The excitement around the AI agent Claude Code is readily apparent. However, figuring out its practical applications proves to be a greater challenge.
One can easily get caught up in the widespread buzz surrounding Anthropic’s coding assistant, Claude Code. This excitement surged during the festive season, echoing the debut of ChatGPT three years prior within technology communities. Many view Claude Code, already dubbed “the new ChatGPT” and “the demise of SaaS,” as a significant leap forward in AI innovation and application.
For those who, like myself, spend considerable time online, Claude Code has recently become a prevalent topic across social media feeds. While algorithms naturally amplify content matching my interests, platforms like LinkedIn and TikTok are inundated with discussions praising Claude Code’s capabilities. This includes AI trendsetters, everyday programmers, inquisitive IT specialists, and, predictably, a continuous flow of less prominent figures aiming to capitalize on the widespread attention.
Users are innovatively developing everything from full websites to small custom applications and discovering novel approaches for their operational processes. Ethan Mollick, a Wharton School professor at the University of Pennsylvania, documented his experience having Claude Code construct an entire startup. Elsewhere, discussions revolve around Gas Town, an orchestration system enabling 20 to 30 agents to operate concurrently. Reports from San Francisco even suggest individuals are allowing “Claude swarms to manage their daily routines.”
Just recently, Clawdbot (now rebranded as Moltbot) likewise joined the wave of excitement; it’s an AI assistant similar to Claude Code, offering to manage your entire digital presence—should you choose to grant it such extensive access.
The entire phenomenon is captivating and frequently astonishing. Coding is widely regarded as the principal application for generative AI (genAI), and witnessing Claude Code’s capabilities makes this clear. Despite multiple recommendations that I “absolutely must” try it, I personally haven’t experimented with Claude Code yet. I’m not especially eager to (though I will likely explore it eventually for learning purposes). Allow me to elaborate on my perspective.
To begin, I am not a programmer, nor am I capable of writing code. While Claude Code’s premise is to generate code on your behalf, the reality isn’t so straightforward. The initial hurdle remains too significant for me, and accessing its more advanced features requires considerable effort. Although I possess a foundational grasp of software development processes and functionalities, my expertise is quite limited.
Anthropic has started tackling this issue with the introduction of Claude Cowork, presented as a version of Claude Code tailored for non-developers. This sounds promising. Yet, my concern is this: the tasks it’s marketed for and the way it’s promoted just seem…uninspiring. Typically, one of the initial examples given for such tools is their ability to “organize files on your computer.” While potentially useful as a one-off task for those with many unorganized files, it barely justifies investing in and learning a new piece of software.
Observing how proponents utilize Claude Code reveals a strong emphasis on personal productivity and “optimization,” both professionally and in daily routines. This particular aspect holds no interest for me whatsoever. Indeed, I too encountered “Getting Things Done” and “The 4-Hour Workweek” two decades ago, but my primary takeaway was that delving too deeply into such self-improvement methodologies increases the likelihood of dedicating excessive time to identifying new optimization strategies. (It’s also disheartening, yet perhaps indicative of contemporary work culture, that individuals resort to “hacks” merely to perform their responsibilities.)
My primary indifference towards Claude Code stems from a missing “software perspective.” This insight struck me recently after reading tech writer Jasmine Sun’s account of her time with Claude Code. She illustrates how parkour practitioners gradually perceive a city differently from others, cultivating a “parkour perspective” where walls and staircases transform into distinct elements. Similarly, she proposes that developers cultivate a software mindset, enabling them to interpret and resolve all challenges through software solutions.
A typical user such as myself doesn’t encounter software-specific issues. I don’t instinctively recognize opportunities for automation or optimization by a bot in my daily tasks. Consequently, when informed that Claude Code can resolve my problems, I struggle to identify any.
This viewpoint, I believe, offers significant insight into AI adoption broadly and is worth considering. It applies equally to enthusiastic early adopters promoting new solutions to those perceived as “falling behind,” and to corporations eagerly pushing their employees to adopt AI. I doubt anyone inherently opposes a new tool that enhances productivity and simplifies tasks, provided they understand its intended application.
Perhaps corporate AI training initiatives should prioritize cultivating the ability to pinpoint issues that AI can address, and possibly even foster a software-oriented mindset, rather than solely emphasizing proficiency with particular tools.
This article originates from CS Veckobrev, a personalized newsletter delivering reading suggestions, useful links, and insightful analyses straight from editor-in-chief Marcus Jerräng’s office every Friday. Subscribe for complimentary access here.
