Outsourced Intelligence Era
I recently graduated with a Masters in Artificial Intelligence and not sure how to feel about it. On the brighter side I deliberately put a lot of efforts to avoid the use of AI as much as possible and force myself to learn and on the other hand there is nothing I couldn’t have done using AI to pass this course. In fact many people just breezed through the course using AI to write assignments which will eventually will be *cough cough assessed and graded by AI. I’m not pointing fingers at any particular organizations or individuals, this is just the sad reality that we are living in right now. At some point I was beginnning to feel like an AI doomer because I did not like the amount of spoon feeding it did for me. Yes I do need help but just point me in the right direction, nobody asked you to solve my problems for me! But sadly it is the future we are going towards because programmers have this compulsive need for automating everything, including themselves.
There is no longer an uncertainty for using AI at work. An overwhelming amount of AI companies or companies that are trying to re-brand as AI companies are popping up every single hour. We are standing at the new frontier of software and engineering and the old institutions are falling and that too glourously. The world or engineers and specialist is coming to an end. Only architects will live (for now). Most startups these days are AI-first whether it is the product or tools used to develop the product. Every single day there is a new AI product automating manual work and to some extent it feels like the same wave as web 1.0 and web 2.0. After the last stronghold, Linus Torvalds, fell into the jaws of LLM coding, it makes a compelling argument as a new paradigm in itself.
Claude code came out an year ago. And it has almost made writing code by hand obsolete. What started as vibe coding has turned into neatly structured high performance output coding which is almost as intuitive and 100x better than what humans were used to producing after hours of focused cognitive efforts. While a lot of companies are including coding agents in their internal dev tooling, some like stripe are looking to replace programmers with architects that only co-author and accept changes from LLMs. When half of the population is rooting for AI’s downfall because it replaces most available jobs, the other majority is fearing it’s absolutness. All the generated content images, text, code has a certain level of skepticism associated with it. While text and images can be objectively broken down, code simply cannot be. Code contains complex logical patterns that are not obvious just by taking a glance at them. It requires years of understanding and muscle memory to figure it out. Human acceptance of LLMs incompetancy is the only barrier that is holding “vibe coding” from becoming the default. We have gone from scarcity of information to scarcity of trust, but that seems to be dissolving little by little, day by day.
Software Engineer is not the only sector that is on the path of being automated. Most other low effort, high friction professions are also being replaced by agents or agentic workflows. My guess is we will see a vastly different world and job landscape in 5 years time, how far we are from that future only time will tell. This is the new kind of industrial revolution which makes a large portion of mental labour obsolete, whether it is just a layer of abstraction on existing world or not, I leave it for you to decide. As of now human consiousness wins in limited ways, but will AI help us understand it better or surpass it? The last year has been a crazy ride with technology shifting from just LLMs in web chat interface to Agents, MCP and now Skills.