Develop 800% faster – with on-premise AI

AI is revolutionizing software development – and not just in the cloud! Imagine you could develop a complex full-stack monitoring dashboard in just a fraction of the time you would normally need for it. Sounds too good to be true? Not with the right AI tools and a well-thought-out on-premise approach!

In this blog post, we dive into the fascinating world of AI-assisted software development, which is not only blazingly fast but also works without any cloud requirement. We take a look at a concrete project: a full-stack monitoring dashboard, developed with React, Node.js, and an AWS database connection – and all of it in just 6 hours!

But the best part? You don’t have to rely on OpenAI, GitHub, or Chinese providers. Instead, we rely on innovative open-source solutions and local development environments that protect your data and give you full control over your development process.

What awaits you:

• An overview of the tech stack used that makes this speed possible
• Insights into the AI tools deployed: Bolt.new for the frontend and Continue.dev with DeepSeek v3 for backend and database integration
• A detailed comparison of development times: AI-assisted vs. manual
• Practical experiences and learnings from the project
• A discussion of the pros and cons of on-premise development compared to cloud solutions

Are you ready to multiply your development speed eightfold while staying independent of the big cloud providers? Then buckle up – we’re heading into the future of software development!

Technology and tools

In the world of AI-assisted software development, there are a multitude of tools, but not all of them are created equal. For our project, we opted for two particularly powerful and flexible solutions: Bolt.new for the frontend and Continue.dev with DeepSeek v3 for the backend and database integration. Let’s take a closer look at these tools.

Bolt.new: The turbo for your frontend development

Bolt.new is a fascinating open-source AI tool that takes frontend development to a new level. Imagine you simply describe what you want to build, and the AI conjures up a working prototype for you – sounds like science fiction, but it’s reality!

Here are some highlights:

• Prompt-based development: You give natural-language instructions, and Bolt.new generates the corresponding code.
• Lightning-fast prototyping: Perfect for quick UI drafts and initial versions of your application.
• Browser-based workflow: From ideation to deployment, everything in your browser.
• Package integration: Automatic installation and integration of external packages without manual intervention.

In our project, Bolt.new proved invaluable for rapidly building the frontend scaffold. However, it’s important to note that the tool sometimes has its own ideas about the ideal tech stack – this is where your expertise as a developer is needed to adjust the output when necessary.

Continue.dev and DeepSeek v3: The dream team for backend and database

For backend development and database integration, we relied on the combination of Continue.dev and DeepSeek v3 – and were not disappointed.

Continue.dev:

• Seamless IDE integration: Work in your familiar environment such as VS Code or JetBrains IDEs (in our case as a VS Code extension).
• Context-aware AI assistance: Understands the context of your project and delivers relevant suggestions.
• Flexibility in model choice: Supports various AI models, including almost all major local and open-source options.
• Customizable commands: Create your own slash commands for frequent tasks.

DeepSeek v3:

• State-of-the-art performance: Even surpasses GPT-4 and other leading models in many areas.
• Huge context window: Can process up to 128K tokens – ideal for complex projects.
• Efficient MoE (Mixture of Experts) architecture: Uses 671 billion parameters but activates only 37 billion per token.
• Versatile applicability: Outstanding in coding, mathematics, and logical reasoning.

The combination of these tools enabled us to generate precise and efficient backend code and to implement the database integration in record time. Particularly impressive was DeepSeek v3’s ability to understand complex relationships and propose tailored solutions.

One important point: although these tools are extremely powerful, they do not replace the expertise of an experienced developer. Rather, they should be viewed as highly efficient assistants that take your productivity and creativity to a new level.

In the next sections, we’ll look at how we used these tools concretely in our project and what astonishing time savings we were able to achieve as a result. Are you ready to revolutionize your development speed?

Project description

Let’s take a closer look at our project: a full-stack monitoring dashboard that we developed in record time with the help of AI. This project is a perfect example of how modern AI tools can accelerate software development without compromising on quality or functionality.

The goal: An interactive dashboard for analyzing user data

Our dashboard was meant to fulfill two main functions:

1. Analysis of registration metrics:

• Display of the number of registrations for any given days or time periods
• Visualization of user distribution by country

2. Overview of user settings:

• Number of users who have set up Brandvoice
• Number of users with Brandvoice enabled by default
• Distribution of preferences for short or long presets
• Statistics on completion of the setup wizard

Technical structure

Frontend:

• React with TypeScript for robust and type-safe development
• Material UI for an appealing and consistent design
• Interactive components such as DatePicker for time-period selection
• Responsive design for optimal use on various devices

Backend:

• Node.js for efficient server-side processing
• Connection to two separate AWS DynamoDB databases:
  1. “user-analytics-db” for registration data
  2. “user-settings-db” for user settings

Database integration:

• Initially mocked up with local JSON files for fast development and testing
• Later seamless integration with AWS DynamoDB for scalability and real-time data
• Implementation of efficient queries for optimal performance

Development process

1. Frontend prototyping with Bolt.new:

• Rapid creation of the UI base scaffold through AI-generated components
• Iterative refinement of the design through precise prompts
• Adaptation of the generated code to specific project requirements

2. Backend development with Continue.dev and DeepSeek v3:

• Creation of the server structure and API endpoints
• Implementation of the database logic, initially with a local JSON server
• Step-by-step migration to AWS DynamoDB integration

3. Fine-tuning and integration:

• Connecting frontend and backend
• Implementation of real-time data updates
• Optimization of query performance and user interface

Challenges and solutions

• Adapting the code generated by Bolt.new to MaterialUI specifications
• Fine-tuning the database queries for optimal performance
• Ensuring data consistency between mock data and the real DynamoDB

The result

After just 6 hours of intensive work with our AI assistants, a fully functional, interactive dashboard was ready. It offers not only an appealing user interface but also robust backend functionality and a scalable database connection.

This project impressively demonstrates how AI-assisted development tools can accelerate the entire development process – from the initial idea to the finished product. In the next section, we’ll look at how this development time compares to traditional, manual development. Spoiler: the difference is impressive!

Comparison of development time

One of the most fascinating aspects of our project was the drastic reduction in development time through the use of AI tools. Let’s take a detailed look at how development times differ with and without AI assistance.

Without AI: The manual development effort

With a traditional approach, our full-stack monitoring dashboard would have taken considerably more time. Based on our experience and estimates, the process would have looked roughly like this:

1. Frontend development:

• UX design: approx. 16 hours
• Implementation in React with Material UI: approx. 24 hours

2. Backend development and database connection:

• Database integration and queries: approx. 8 hours

Estimated total time without AI: approx. 48 hours

That corresponds to about 6 full working days for a junior full-stack developer – and that’s still an optimistic estimate!

With AI: How Bolt.new and Continue.dev accelerate development

Now let’s look at how development time changed with our AI assistants:

1. Frontend with Bolt.new:

• Initial UI design and component generation: approx. 30 minutes
• Iterative refinement and adjustments: approx. 30 minutes
• Total time frontend: approx. 1 hour

2. Backend and database integration with Continue.dev and DeepSeek v3:

• Database connection (incl. transition from mock data to DynamoDB): approx. 4 hours
• Refinement and adaptation of the frontend: approx. 1 hour
• Total time backend: approx. 5 hours

Actual total development time with AI: 6 hours

The numbers speak for themselves: we accelerated development time by an impressive 800%!

Detailed time-savings analysis:

• Frontend development: 95% faster (40 hours vs. 2 hours)
• Backend development: 50% faster (8 hours vs. 4 hours)
• Overall project: 83.33% faster (48 hours vs. 6 hours)

What does this mean in practice?

1. Drastically shortened time-to-market: Projects that would otherwise have taken weeks can now be implemented in days or even hours.
2. Increased agility: Faster iterations and adjustments to customer feedback are possible.
3. Cost efficiency: Fewer developer hours mean lower project costs.
4. Focus on creativity and strategy: Developers can concentrate on higher-value tasks while repetitive coding work is taken over by the AI.
5. Faster prototyping cycles: Ideas can be turned into functional demos more quickly.

It’s important to note that these time savings do not come at the expense of code quality. The solutions generated by AI are often at the level of experienced developers and can, in many cases, even lead to cleaner and more efficient implementations.

Conclusion

The combination of Bolt.new for the frontend and Continue.dev with DeepSeek v3 for backend and database integration has proven to be a real game-changer. It enables a development speed that seemed unthinkable just a few years ago.

However, it’s important to emphasize that these tools do not replace the developer but rather empower them. The role of the developer shifts from a pure coder to an architect and problem-solver who deploys the AI tools strategically to achieve optimal results.

In the next section, we’ll take a closer look at the experiences and insights we gained from this project and how you can deploy these technologies in your own development process.

Results and learnings

Our project with Bolt.new and Continue.dev was not only a success in terms of development speed but also a source of valuable insights into the use of AI in software development. Here are our most important learnings and observations:

Bolt.new: Lightning-fast frontend prototyping

Strengths:

• Extremely fast generation of complete UIs
• Intuitive, semantic prompt interaction
• Flexibility in design adjustments

Challenges:

• Occasional deviations from the desired tech stack
• The need for precise prompts for optimal results

Key learnings:

1. Creativity and efficiency: Bolt.new particularly shines at quick UI prototypes and makes it possible to visualize creative ideas rapidly.
2. Prompt engineering: The quality of the output depends heavily on the clarity and precision of the prompts. It pays off to invest time in formulating good prompts.
3. Post-processing: Although the generated UIs are often impressive, a review and fine-tuning by experienced developers is essential to ensure consistency and best practices.

Continue.dev and DeepSeek v3: Precision in the backend

Strengths:

• High-quality code suggestions, often at expert level
• Excellent performance on complex tasks such as database integration
• Flexibility through support for various AI models
• Very up-to-date training data (July 2024), important for development tasks

Challenges:

• Learning curve when using complex implementations
• Occasional delays in code generation, especially with local models (due to the high graphics requirements)

Key learnings:

1. Context awareness is crucial: The more context (e.g. project structure, dependencies) the model has available, the more precise and useful its suggestions are.
2. Model choice makes a difference: DeepSeek v3 showed impressive performance, especially on complex coding tasks and logical reasoning.
3. Human-AI collaboration: We achieved the best results through close collaboration between developer and AI, with the AI acting as a powerful assistant.

Comparison with other AI models

In our test, DeepSeek v3 even performed better than GPT-4 in terms of code quality, especially on complex programming tasks. However, there are trade-offs:

• Speed: DeepSeek v3 can be slower than some cloud-based alternatives.
• API availability: Usage can be more involved than with established cloud services, if you cannot run it locally.

Practical tips for deployment

1. Iterative working: Use the speed of the AI tools for fast iterations and continuous improvements.
2. Combining tools: Using specialized tools for frontend (Bolt.new) and backend (Continue.dev) enables optimal results in both areas.
3. Code review: Even AI-generated code should be carefully reviewed. The AI can produce excellent code, but human judgment remains indispensable.
4. Adaptation to the workflow: Integrate the AI tools step by step into your existing development process and adapt them to your specific needs.
5. Continuous learning: The AI landscape is evolving rapidly. Stay up to date on new features and models to get the most out of these tools.

Conclusion

Our project has shown that AI-assisted development tools such as Bolt.new, Continue.dev, and DeepSeek v3 have the potential to fundamentally change the way we develop software. They not only enable a drastic acceleration of the development process but also open up new possibilities for creativity and innovation.

Nevertheless, it’s important to emphasize that these tools do not replace the expertise and experience of human developers but rather complement them. The key to success lies in the skillful combination of human creativity and judgment with the efficiency and precision of AI.

In the next section, we’ll look at how this approach of on-premise development with AI compares to cloud-based solutions and what advantages it offers in terms of data protection and independence.

On-premise development vs. cloud solutions

In an era where cloud-based development environments and AI services are ubiquitous, our approach of relying on on-premise solutions may seem surprising at first. But this decision brings decisive advantages, especially when it comes to data protection, control, and independence. Let’s take a closer look at how on-premise development with AI compares to cloud solutions.

Advantages of local development with AI

1. Data protection and security

• Full control over sensitive data: Your code and your trade secrets stay within your network.
• Compliance with strict data protection guidelines: Ideal for companies with high compliance requirements.
• No risk of data leaks through third-party providers: You are not dependent on the security infrastructure of external cloud providers.

2. Independence from cloud providers

• No vendor lock-ins: You are not tied to a particular cloud provider.
• Flexibility in model choice: Freedom to test and switch between various AI models.
• Cost control: No unexpected cloud usage fees or billing surprises.

3. Adaptability and control

• Tailored solutions: The ability to adapt AI models to specific project requirements.
• Complete control over the infrastructure: You decide on updates, security patches, and configurations.
• Offline capability: Development is possible even without a constant internet connection.

4. Performance and latency

• Potentially lower latency: Especially important for real-time applications and fast feedback during development.
• Consistent performance: No dependency on network quality or cloud server load.

Challenges of on-premise development

Of course, the on-premise approach also brings some challenges with it:

1. Initial setup and maintenance

• Higher effort during the initial configuration of the development environment or AI server infrastructure
• The need for regular updates and maintenance of the local infrastructure.

2. Hardware requirements

• Powerful hardware required, especially for compute-intensive AI models such as DeepSeek v3.
• Potentially higher acquisition costs compared to cloud-based pay-as-you-go models.

3. Scalability

• Limited ability to scale quickly during load peaks, in contrast to elastic cloud solutions.

4. Collaboration

• Additional effort for setting up collaborative development environments compared to cloud-based solutions.

Comparison with cloud solutions

Cloud-based development environments and AI services undoubtedly offer advantages such as easy scalability, global availability, and often a broader range of ready-to-use services. However, they also come with their own challenges:

• Data protection concerns: Sensitive data leaves the corporate network and may possibly be used for training purposes.
• Dependency on internet connection and external services.
• Potentially higher long-term costs with intensive use.
• Limited control over the underlying infrastructure.

Our conclusion

For our project, the on-premise approach proved to be the ideal solution. It enabled us to achieve:

1. Maximum data security and compliance with strict requirements.
2. Flexibility in choosing and adapting the AI models.
3. Full control over our development environment and data.
4. Consistent performance without dependency on external factors.

Despite the initial challenges of setting it up, the long-term advantages in terms of security, control, and independence more than outweighed the effort.

It’s important to emphasize that the choice between on-premise and cloud solutions depends heavily on the specific requirements and circumstances of a project or company. For many scenarios, a hybrid solution that combines the advantages of both worlds can be the optimal path.

In our case, the combination of a local development environment, on-premise AI tools such as Bolt.new and Continue.dev, and the powerful DeepSeek v3 model created a development environment that perfectly unites speed, security, and flexibility.

In the final chapter, we’ll draw a conclusion and take a look at the future of AI-assisted software development.

Conclusion and outlook

Our project of developing a full-stack monitoring dashboard in just 6 hours impressively demonstrated the potential that lies in the combination of AI-assisted development tools and on-premise solutions. Let’s summarize the most important findings and take a look into the future.

Summary of the advantages

1. Drastic time savings: A reduction in development time by over 80% is a game-changer for the software industry. This enables faster market launches, more agile adaptations to customer needs, and more room for innovation.
2. Quality and efficiency: The AI-generated solutions were not only fast but also of high quality. Tools like Bolt.new for the frontend and Continue.dev with DeepSeek v3 for the backend delivered expert-level code with state-of-the-art frameworks.
3. Data protection and control: Our on-premise approach ensured the highest security standards and full control over sensitive data and trade secrets.
4. Flexibility and independence: The freedom to choose and adapt various AI models without being tied to a cloud provider proved to be a major advantage.
5. Cost efficiency: Despite the initial investments in hardware, the on-premise approach offers better cost control in the long term and avoids unexpected cloud fees.

Challenges and solution approaches

1. Initial setup: The initial effort for configuring the on-premise environment can be minimized through well-documented setups and automated setup processes.
2. Hardware requirements: Investments in powerful hardware pay off through flexibility and long-term cost savings.
3. Scalability: Hybrid solutions that combine on-premise systems with cloud resources for load peaks can address this issue.
4. Continuous training: The rapidly evolving AI landscape requires regular training and updates in order to exploit the full potential of the tools.

Outlook: The future of AI-assisted software development

1. Even smarter AI assistants: Future versions of tools like Bolt.new and Continue.dev will presumably be even better able to understand context and proactively suggest solutions.
2. Improved on-premise solutions: We expect a further development of powerful, local AI models that will continue to close the gap to cloud-based solutions.
3. AI-assisted software architecture: In the future, AI could help not only with code generation but also with the optimization of system architectures and the prediction of potential scalability or security problems.
4. Strengthened human-AI collaboration: The role of developers will continue to shift toward higher-value, strategic tasks, while AI increasingly takes over repetitive tasks.
5. Ethics and responsibility: With the growing importance of AI in software development, questions of ethics, responsibility, and transparency will move to the foreground.

Closing thoughts

Our project has shown that integrating AI tools into the development process, combined with a well-thought-out on-premise approach, has the potential to fundamentally change software development. We stand at the beginning of a new era in which the boundaries between human creativity and machine efficiency are increasingly blurring.

The challenge for companies and developers will be to use these powerful tools responsibly and effectively. It’s not about replacing human developers but about empowering them to solve more complex problems and create more innovative solutions.

For companies, especially in the SME sector, this development offers enormous opportunities. The ability to develop high-quality software in a fraction of the time can be a decisive competitive advantage. At the same time, the on-premise approach enables compliance with strict data protection guidelines and offers the necessary flexibility to respond to changing market requirements.

The future of software development lies in the skillful combination of human expertise with AI-assisted tools. Companies that recognize and leverage this synergy early will be ahead in the rapidly evolving digital landscape.

We invite you to be part of this exciting development. Experiment with AI tools, explore the possibilities of on-premise solutions, and find your own way to integrate these technologies into your development process. The future of software development begins now – and it’s more exciting than ever!

The AI Software Company supports small and medium-sized software firms in the DACH region in making their development processes more efficient, faster, and future-proof with AI.

Curious about how to use AI meaningfully in your software team? Sign up for our newsletter and receive valuable tips, insights, and updates!