Development Log
The Build Story
41 sessions. 111 source files. 29,740 lines of code. 423+ tests. Built across March 13-22, 2026 with Claude Code as the agent harness.
41
Sessions
111
Files
29.7k
Lines
423+
Tests
Architecture & Full Scaffold
- 12-strategy Darwinian competition model with 3 main strategies x 4 roles
- Dual-AI architecture: Claude for evolution, Venice AI for execution
- 3 Solidity contracts (717 LOC) + full agent core, trading layer, chain layer
Click to expand
Architecture & Full Scaffold
- 12-strategy Darwinian competition model with 3 main strategies x 4 roles
- Dual-AI architecture: Claude for evolution, Venice AI for execution
- 3 Solidity contracts (717 LOC) + full agent core, trading layer, chain layer
Click to expand
ERC-4626 Vault & React DApp
- ERC-4626 tokenized vault (DarwinVaultV2) -- deposit USDC, receive dvUSDC shares
- Agent borrow/return pattern with high water mark performance fees
- Full React DApp: Home, Portfolio, Tournament pages with Wagmi + RainbowKit
Click to expand
ERC-4626 Vault & React DApp
- ERC-4626 tokenized vault (DarwinVaultV2) -- deposit USDC, receive dvUSDC shares
- Agent borrow/return pattern with high water mark performance fees
- Full React DApp: Home, Portfolio, Tournament pages with Wagmi + RainbowKit
Click to expand
Full Test Suite & Audit Sprint
- 59 tests across 4 modules: performance, circuit-breaker, strategy-manager, paper-engine
- Proving Ground Rule: strategies must profit in paper before going live
- Circuit breaker system with 3 independent trip conditions + stale price detection
Click to expand
Full Test Suite & Audit Sprint
- 59 tests across 4 modules: performance, circuit-breaker, strategy-manager, paper-engine
- Proving Ground Rule: strategies must profit in paper before going live
- Circuit breaker system with 3 independent trip conditions + stale price detection
Click to expand
DApp Frontend & Wallet Integration
- Dashboard redesign: dark navy + neon cyan/purple, Press Start 2P + JetBrains Mono fonts
- Strategy Tournament Leaderboard, Recent Trades Feed, Evolution Timeline
- Structured conversation log audit trail for hackathon judge review
Click to expand
DApp Frontend & Wallet Integration
- Dashboard redesign: dark navy + neon cyan/purple, Press Start 2P + JetBrains Mono fonts
- Strategy Tournament Leaderboard, Recent Trades Feed, Evolution Timeline
- Structured conversation log audit trail for hackathon judge review
Click to expand
Cost Optimization & Multi-AI Pipeline
- Three-tier loop: 30s rule-based (free), 2min Claude CLI (free), 4h Venice evolution (~$0.30/day)
- Claude CLI as signal engine via ClaudeCliEngine (Haiku batch eval)
- Venice API for evolution only -- sponsor showcase at minimal cost
Click to expand
Cost Optimization & Multi-AI Pipeline
- Three-tier loop: 30s rule-based (free), 2min Claude CLI (free), 4h Venice evolution (~$0.30/day)
- Claude CLI as signal engine via ClaudeCliEngine (Haiku batch eval)
- Venice API for evolution only -- sponsor showcase at minimal cost
Click to expand
IPFS, Filecoin & Sponsor Integrations
- IPFS genome pinning: CID recorded on-chain via PerformanceLog contract
- Sponsor audit: 6 genuine integrations, 2 skipped as architectural mismatches
- Verifiable evolution: genome JSON on IPFS -> keccak256 hash -> on-chain match
Click to expand
IPFS, Filecoin & Sponsor Integrations
- IPFS genome pinning: CID recorded on-chain via PerformanceLog contract
- Sponsor audit: 6 genuine integrations, 2 skipped as architectural mismatches
- Verifiable evolution: genome JSON on IPFS -> keccak256 hash -> on-chain match
Click to expand
Brutal Audit Sprint
- Root cause: RPC 503 -> no snapshots -> empty AI data -> zero buy signals -> no trades
- Built technical indicators engine: RSI-14, MACD, Bollinger Bands, EMA-9/21
- Added 4 high-volatility tokens (DEGEN, BRETT, VIRTUAL, HIGHER) + rule-based entry bypass
Click to expand
Brutal Audit Sprint
- Root cause: RPC 503 -> no snapshots -> empty AI data -> zero buy signals -> no trades
- Built technical indicators engine: RSI-14, MACD, Bollinger Bands, EMA-9/21
- Added 4 high-volatility tokens (DEGEN, BRETT, VIRTUAL, HIGHER) + rule-based entry bypass
Click to expand
DApp UI/UX Overhaul
- 5-department biological model: Senses, Reflexes, Cortex, Nerves, Marrow
- Adaptive evolution with 3 triggers: timer, emergency (accuracy < 30%), weight shift (> 20%)
- Venice AI for Grok/X sentiment scraping, source fitness scoring
Click to expand
DApp UI/UX Overhaul
- 5-department biological model: Senses, Reflexes, Cortex, Nerves, Marrow
- Adaptive evolution with 3 triggers: timer, emergency (accuracy < 30%), weight shift (> 20%)
- Venice AI for Grok/X sentiment scraping, source fitness scoring
Click to expand
Live Deployment & Battle Testing
- 3 contracts deployed to Base mainnet, vault funded with 73 USDC
- 7-division Immune System: Patrol, Antibodies, Thymus, Platelets, Membrane, Lymph, Genome
- Self-healing engine with fix-then-verify pattern and 12h self-evolution
Click to expand
Live Deployment & Battle Testing
- 3 contracts deployed to Base mainnet, vault funded with 73 USDC
- 7-division Immune System: Patrol, Antibodies, Thymus, Platelets, Membrane, Lymph, Genome
- Self-healing engine with fix-then-verify pattern and 12h self-evolution
Click to expand
Frontier, Championship & Final Audit
- 4 frontier archetypes: Abiogenesis (moonshots), Mitosis (HFT), Cambrian (volatility), Symbiont (whale following)
- Multi-chain: ChainRegistry for Base, Arbitrum, Optimism with 1inch aggregation
- Cross-team championship: Team 4 winner competes against Teams 1-3 champion
Click to expand
Frontier, Championship & Final Audit
- 4 frontier archetypes: Abiogenesis (moonshots), Mitosis (HFT), Cambrian (volatility), Symbiont (whale following)
- Multi-chain: ChainRegistry for Base, Arbitrum, Optimism with 1inch aggregation
- Cross-team championship: Team 4 winner competes against Teams 1-3 champion
Click to expand
Instinct System
- 5 departments: Senses, Reflexes, Cortex, Nerves, Marrow
- Multi-timeframe predictions (1m, 5m, 15m, 1h)
- Adaptive evolution with emergency triggers on accuracy drops
Click to expand
Instinct System
- 5 departments: Senses, Reflexes, Cortex, Nerves, Marrow
- Multi-timeframe predictions (1m, 5m, 15m, 1h)
- Adaptive evolution with emergency triggers on accuracy drops
Click to expand
Instinct Dashboard UI & Evolution Tuning
- Instinct dashboard with live prediction feeds
- Source fitness scoring visible per department
- Evolution cycle acceleration (4h -> 1h during qualification)
Click to expand
Instinct Dashboard UI & Evolution Tuning
- Instinct dashboard with live prediction feeds
- Source fitness scoring visible per department
- Evolution cycle acceleration (4h -> 1h during qualification)
Click to expand
Immune System
- 7 divisions: Patrol, Antibodies, Thymus, Platelets, Membrane, Lymph, Genome
- Self-healing with fix-then-verify pattern
- 12h self-evolution cycle for threshold tuning
Click to expand
Immune System
- 7 divisions: Patrol, Antibodies, Thymus, Platelets, Membrane, Lymph, Genome
- Self-healing with fix-then-verify pattern
- 12h self-evolution cycle for threshold tuning
Click to expand
Frontier System
- 4 archetypes: Abiogenesis, Mitosis, Cambrian, Symbiont
- ChainRegistry for multi-chain with 1inch aggregation
- Cross-team championship bracket
Click to expand
Frontier System
- 4 archetypes: Abiogenesis, Mitosis, Cambrian, Symbiont
- ChainRegistry for multi-chain with 1inch aggregation
- Cross-team championship bracket
Click to expand
DApp Deployment
- React SPA deployed at corduroycloud.com/darwinfi/
- VaultV2 (ERC-4626) integration with deposit/withdraw
- RainbowKit wallet connect + Base chain
Click to expand
DApp Deployment
- React SPA deployed at corduroycloud.com/darwinfi/
- VaultV2 (ERC-4626) integration with deposit/withdraw
- RainbowKit wallet connect + Base chain
Click to expand
VaultV3 Contract & Evolution Enhancements
- VaultV3: 1% management fee + 5% performance fee (high water mark)
- 51 contract tests passing
- Deploy script ready for mainnet launch
Click to expand
VaultV3 Contract & Evolution Enhancements
- VaultV3: 1% management fee + 5% performance fee (high water mark)
- 51 contract tests passing
- Deploy script ready for mainnet launch
Click to expand
Self-Evolution Engine & Frontier Freedom
- Venice AI code generation with git worktree sandboxing
- Canary deploy with auto-rollback on degradation
- 18 smoke tests gate every mutation
Click to expand
Self-Evolution Engine & Frontier Freedom
- Venice AI code generation with git worktree sandboxing
- Canary deploy with auto-rollback on degradation
- 18 smoke tests gate every mutation
Click to expand
AI Router & Autonomy Gap Closures
- AI Router: Claude -> Ollama/Gemma2 -> rule-based fallback chain
- Ollama engine for $0 local AI inference
- Autonomy audit: 40% -> closing 6 critical gaps
Click to expand
AI Router & Autonomy Gap Closures
- AI Router: Claude -> Ollama/Gemma2 -> rule-based fallback chain
- Ollama engine for $0 local AI inference
- Autonomy audit: 40% -> closing 6 critical gaps
Click to expand
VaultV4 Security Hardening
- 48-hour timelock on agent/feeRecipient changes
- 80% maximum borrow ratio with bad debt write-off
- 80 Solidity tests covering all V4 security features
Click to expand
VaultV4 Security Hardening
- 48-hour timelock on agent/feeRecipient changes
- 80% maximum borrow ratio with bad debt write-off
- 80 Solidity tests covering all V4 security features
Click to expand
Showcase Site v1
- Next.js 15 with App Router and Tailwind CSS
- Interactive organism diagram with 6 organs
- Live vault telemetry polling real API data
Click to expand
Showcase Site v1
- Next.js 15 with App Router and Tailwind CSS
- Interactive organism diagram with 6 organs
- Live vault telemetry polling real API data
Click to expand
Lit Protocol Chipotle v3 Integration
- Programmable Key Pair (PKP) minted on Lit naga-dev network
- IPFS-hosted Lit Action with whitelisted contracts and tokens
- SwapRouter calls restricted to exactInputSingle and multicall
Click to expand
Lit Protocol Chipotle v3 Integration
- Programmable Key Pair (PKP) minted on Lit naga-dev network
- IPFS-hosted Lit Action with whitelisted contracts and tokens
- SwapRouter calls restricted to exactInputSingle and multicall
Click to expand
Judge Assessment Framework
- 7 judge personas with distinct scoring rubrics
- Automated HTML email report with score visualization
- Score tracking across assessment rounds
Click to expand
Judge Assessment Framework
- 7 judge personas with distinct scoring rubrics
- Automated HTML email report with score visualization
- Score tracking across assessment rounds
Click to expand
Wave Sprint R1: Score 5.8 to 6.8
- 4 parallel worktree agents executing simultaneously
- Test suite expanded from 325 to 380+ tests
- Circuit breaker documentation and Lit Protocol docs added
Click to expand
Wave Sprint R1: Score 5.8 to 6.8
- 4 parallel worktree agents executing simultaneously
- Test suite expanded from 325 to 380+ tests
- Circuit breaker documentation and Lit Protocol docs added
Click to expand
Wave Sprint R2: Score 6.8 to 7.7
- VaultV4 security hardening (6 audit fixes)
- DApp UI polish and responsive design
- Code quality improvements across agent core
Click to expand
Wave Sprint R2: Score 6.8 to 7.7
- VaultV4 security hardening (6 audit fixes)
- DApp UI polish and responsive design
- Code quality improvements across agent core
Click to expand
Wave Sprint R3: Score 7.7 to 8.0
- Darwin chatbot with Claude Haiku SSE streaming
- Showcase organism diagram interactivity
- Sponsor integration documentation
Click to expand
Wave Sprint R3: Score 7.7 to 8.0
- Darwin chatbot with Claude Haiku SSE streaming
- Showcase organism diagram interactivity
- Sponsor integration documentation
Click to expand
Wave Sprint R4: Score 8.0 to 8.3
- Self-Evolution Engine smoke tests and canary deploy
- Live Results page with PnL chart and leaderboard
- Under the Hood technical deep-dive sections
Click to expand
Wave Sprint R4: Score 8.0 to 8.3
- Self-Evolution Engine smoke tests and canary deploy
- Live Results page with PnL chart and leaderboard
- Under the Hood technical deep-dive sections
Click to expand
Wave Sprint R5: Score 8.3 to 8.4
- Development log expanded to cover all 40 sessions
- Product page with feature showcase and tech stack
- Edge case test coverage for vault operations
Click to expand
Wave Sprint R5: Score 8.3 to 8.4
- Development log expanded to cover all 40 sessions
- Product page with feature showcase and tech stack
- Edge case test coverage for vault operations
Click to expand
Wave Swarm: 7 Parallel Agents
- 7 parallel worktree agents executing simultaneously
- Showcase accuracy fixes across all pages
- Dev log, Notion battle plan, and briefing page
Click to expand
Wave Swarm: 7 Parallel Agents
- 7 parallel worktree agents executing simultaneously
- Showcase accuracy fixes across all pages
- Dev log, Notion battle plan, and briefing page
Click to expand
Demo Video Pipeline
- ElevenLabs TTS narration with Daniel voice
- Playwright-based DApp screen recording
- ffmpeg video composition pipeline
Click to expand
Demo Video Pipeline
- ElevenLabs TTS narration with Daniel voice
- Playwright-based DApp screen recording
- ffmpeg video composition pipeline
Click to expand
Final Sprint: Deadline Day Polish
- Live Results migrated from showcase to DApp
- Chatbot evolution thinking words
- Logo spacing and organism diagram readability
Click to expand
Final Sprint: Deadline Day Polish
- Live Results migrated from showcase to DApp
- Chatbot evolution thinking words
- Logo spacing and organism diagram readability
Click to expand
What Comes Next
The Frontier bots are hunting across chains. The Immune System is self-healing. The vault is open. DarwinFi is a living financial organism -- and it never stops evolving.