“Free Code Agent: Build Faster Right in Your RAD Studio IDE”
Presented by Yai (CTO at Solutions, Embarcadero MVP)
Core Idea
* AI Code Agents are tools that accelerate coding tasks by generating, refactoring, and scaffolding code inside your IDE.
* They do not replace developers; they augment productivity.
* Example used: Gemini CLI (Google AI) integrated with Delphi IDE.
Key Demonstrations
1. Practical Example:
* Used Gemini to generate a Delphi API client based on provided specs.
* AI created:
* DTO classes
* JSON serialization methods
* UI form with components
* Compilation errors occurred due to wrong specs, proving:
* AI follows instructions strictly.
* Developer responsibility: write accurate specifications.
2. Bug Fixing & Refactoring:
* AI used to remove compiler hints/warnings.
* Demonstrated auto-fix loop for bugs.
* AI can update documentation dynamically based on current code state.
3. Additional Use Cases:
* Language translation for UI elements.
* Unit test generation.
* API integration scaffolding.
* Batch scripts for CI/CD with AI-assisted bug fixing.
What a Code Agent Is (and Isn’t)
* Not low-code/no-code or VB-like coding.
* Designed for experienced developers.
* Works as a feedback loop:
* Reads context
* Executes tasks
* Learns from previous instructions
* Can use tools (search, file generation, etc.) and maintain memory for context.
Benefits
* Productivity boost:
* GitHub reports: AI-assisted devs deliver 50% faster.
* Adora report: 98% more pull requests.
* Quality improvement:
* Reduces human typos and fatigue errors.
* Faster deployment & precise refactoring.
* Dynamic documentation generation.
Risks & Challenges
* Security concerns:
* Free tools (like Gemini) may use your code for training.
* Recommendation: filter sensitive code (move unique logic to separate folders).
* Complexity:
* AI often writes code using advanced patterns (generics, best practices).
* Junior devs may struggle to understand AI-generated code.
* Skill erosion:
* Over-reliance on AI can reduce coding proficiency.
* Bugs introduced by AI:
* Always keep a human in the loop.
Cost Models
* API per token: Best for enterprise (robust, secure, expensive).
* Subscription plans:
* Pro Plus: $20/user/month.
* Max plans: $200/user/month.
* Free options:
* Gemini via Google AI Studio (limited quota).
* Local LLMs (e.g., Ollama) – require strong hardware.
Integration Tips
* Best approach: Use a code agent with context (Gemini CLI).
* IDE integration via Tools > Options > Smart Code Inside:
* Configure AI editor with API keys.
* Objective prompts:
* Clear, complete instructions = better results.
* Use auto-fix loops for CI/CD pipelines.
Final Considerations
* AI is not magic:
* It won’t replace developers.
* It’s a productivity accelerator, not a full solution.
* Developers can choose their role:
* Delegate boring tasks to AI.
* Focus on core logic and architecture.
* Stay competitive:
* Companies using AI will outpace those who don’t.
* Future-proof your skills:
* Use AI wisely, keep coding sharp.
Key Message
“AI won’t replace you. Developers using AI will.”
> Use AI to deliver more, faster, and with better quality, while maintaining control and security.
See the rest of the sessions here.