Abhijit and I were in the same first year class at Harvard, and I have two especially strong memories of him from that time.
First, he was always willing to help out those who were not as advanced in the class work as he was. Furthermore, that was literally everyone else. He was very generous with his time.
Second, when it came to the first-year Macro final (I don’t mean the comprehensive exams), Andy Abel wrote a problem with dynamic programming, which was Andy’s main research area at the time. Abhijit showed that the supposed correct answer was in fact wrong, that the equilibrium upon testing was degenerate, and he re-solved the problem correctly, finding some multiple equilibria if I recall correctly, all more than what Abel had seen and Abel wrote the problem. Abhijit got an A+ (Abel, to his credit, was not shy about reporting this).
One of my favorite Abhijit papers is “On Frequent Flyer Programs and other Loyalty-Inducing Economic Arrangements,” with Larry Summers. I believe it was published QJE 1987, but somehow the jstor link does not show up from google searches. This was one of the first papers to show how consumer loyalty programs could segment the market and have collusive effects.
Another favorite Abjihit paper of mine is his job market paper, “The Economics of Rumours,” later published in ReStud 1993. Have you ever wondered “if this rumor is true, why haven’t I heard it before?” Abhijit works through the logic of the model on that one, in a scintillating performance. It turns out this paper is now highly relevant for analyzing information transmission through social media.
Abhijit is the clearest case I know of a brilliant theorist who decided the future was with empirical work — he was right. Nonetheless his early theory papers are still worthy of attention. When Abhijit went on the job market, his letter writers suggested he might someday win a Nobel Prize, so strong were his talents. They were right, but I suspect they had no idea for what the prize in fact would turn out to be.