reading time: 4 minutes

Have you ever stared blankly at a screen, despairing over PHP syntax, or a CSS property that refuses to behave? “Ah! If only there was someone, or something, to guide me,” you’ve likely lamented. Enter the realm of Large Language Models, the curious AI sprites that seem like something out of a sci-fi movie, always keen to assist. But does their utility match their intrigue?

At the onset of META-STRAND, I wasn’t merely hunting for a tool. I needed a guide. A mentor. The Morpheus to my Neo, guiding me through the digital rabbit hole. ChatGPT quickly became that mentor. Believe me, I approached it as skeptical as Sherlock Holmes eyeing an ‘obvious’ clue, but this chatbot? It seemed to pull code from the very ether. One moment, ChatGPT serenaded me with flawless code that Mozart would envy if he were, you know, a programmer. The next, offering CSS tweaks that felt eerily like sage advice from a seasoned developer.

But META-STRAND wasn’t just about code. The essence of any site is its content. And here, my non-native English self – juggling multiple linguistic balls, with English being an elusive one – yearned for fluency. Enter Claude. Not a beret-wearing French linguist, but a chatbot cut from the similar digital language cloth as ChatGPT. If ChatGPT was my coding sage, Claude was the meticulous editor, ensuring each piece I typed is as refined as a cup of Earl Grey. A touch of Neil Gaiman here, a sprinkle of J. R. R. Tolkien there, Claude refined my words with finesse. Claude often took me by surprise. For every moment of exasperation, there were ten of astonishment. I’d submit a piece, confident in its perfection, only to be gently nudged to consider a nuance, a shade, a flair – ever so meticulous in its critique, yet seemingly endlessly endurant.

While META-STRAND’s bond with the likes of ChatGPT and Claude has been fruitful, it’s essential to acknowledge the broader landscape of AI-driven code tools that are shaping the modern developer’s toolkit. In recent times, we’ve witnessed GitHub’s Copilot X, an offspring of OpenAI’s GPT-4, which arrived in March. Beyond the automated completion of code and comments, Copilot X also lends its expertise directly within integrated development environments like Visual Studio and VS Code, revolutionizing the way developers navigate the coding process. Not to be outdone, Amazon threw its hat into the ring with CodeWhisperer, another AI tool keen on writing, verifying, and updating code. On the other hand, Google’s DeepMind team showcased their prowess with AlphaCode, a tool that crafts code from plain language instructions. Lastly, let’s not forget Code Llama. Underpinned by Meta’s Llama 2 language model, it boasts of a flexibility that surpasses its predecessor. Made available to the public, Llama 2 has proven its mettle as a robust, open-source tool. This inclusive approach by Meta, in collaboration with Microsoft, signifies a shift towards democratizing AI and making it accessible to all.

“Now, hold up,” I can hear a voice from the back (probably you, m). “Why use chatbots for PHP and CSS code checking when there are other, resource-saving options like linters, IDE-based tools, or even community-driven platforms like Stack Overflow? Aren’t they more efficient, not to mention environmentally conscious given the energy-intensive nature of large language models?” Indeed, m, you raise a valid concern. Each of these AI entities consumes energy, and as stewards of both technological and natural realms, we ought to use these resources judiciously. But while code analysing tools or Stack Overflow are undeniably valuable and can be more efficient in certain contexts, the allure of chatbots lies in their adaptability and dynamic interaction – and also in their patience. They offer immediate feedback, sometimes in a more comprehensible manner for beginners. Plus, there’s a certain charm in conversing with a digital entity that mimics a certain conversationality, even though it’s miles away from being sentient. Yet, it’s crucial to strike a balance and be aware of the carbon footprint that such interactions might carry, ensuring we make conscious choices in the digital tools we utilize.

“And are we not concerned about trusting machines for such pivotal tasks?” An excellent inquiry, reminiscent of the great Turing Test debates of the ’50s! It’s a tango, really, entrusting machines whilst maintaining an intellectual distance. Trust me, the irony isn’t lost on me. Here I was, a non-native English speaker, entrusting a digital entity to refine my words. It’s like asking a robot to teach you to dance; you wonder if it has the rhythm, but damn, can it execute a perfect moonwalk! And with META-STRAND, our chatbot counterparts are less about terminating and more about… iterating. As I’ve often mused, if HAL 9000 were more like ChatGPT, Dave would’ve been less “I’m afraid I can’t do that” and more “Did you consider this PHP function, Dave? And, also, maybe let’s tweak that CSS for a snazzier homepage, yes, shall we?

In wrapping up this digital diary, META-STRAND’s collaboration with chatbots like ChatGPT and Claude, ocassionally also Bard and the chat function in Bing, isn’t just about getting things done efficiently. It’s recognizing where technology can help me with my incompetencies, how it can take my raw thoughts and mold them into clearer expressions. It’s the practical choice of using the best tools at hand to build our narrative in this digital age.