Keeping up

I checked today and it turns out, I’ve read more than 300 AI research papers in the past year (GitHub repo).

Back in college, I suspected that being in software would mean constantly updating yourself. That has proven true over the years, but AI has taken it to a whole new level.

As a leader, being effective at this moment requires having a sense of where things are headed (my reasons why). Reading papers (along with working on hacks and experiments) has been my way of staying informed. Finding time hasn’t been easy — I traded Netflix/Hulu watching for research reading sessions. But it’s been worth it.

On several occasions, I’ve been able to point my team to relevant papers that helped steer their efforts more productively. Some papers prompted ideas for hackathon projects, while others sparked concepts that eventually led to patents.

In moments of rapid change, referencing specific, actionable insights is far more effective than relying on time-tested, generic advice like “collaborate” or “focus on fundamentals.” Disruption often rewards those who prioritize value creation over perfection—”Move fast with stable infrastructure” was “Move fast and break things” during Facebook’s hyper-disruptive phase, battling MySpace, Friendster, and others.

Yet, engaging deeply with emerging trends can be challenging. Some leaders hesitate, opting to wait or take a cautious, overly critical stance without fully exploring the potential. That mindset risks missed opportunities and eventual disruption, especially in the current AI wave, where models’ code-authoring capabilities are accelerating AI adoption itself.

As an aside, I suspect this self-perpetuating dynamic will accelerate further with the rise of coding agents like Devin, as the role of GenAI shifts from assisting humans as copilots to autonomously raising PRs for human review and merging. The net output of the same human team will likely increase as coding agents get more adoption.

I organized these papers into a word cloud (see header image above), and as expected, “Reasoning” emerged as my main area of interest. It still amazes me that something originally designed for next-word prediction is now demonstrating complex reasoning capabilities.

With so much happening so quickly, the question is…

How to Keep Up?

Unlike previous technology waves, where you had time to wait for books to be written and tech to be adopted, that’s simply not the case with AI. New breakthroughs are emerging every week, and it can seem overwhelming. Here are some ways to keep up:

1. Subscribe to the Hugging Face Daily Papers Newsletter: https://huggingface.co/papers.
2. Follow Specific Accounts on X: Follow people who post about AI papers, such as Rohan Paul (https://x.com/rohanpaul_ai) and AK (https://x.com/_akhaliq).
3. Use AI to Make Sense of AI: This is the big one! Leverage AI-powered tools to summarize and analyze papers efficiently.

But wait, AI hallucinates! — you might say. The key with #3 is to realize that, with citations, AI’s occasional hallucinations are easy to check. Far more often, it explains concepts clearly and makes them easier to grasp — tirelessly, like the best and most patient tutor you’ve ever had. I’m not asking you to believe me; instead, I’m asking to try it out for yourself. Andrej Karpathy realized the same thing:

A simple setup:

The AI journey is intense but fun.