Stop Asking AI to Write Your Code (And What to Do Instead)

Stop Asking AI to Write Your Code (And What to Do Instead)

From the Codecamp Stage_

Talks That Make You Think and Build Better

Identifying and fixing Issues in Code using AI-based tools, with Venkat Subramaniam

“I find it to be really horrible as a code generator, but it's fantastic as a code analyser.” Venkat Subramaniam shares practical insights on the true power - and the hidden dangers - of using AI in software development. Learn why relying on AI to write production code from scratch is a trap, and how to instead use it for ideation and debugging. Discover how AI can become your ultimate code reviewer, specifically designed to handle the heavy cognitive load that developers face when dealing with complex or legacy code.

Key takeaways:

  • Stop asking AI for solutions; use it to generate ideas, evaluate options, and identify potential issues.
  • AI excels at reducing cognitive load by analysing, explaining, and documenting highly complex or unreadable code.
  • Critical thinking is your most valuable skill - you must understand the fundamentals to validate AI-suggested code and prevent side effects.
  • AI tools are surprisingly good at spotting logical bugs (like mutating shared state in parallel streams) that humans easily miss.
  • When dealing with unknown codebases, AI can rapidly identify problems and even help generate unit tests to make the system safer to maintain.

Why it matters: This talk is essential for programmers, tech leads, and architects who want to integrate AI into their workflow safely and effectively. It provides a realistic, hands-on perspective - steering you clear of the hype - so you can leverage AI to solve problems without losing control of your codebase.


Masterclasses That Turn Insight Into Impact_

Deep Dive: AI Assisted Programming, with Venkat Subramaniam

Article content

Accelerate your development while maintaining total control. In this two-day masterclass, Venkat Subramaniam guides you through hands-on exercises and practical examples to harness the power of AI tools without being burned by their unpredictability.

You’ll explore:

  • The true power and perils of AI, clearly identifying where it shines and where it fails in application development.
  • Techniques for using AI to analyse, reason about, and document legacy code to reduce cognitive load.
  • How to evaluate generated code and set up guard rails, system prompts, and contexts to steer AI tools toward better quality.
  • The capabilities and issues of "vibe coding," rapid prototyping, and specification-driven development using state-of-the-art tools (Claude, Copilot, Cursor, etc.).
  • Practical exercises that combine creating tests, finding bugs, and solving real problems while evaluating the effectiveness of the generated results.

📍 Session available:

Bucharest: 5 - 6 May 2026 - Hotel Caro

Small groups • Hands-on coding • Practical strategies for safe AI adoption

Spots are limited. Secure your seat:

Codecamp_Masterclasses

Great presentation!! the use of AI significantly simplifies and reduces the cognitive load of reading and understanding old code, also it is really good at challenging your solutions, finding cases your never think about ( that's why it is really good at writing tests ).

Wow 🤩 sir … your talk was bang on… 3 things which really were pretty insightful (from the 15 mins I watched) AI shouldn’t be referred to as Artificial intelligence, but it should be Accelerated Inference Humans should think (though it’s considered optional 😅) and AI should execute Don’t use AI to solve but use it to ideate and brainstorm… And one key thing which I was also feel and you validated is that, we humans are accountable for our work and hold reputation and for me sometimes the fear of putting my reputation at stake motivates me to push myself , I don’t think AI has that and it can just get away with a sorry … 🤣

Like
Reply

To view or add a comment, sign in

More articles by Codecamp Romania

Others also viewed

Explore content categories