I have recently re-engineered my whole blog. The whole engineering stack was moved from Hugo to NextJS, while maintaining all the looks and functionalities (not many since it’s a frontend-only website). Claude Code was very helpful for this; it’s quite surprising to see it was done in like one hour, about 3.6k lines of code in 65 files, working one shot. I wonder what we can do now as a society, now that software is so cheap to produce, at least the small ones. I would say this is also a reason to start to publish a little bit more on this empty desert blog. I am still not sure about the rates of publishing, but at least some times every while, like once a week as a lower bound perhaps. I should have also now differentiated the archives in a way such that notes and posts like this are now separated, and users can more easily look around and explore my thoughts, or the evolution of my thoughts, as I come to discover more and more about this world.
One commitment I can make here is not to use LLMs to write whole paragraphs at all. I think my commitment is mostly to limit them as grammar correctors, or oracles to discuss with. Let’s start with some of today’s discussions, then.
I think in the past few years, the fear of contributing to the bad quality of the content online has prevented me from posting a lot here. I think now I can separate with tags the good ones with the ones like the following, where I just have some random thoughts about some topics, so that the reader can have expectations on what he/she/they read here. Let’s start with today’s random thoughts, then.

The Tragedy of The Academia Commons

The current world is very much focused on results. This is justified since usually actions, and eventually the results that these actions produce, are what indeed produce general benefits to one’s society, and on the side, to one’s self. Yet in the academic field, the mere concept of result is a little more difficult to conceive compared to the raw physical results of the production of the industry or the work of the soil. While the soil’s production is physical, tangible, the work of academia is mostly virtual, in the concepts and ideas that are put together.
Particularly in the CS world, my world, the result of the academic work also has some almost tangible production. Often, it is some new technique, new architecture, or new framework that allowed that lab to achieve something that was not possible before, or at least, it was not known by the majority of the readers engaging with that specific problem. Some other times it is a conceptual shift that then allows you to explore some other directions, and very rarely it is some paper that closes an open problem forever. One piece of advice from Professor Lorenzo Alvisi that I personally received and won’t forget is the following: there are only three kinds of papers worth publishing: the one that opens a new field, the best one for the current SOTA, and the paper that ends a field. This has shaped a lot of my research taste, and the presupposition of the result of impact is quite clear.
Yet, the topic I wanted to put the marker on today is another. It’s the tragedy of the commons, where everybody is mostly just pushed to publish, without engaging in the real result of their work, in terms of impact. Claude told me this is a phenomenon that is already studied (e.g. Andrew Abbott’s Chaos of Disciplines). If everybody just publishes without any concrete novel new impactful results, or if the result is mostly just framing, or agenda setting, with more and more researchers, everybody would mostly just get into building their own new field, with their own community, while in reality, everybody is doing something very similar to one another conceptually. This is some ugly future. Hopefully, the institutions that we have know about this problem, and filter out works that are actually producing meaningful advancements not only in their newly established very interesting new field, but also on a larger scale across the sciences. Now, LLMs provide some quite interesting takes on cross-contamination across fields, since everybody has an expert-sounding tool that is able to summarise the consensus across different subfields.
Reading Binmore 1994 more broadly, it seems the only way to play this game fairly is to establish a social contract that makes not-resourceful publication not an equilibrium of the game of morals we are playing now. In another sense, everything that anybody publishes should have some standards that each researcher should have self-verified and checked, before even attempting to publish: for example, novelty, clarity of the writing, and relation to the state of the art. What we need is just the basics to not flood the public good of shared open research with plain trash. Just some personal professional ethics to prevent such things, or not contribute to making this problem worse than it might be. This also echoes my recent paper on prosociality and rules: Huang et al. 2026.

References

[1] Binmore “Game Theory and the Social Contract: Just Playing” MIT Press 1994

[2] Huang et al. “Mechanism Design Is Not Enough: Prosocial Agents for Cooperative AI” 2026