPU-based. GPU acceleration for proof generation requires additional work (Halo2 GPU support or custom GPU kernels)
+2. **Complete proof generation**: Connect chunk execution to actual proof generation ([#8](https://github.com/ray-project/distributed-zkml/issues/8))
+3. **Ray-Rust integration**: Connect Python Ray workers to Rust proof generation ([#9](https://github.com/ray-project/distributed-zkml/issues/9))
+4. **GPU acceleration**: Current implementation is CPU-based. GPU acceleration for proof generation requires additional work ([#10](https://github.com/ray-project/distributed-zkml/issues/10))
 
 ### GPU Performance Benchmarking