Diary Entry 1

A diary entry on research, ambition and the strange comfort of running one more experiment.

31 August 2026

I am twenty-five years old, and lately I have acquired the ridiculous conviction that I am late for my own life.

Late for what, precisely? This I cannot say. Nobody has given me an appointment. There is no letter on my desk saying: Mr. Vartak, you were expected to have become somebody by twenty-four. Nevertheless I feel it. I wake with the sensation that somewhere an invisible clock has been running without my permission, and that other men (often younger men, which is especially offensive) have understood its rules while I have been occupied with foolish things: degrees, papers, experiments, visas, applications, research projects that fail after three days, research projects that fail after three weeks, and occasionally projects which succeed just enough to make the next failure more painful.

The most absurd part is that, if someone were to describe my life to me as though it belonged to another person, I might think he was doing rather well. IIT Bombay, then Duke, research, publications, professors, America, Abu Dhabi, India again; there is movement in it, at least. It would sound like one of those biographies I torment myself by reading online. I might even envy him. Yet because the unfortunate man is myself, I know every defect in the construction. I know which paper was not really mine in the way I wanted it to be mine, which project went nowhere, which opportunity I nearly had, which field I entered too late, which paper I have not read, which researcher understood something at twenty-one that I am only now trying to understand at twenty-five. Biography is merciful because it leaves out Tuesdays.

This MATS application has become particularly grotesque. I am convinced I will be rejected. I say this repeatedly, with all the confidence of a prophet, and then spend the entire day trying to improve my application. What sort of behavior is this? A condemned prisoner polishing his shoes because he wishes to make a respectable appearance at the gallows. I read the work of previous applicants and immediately discover that they are all geniuses. Their ideas appear inevitable and mine accidental. They seem to have been born understanding mechanistic interpretability. I imagine one of them at fourteen opening a transformer and saying, “Ah, yes, naturally.” At fourteen I was presumably doing something much less useful.

And yet I continue.

This is perhaps the part of myself I understand least.

If I truly believed there were no hope, why do I run another experiment? Why do I fix the graph? Why do I wonder whether the framing could be cleaner? Why, after announcing to myself that the entire academic project is finished, do I open another paper twenty minutes later? Either I am dishonest about my despair or dishonest about my hope. Most likely I am both.

I have become very skilled at conducting trials against myself. There is always a prosecution and never a shortage of evidence. If I spend a week exploring, I am wasting time that should be spent publishing. If I force a publication, I am producing shallow work for credentials. If I read papers, I accuse myself of hiding in the literature instead of doing research. If I stop reading papers and build something, I immediately discover a paper from six months ago proving that my idea is obvious. If I apply for a PhD now, perhaps I am underprepared. If I wait another year, I will have wasted another year. If I remain in research, I may never become rich. If I leave research, perhaps I will spend the rest of my life wondering whether I might have become good at it.

This court in my head is very efficient because it permits contradictory charges.

I sometimes wonder whether I even want a PhD or whether I want the acceptance letter. There is an important difference, and I am not always noble enough to pretend otherwise. The letter would say something that the PhD itself could never guarantee. It would say: Yes, you were right to keep going. Yes, all of this amounted to something. Yes, somebody whose judgment you respect has looked at you and decided that you belong here. What an enormous thing to ask from a PDF attachment.

Perhaps this is why I keep searching for the right advisor with an almost romantic stupidity. I imagine that somewhere there exists a professor for whom my strange collection of interests will make immediate sense. He will see what I am trying to become before I have become it myself. We will work on exactly the right problems; I will become sharper; the projects will finally join together into something resembling a research identity. I know, of course, that this is absurd. Advisors are human beings. They miss emails. They have grants and students and bad moods. They do not stand on hilltops waiting to complete the intellectual destinies of anxious twenty-five-year-olds. Still, I cannot entirely kill the fantasy.

Then there is money, which intrudes upon all of this like an indecent relative at a funeral.

I want to do intellectually serious work, yes. I also want to be rich. There, I have written it plainly, and perhaps that is better. I want enough money that choices cease to feel like traps. I want to live where I choose. I do not want to spend my thirties calculating whether losing a job also means losing a country. I want to be able to walk away from people and institutions I dislike. I want comfort. I want freedom. And yes, God forgive me, I want the sports car too. One can construct elegant theories about “optionality,” but the Porsche remains stubbornly visible beneath them.

This complicates America. I loved something about living there, not every part, certainly, but the feeling that enormous things were being built everywhere and that perhaps one might join them. Yet America demands paperwork in exchange for dreams. There is something comical about being told: here are laboratories working on machines that may transform civilization; now please prove again that you are legally permitted to sit in North Carolina. I cannot bear the idea of spending decades under immigration anxiety. But then I imagine remaining in India and immediately worry that I have abandoned the center of the field. When my parents tell me India is perfectly fine, I become passionately American. When they tell me to apply abroad, I discover within myself a deep commitment to India.

I am, geopolitically, extremely easy to manipulate.

And sometimes, in moments of exhaustion, I decide I will abandon everything and start a business. Not even an AI company necessarily. A factory! Why not a factory? There is something enormously attractive about a factory when one has spent months thinking about language models. Steel enters. Something is manufactured. It leaves. One can count it. Nobody asks whether the causal interpretation is sufficiently justified. Reviewer Two cannot reject a shipment because “the novelty remains unclear.”

For an hour I become an industrialist. I see myself inspecting machines, discussing production numbers, making money, perhaps wearing a watch more expensive than necessary.

Then I wake the next morning and think: I wonder whether unfaithfulness has a linear direction in activation space.

And the factory disappears.

This is how I know I am in trouble.

The truth, which I resist because it deprives me of several excellent excuses, is that I actually like research. Not all of it. Certainly not the endless reading or the vague projects or the humiliating periods when nothing works. But I like the moment when a question becomes concrete. I like constructing a small experiment and waiting to see whether the machine behaves as I expected. I like discovering that I was wrong when the wrongness is interesting. I like the feeling of a project becoming clean enough that I can explain it in five sentences. Perhaps this is why the recent sprint projects have felt better. For a few days there is only one question. My entire future is temporarily forbidden from entering the room.

This appears to be good for me.

I have spent far too much of my life several years ahead of myself. I am twenty-five but am constantly negotiating with thirty-five. Thirty-five-year-old Rohit is an unbearable man. He has very strong opinions about what I should be doing. He wants money, citizenship, career satisfaction, intellectual achievement, a family perhaps, physical fitness, a house, and apparently several cars. He blames me personally for not having arranged all of this already. I have never met him, yet he dominates my schedule.

Meanwhile twenty-five-year-old Rohit has an experiment to run tomorrow.

This ought to matter more than it does.

I think what terrifies me most is not failure itself but the possibility of having wasted whatever ability I possess. If I try seriously and fail, perhaps I could endure it. But what if I chose the wrong field? What if I should have spent another year doing research? What if I should have gone directly into industry? What if I could have become an excellent researcher but became impatient? What if I spend years chasing a PhD and emerge poorer, older, trapped by immigration, and discover that the life I wanted was elsewhere? There are infinitely many versions of myself, and every decision murders all but one of them.

It is exhausting to feel responsible for the dead.

Perhaps that is why I crave institutions so much. An acceptance letter absolves you temporarily from choosing. Someone says: come here. Work on this. For four years the road narrows. This is called opportunity, and perhaps it is, but sometimes I suspect I also desire the relief of being told what to do.

There is a humiliation in admitting that.

Still, these last few weeks have taught me something, though I am reluctant to dignify it by calling it wisdom. I can finish small things. This sounds pitiful written down. After all these degrees and papers, my great revelation is that one may work on one project for a week. But perhaps the obvious truths are the ones a man spends longest avoiding. I do not need to understand the entire safety literature before touching a problem. I do not need to know my dissertation before applying for a PhD. I do not need to decide whether I will live in India at forty before running tomorrow’s experiment.

A strange thought has occurred to me: perhaps my life is not actually waiting to begin.

I dislike this thought.

It makes me responsible for the present.

It would be much more convenient if this period were merely preparatory, the unfortunate corridor between Duke and the PhD, between uncertainty and arrival. Then I could forgive myself for being miserable because the real rooms lie ahead. But what if there are only corridors? What if after the PhD there is the postdoc or job search, after that promotion, immigration, funding, papers, money, something else? What if every room turns out to contain another door?

Then postponing life would be a very serious mistake.

I keep thinking of Gusteau from Ratatouille, which is perhaps not the authority one expects to cite after IIT and Duke, but I have reached the stage of my intellectual development where a dead cartoon chef provides more useful career advice than most university websites. Gusteau’s creed was wonderfully simple: “Anyone can cook.” In my increasingly desperate translation: anyone can do research. I find this enormously comforting for approximately ten seconds.

Then, unfortunately, I remember Anton Ego.

Ego, being a critic and therefore professionally obligated to destroy whatever small happiness has survived until the final act, explains what Gusteau actually meant: “Not everyone can become a great artist; but a great artist can come from anywhere.” Applied to research, this is considerably less reassuring. Anyone may attempt research; this does not mean everyone will become a good researcher. It means only that a good researcher might emerge from anywhere, which still leaves unresolved the rather important administrative question of whether “anywhere” includes me.

This is where Ratatouille, ostensibly a children’s film about a rat cooking soup, begins to feel personally malicious.

Still, Ego offers one final consolation. Earlier in the same review he admits, in effect, that even an average piece of work brought into existence can carry more meaning than the criticism that merely pronounces it worthless. I cling to this because at present there is very little else available. A mediocre experiment I actually run has more value than the magnificent experiment I design in my head and never attempt. An ugly graph exists. A failed project contains information. A stupid question, pursued honestly, might at least give birth to a slightly less stupid question.

Meanwhile I can sit in judgment forever. I can explain with increasing sophistication why the project is naïve, why somebody better has probably already done it, why the result will not matter, why MATS will reject it, why a professor will not care, why the entire field will regard it with indifference. I can become an exceptionally perceptive critic of work that does not exist.

Perhaps that is what self-criticism becomes when cultivated too carefully: an intelligence whose principal achievement is preventing the production of evidence against its own pessimism.

And so, humiliatingly, my current philosophy of research has been assembled from a Pixar film about a French rat.

I have had worse advisors.

Tonight I am at home. The MATS deadline is near. PhD applications will follow. I have no idea where I will be a year from now. Perhaps I will be accepted somewhere excellent and laugh at how melodramatic I was. Perhaps I will be rejected everywhere again, and then I will discover that the tragedy I rehearsed so carefully was merely an inconvenience followed by breakfast. Perhaps I will leave academia. Perhaps I will find the advisor I keep imagining. Perhaps I will start the ridiculous factory after all.

I know nothing.

This, apparently, is the unbearable fact.

And yet today I worked. Tomorrow I can work again. There is another experiment I want to try. I have a few ideas that may be stupid, which is at least an improvement over having no ideas. I have discovered that projects can fail without requiring me to fail alongside them. These are small facts. They do not answer the questions I keep demanding that life answer.

Perhaps life has refused because the questions are badly formed.

I keep asking, Will I become successful? Will I become a great researcher? Will I become rich? Will I get the PhD? Where will I live? Was all this worth it?

And life, with extraordinary rudeness, answers:

Run the ablations now.

I suppose I will.