September 20, 2024

How to Use AI to Level Up: A Tech Startup Perspective

How to Use AI to Level Up: A Tech Startup Perspective

In today’s day and age AI is the new hot topic. The next big thing. The Sabrina Carpenter or Lamin Yamal of the tech world. And while it may not be able to serenade an arena full of screaming fans (yet) or dribble down the wing and whip a cross in (yet!), it is quite the handy tool to help startups level up and scale their ops rather quickly. However, you’ll still find a small pocket of business owners, startup gurus, and tech talkers disparaging AI on various soapboxes across the internet and playing down its utility as a business booster.

In the opposite corner of these naysayers, the team at HTCD has been actively using AI to enhance our workflow, productivity, and overall efficiency. Through the perspectives of key members of the engineering team, we gain a deeper understanding of how AI is transforming the way we approach development, ideation, and problem-solving.

"Those Who Don't Like It Don't Know How to Use It."

William Becerra, HTCD’s Director of Engineering, emphasizes that AI, in its current state, is a tool that can enhance a developer's capabilities rather than replace them. He notes that "AI will not replace human developers soon; it will just make us who use it 10x more efficient." However, he also adds the disclaimer that this opinion is based on the current capabilities of AI. He jokingly adds, “If GPT-5 comes with some Sci-fi tech. I don't know”.

William observes distinct patterns in AI adoption among developers:

  1. Resistance due to Lack of Understanding: AI is a catalyst, not a replacement, and "Those who don't like it don't know how to use it." There's a clear correlation between familiarity with AI tools and their acceptance.
  2. Fear of Over-Reliance: Some developers are cautious about becoming too dependent on AI, fearing it could stifle their own problem-solving abilities.
  3. Shallow Prompts Lead to Shallow Results: The quality of the AI’s output is highly dependent on the depth of knowledge embedded in the prompts. As William puts it, "Lack of depth in knowledge makes your prompt shallow and therefore results in naive solutions," which can lead to the misconception that AI isn't as capable as it actually is.

“AI as a Coding Assistant and an Ideation Tool.”

Subham Kundu, Principal AI Engineer, highlights AI's role as a coding assistant, noting its usefulness in day-to-day coding tasks and initial feature ideation. AI helps convert abstract ideas into concrete code and provides a solid foundation for new concepts.

Beyond coding, Subham leverages AI for writing and editing, particularly for transforming complex ideas into well-structured blog drafts and performing grammar checks. This dual functionality makes AI an invaluable tool in both technical and communication tasks. AI ensures clarity and precision in not just helping spur ideas but also in documenting them.

“AI Is a Professional Solution Evaluator and a Code Companion.”

William further expands on AI's role as a professional solution evaluator. He appreciates how AI can provide pros and cons for various aspects of development:

  • Database schema relationships
  • Architecture considerations
  • Industry patterns

William also uses AI as a code companion, where it excels in generating code from various inputs:

  • Generating code from UI mockups and tech specs: This allows developers to quickly prototype interfaces.
  • Database queries: AI’s proficiency in SQL or Prisma query generation is particularly highlighted.
  • Postman API tests: AI can generate specific test cases from example responses and API definitions, speeding up the testing phase.
  • Code review and understanding: AI assists in identifying potential failure points in code and explaining complex code snippets.

“AI Helps Exploring Alternative Implementations.”

Regan van Heerden, Principal Software Engineer, enjoys using AI to explore different implementation strategies. By presenting AI with the requirements of a feature or architectural element, he can obtain various approaches, each with its own set of pros and cons. This helps in making more informed decisions and understanding the trade-offs involved in each option.

Comparing AI Tools: Copilot, ChatGPT, and Gemini

William shares his experiences with some of the most popular AI tools available today:

  • GitHub Copilot: Described as "intensiveness in steroids," Copilot enhances productivity significantly by offering code suggestions as you type.
  • Gemini: Known for its ability to generate well-articulated English explanations, though its code suggestions may sometimes be naive or impractical without further refinement.
  • ChatGPT: William finds ChatGPT particularly effective in generating full solutions from screenshots. He recounts successful projects where ChatGPT generated nearly complete UIs, requiring only minor adjustments to fit their specific needs.

Final Thoughts: The Human-AI Synergy

The consensus among HTCD’s engineering team is that while AI is a powerful tool, its effectiveness is largely determined by how well it is used. A deep understanding of the problem at hand, combined with clear and specific prompts, allows AI to deliver solutions that are not only practical but also innovative.

As William succinctly puts it, "AI will not replace human developers soon; it will just make us who use it 10x more efficient." And it most certainly does, as long as you know how to use it.

William Becerra

LinkedIn logo
Director of Engineering

Related Articles

Back to blog