I used to believe I was the one on the planet who loved this field the most. Now I suspect I care the least. I cannot see the future, sprinting toward a horizon that keeps receding into darkness. Part of this text was written one year ago, originally meant to be the script for a video. I found it too clumsy and in constant need of revision, so I reformed it into an essay.
Before I begin, let me summarize a remark of one brilliant peer, that variance is enormous, and joining the frontier requires a otherworldy stainless résumé. It is never too much to repeat this. However, since anyone actually aiming for that perfection is unlikely to read a post like this, I only deal with the question how large the the gap/residual will be if you fall short.
The industry lacks foundations
You have probably heard of the cliché that academia is ‘out of touch’ with practice. For computer science, I would be better say much of practice is not based on sounded discipline. Looking back: even Dijkstra’s algebraic proofs remain a distant dream. The flaws in productions are daily fodder for public mockery, but how can we demand rigor from a production pipeline that is less disciplined than alchemy? That MIT retired 6.001 (SICP) was an admission that there are no choice but to conspire with the rules of launching rat race. Looking closer: many people are simply unequipped to think about cutting-edge problems. No comment is needed on how much sense AI hype-sters perceive for the theoretical big picture.1
I should have found a reference for this. I selected this only by its title, which is not a quality guarantee to the author Ed Zitron.
Hence the chronic disease of our field: rounded square résumés, glittering portfolios, and interview questions that might as well be stand-up comedy. I strongly recommend the Undergraduate Self-Help Guide from SUFE, for we are converging with business studies, unless you arrive with Fields-medal-level mathematical force. The barrier erected by school rankings is more and more insurmountable, and that is only the tiniest symptom of the convergence.
The public panic of professional moat as the bedrock of personal competitiveness misses the larger picture. A discipline without foundations does not just lack a moat. Its bigger problem is that aiming for the middle lands you at the bottom. Conformism everywhere, and no place for sublimity. The suspicious on technical salons yester-year is glue sustaining even-swelling clusters today. For many, ‹two bugs make a right› is the goal. Grinding for GPA or flaunting ego are merely corollaries of lacking goal.
Contrary to the popular claim that self-studying computer science is impossible, the myth is real. However, that is acquired merely by reading profound-sounding materials is all but serious. Even worse is the invented loftiness: mimeographed canonical books, or cult of an overnight genius. If you are unlucky enough to take a course built on cult, exercise your own judgment.
Full disclosure: the above is the zeal-of-a-convert view from someone lacking training in mathematics, natural science, or economics. I hope the reader is better equipped, yet this field is full of people like me, and there is no escape hatch.
Prometheus and clay
Of course, the joy of making things is real. Perhaps you still remember the first time a terminal lit up under your fingers, when you felt like the hero of a tiny universe. Eventually we discover we are only rivets, but even a rivet must once have played emperor. Let me correct the earlier tilt toward theory: creation is equally important, which demands multidisciplinary skills that formal education rarely provides. I doubt, however, whether ‹multidisciplinary ability› is a feasible objective function. There is an NFL-theorem feel to it: one can even construct a task for which the dancer-engineer outperforms the non-dancer.
In case the scale is not balanced enough, I would argue that though no math-y things are wrong, not all theories are useful. I have read an epigram comment saying ‹some CS students are baffled by category theory, others by pointers›. The original words might be homotopy type theory, but I question this recollection much for reputation of the commentator.
Time is the first and final resource determining what you can actually build. The four undergraduate years are probably the last chance in your life for free exploration, but that freedom must be rationed by you alone. You might be the next Bohemian-dance virtuoso, but a few short years may not suffice to fuse computer science with Bohemian dance. I have a good deal of sympathy for people like this. A decade ago seniors still had time to roam the ‘pan-computer’ playground, while in the ‘thoroughly collapsed’ era, the curtain has fallen and miracles are not allowed. One can only forget the vastness of the sky, picking a single pursuit and hoping for the best. I know no much about the status quo: only one year ago, when I started drafting this post, Copilot-generated freaks had not flooded ranking lists or even hackathons.
For concrete advice on how to create, see ByVoid’s slides.
Work or any act of creation derives its value from what it offers others, not from private whim. Your pet project may be a beautiful hack or a mere trinket, but Spring Cloud instances from all over the world are still emitting steaming ProtoBuf packages.
A dream
When all roads are closed, dream one more dream. While procrastinating on my bachelor’s thesis, I once stumbled upon a blog post by Stephen Wolfram on the centennial evolution of combinators (I failed to find it now), and in a flash I realized: this is the theory I had been yearning for. Sometimes I wish a few more Wolframs would appear to rouse the field, and I even allow myself to imagine that one of them might be reading this now. Alas, reality is not a heroic novel. Only a handful proudly declare themselves future Turing laureates, and they are precisely the ones who would not have opened this essay. Wolfram, after all, came from physics.
I have avoided mentioning social system. Neither a rebel nor a sharp critic, I am still undeniably its casualty. I possess no lifeline — nothing brighter than background radiation. I mean: a friend whose grasp of mathematics goes beyond rote memorization of analysis and linear algebra, a circle of contest hunters, a NAS filled with books where there are no disproportionate computer scientist-to-philosopher ratio, or simply the innate gift of dancing under sunlight. Background radiation is the system. Make better choices than me.
I know I have failed this discipline. Love does not outlast the erosion of years. At my level, the real thing never comes within reach. I hope that when you discover how dearly transcendence costs, you will not abandon the pursuit. This is not the platitude of purists, but a plea that, on this bleak road, you keep searching for an answer once thought unimaginable.
The text was originally in Mandarin, which is written without LLMs. But the current version is malodorous. I will be more cautious in drafting and translation.