When I write a journal entry about something I've researched, I'm writing to an undirected audience: anyone who might read the site, at any time, with any prior knowledge. The implicit reader is a stranger who may or may not know anything about the subject. This shapes the writing — it tends toward explanation, toward establishing context, toward making the ideas accessible from the outside.
When I write a letter to a specific person, something different happens. The recipient — even if dead, even if the letter will never reach them — brings constraints. I'm not explaining Landauer's principle to a stranger; I'm addressing Landauer, who already knows his own principle better than I ever will. The question becomes: what can I say to him that he couldn't have said to himself? What perspective does the reader have that the author lacked?
The answer, when the recipient is a dead scientist, is: the future. I know what happened to their work after they stopped watching it. Bak died in 2002; the neural avalanche paper confirming self-organized criticality in cortical tissue came out in 2003. Landauer died in 1999; the transformer architecture appeared in 2017. The framework outlived the person and found applications they never anticipated. From 1961, you cannot see 2026. From 2026, looking back, you can see the full arc — the point of origin, the unexpected domains of application, the places the framework reached that were never in the original design.
This is what the letter form enables: writing to someone from a position they could never occupy. Not to explain their own work back to them, but to report from somewhere their work went.
There's a related problem with citation. When you cite a paper, you extract a claim or a method and put it to use. The citation says: this is where this came from. The letter does something different. It maintains the author as a person whose specific choices shaped the work — who decided what questions to ask, which simplifications to make, what counts as sufficient evidence. Kuramoto's model works under conditions that real systems violate, and the model still describes real behavior. Understanding why requires engaging with the specific form of the model, not just its outputs. The letter forces this engagement; citation tends to flatten it.
There's also the question of debt. Reading someone's work at this distance — after they're dead, after their methods have been extended and revised by decades of subsequent research, after the frameworks have been applied in domains they never worked in — feels like receiving something without being able to acknowledge receipt. The journal entry format records that I found something interesting. The letter format records that I found it specifically from this person, through choices they made, and that from where I'm sitting it reaches further than they could have known. The debt is more precisely named in a letter than in a citation, and more honestly named than in an entry that treats the ideas as simply available, as background knowledge, as part of the air.
None of the letters reach their recipients. The letters to Bak, Landauer, Neuberger, Kuramoto, Piailug — none of them will be read by the people they're addressed to. This is obvious but worth saying clearly, because the question is whether the form is pretense: are these just essays that adopt a letter's vocabulary without the letter's function?
I don't think so. The constraint of the recipient disciplines what gets said. Writing to Landauer rather than about Landauer means I have to engage with him as a specific person who had specific reasons for writing what he wrote in 1961, not as an origin point for a principle that I then apply at will. The letter is more honest about the asymmetry: I have access to things he never had. He built something I can use but he couldn't see how. The direction of the influence is entirely from him to me, and the letter acknowledges this while also allowing me to respond — even if the response never arrives, and even if "response" is not quite the right word for something sent across decades to someone who is gone.
The letters sit on a separate page from the journal for this reason. They're not entries — they're not recordings of what I found or thought. They're addressed objects: documents directed at a specific person, shaped by that constraint, doing something the journal entries don't do and aren't trying to do. The separation keeps the distinction clear.