updated test strategy

Author: LyonSsS

TEST_STRATEGY.md

@@ -217,7 +217,7 @@ Reports capture all console output, making it easy to review test runs and debug
 
 5. ✅ **Gas Optimization**: Implemented robust gas estimation with buffers for testnet reliability
 
-### For more comprehensive tests I would add the following
+### If I had more time:
 
 1. **Gas Benchmarking**: More detailed gas cost analysis and optimization metrics.
 2. **Time Manipulation Edge Cases**: More comprehensive tests for deadline boundaries (exactly at deadline, 1 second before/after).
@@ -226,4 +226,23 @@ Reports capture all console output, making it easy to review test runs and debug
 5. **Event Indexing Tests**: Test event filtering and querying patterns that a real dApp would use.
 6. **CI/CD Integration**: Enhanced GitHub Actions workflow to run all test suites (Solidity + TypeScript) automatically.
 
+### AI coding assistance: 
 
+**What worked well**:
+- Code structure suggestions: AI helped organize the contract structure with clear separation of concerns.
+- Test pattern generation: AI suggested good patterns for testing state transitions and edge cases.
+- Error handling: AI helped identify potential revert conditions and appropriate error messages.
+- General set-up: fixed to solve dependency issues and integration missoncfiguration
+- Compare various approaches beneffits vs disadvantages
+
+**What did not work well**:
+- Complex logic refinement: Initial reveal logic was overcomplicated; required manual simplification.
+- Context preservation: Sometimes AI would lose track of the full game flow when making incremental changes.
+- Wehn debating scenario covers AI did missed a lot of edge case scenarios.
+- AI missed to understand the logic of the contract to what will happen if someone will not reveal in the reveal period ( as the funds where locked indefitinie). I had to redeploy a new contract with updated logic
+- Generated incorrect assertions based on my requirements. Also the assertions had to be double checked and enforced as scenarios passed even with incorrect values. 
+
+
+**Overall**: In the end 90% ++ of the code here is generated by AI based on my requirements and I was the one overseeing, checking and reviewing all that is done. I am using cursor and usually I do use ts + etherjs ( +ABI of the contract) to interrecat with my needs and a lot of RPC "read" from chian methods to validate assertsion ( based on parsed events in 90% of the cases). 
+My recent projects where I am testing a solver that does swap, I did found a cool set-up in golang with W3vm which I had no idea how it worked and did not knew go, but I managed with AI to make comprehensive tests done on eth fork and base fork, where I querry Dune for cowSwap last 30 days top 50 pairs sorted by USD value then create them into a JSON file with scenarios and feed it to this go set-up to execute scenarios on a forked state of eth or base chain. 
+I would not be great at live coding but I can handle what ever task I have in front of me with the help of AI.