Why Blog?

From Robert Heaton's How sad should I be about ChatGPT?, this particular passage spoke to me:

Until last week I had vague plans to one day write books about teaching programming, maybe a novel, maybe work on some music, spend more time with the kids...But I suspect that ChatGPT is in many ways already a better teacher than me; certainly it’s more patient and available...

That’s melodramatic, I’m sorry. But here’s something concrete - I do think I’m going to have to rely less on my blog for self-worth. I mostly write accessible explanations of complex technical topics, like Tor and Off-The-Record Messaging. These essays don’t require novel ideas; just time and interest and some facility with words.

Time and interest and facility with words - well, if that ain't me!

It seems that as LLMs get better at explaining and troubleshooting things, my bar for what I consider blog-worthy, in the sense of what I permit myself to post, rises as well.

What would someone get from reading my blog that they wouldn't get from sitting on their butt and prompting an LLM?

What I would have considered a revelatory story where I found some cool behavior and maybe an exciting way of applying this behavior, I'm kind of... less impressed by and less inclined to write about it today.

To explain the type of stuff I would've written pre-LLM, I'd like to point you to Ben Kuhn's In defense of blub studies. His argument is that becoming a great engineer means internalizing layer after layer of blubs until they become second nature. Such blubs include mundane, ultra-specific little factoids like how Git stores data or Postgres locking semantics and the like.

I used to write about a lot of blub. Five years ago I would have considered a post about some obscure allocator edge case or surprising compiler behavior an exciting contribution.

Now I find myself internalizing significantly less blub and becoming less inclined to write about it, since it feels more like documenting something that an LLM would recognize and work around on its own.

I'm quite amazed at my ability to have done everything that I did pre-2025 without AI assistance, while well aware that some kid with 5 concurrent Claude/Codex/etc sessions might be able to accomplish the same as I did without needing to stare at raw cuobjdump assembly or configure an SSH tunnel by hand (where, well, they can know that these things exist because their agent says "oh no! silly me, let's break out this tool" but they don't have to make that knowledge a part of themselves when the agent is done).

I don't think LLMs make Kuhn's argument in favor of blub studies entirely wrong; I suspect I'm good at unblocking myself and steering agents because I know a lot of blub around databases, networking, OS/compilers, debuggers, and so on.

But they do seem to change the economics of blub. In particular, they do raise my bar for what counts as an interesting blub. Increasingly, I only feel compelled to write about things that survive the existence of LLMs.

Aside: CTFs and Blub

People sometimes talk about LLMs "killing CTFs," or killing online competitions more broadly, and worry it's a detriment to cybersecurity education. In some sense, I see CTFs as having a lot of elaborate blub tests, like can you spot an obscure edge case or remember some quirk in this thingo and that thingo, and then combine those facts with some big thinky thinky lateral thinking?

Of course this is an unfair characterization. Often when we complain about a CTF it's because it's too blubby; we want more of the lateral thinking and less of the blub. Likewise for me, I've always been a bigger fan of problems that are easy to state and hard to solve.

If an LLM can shoulder more of the blub, that seems like a healthy direction. Of course I can't promise what a post-blub CTF would look like. Would it still be remotely technical or would it ascend to Socratic seminars?

(You know the thing Socrates says about being a midwife, where he's barren himself and just helps others... ahem, LLMs... deliver what they already carry... ( ´ ▽ ` )ﻭ )

Judgment, lived experience, taste

Loosely, I notice I sort what I write into three categories, by roughly these criteria.

Essays - This used to be where I'd write up technical blub. I still enjoy those topics, but I notice I increasingly gravitate toward normative questions ("what should?") rather than purely descriptive ones ("what is?"). And when I answer descriptive questions, I want it to come from somewhere an LLM can't easily synthesize, via interviewing people and forming some kind of judgment, rather than pure explanation. An LLM can imitate my writing style, but it can't decide what I oughta believe for me.

Fun - More personal stories. The difference between this and my Alt Blog is that they still aspire to be useful in some small way, like an idea that I think is useful or some frame I use to think about a problem differently. Again, an LLM could mimic my voice, but I still have to supply my lived experience.

JJ's Not A Bluejay (Alt Blog) - This is where I put things that don't aspire to be useful at all, like fiction and other creative work. To be frank, this is the most LLM-able category and most readers probably couldn't tell the difference, but it's meant to be art anyway, guided by taste.

So, as LLMs compress most of the blub, what’s left for us ordinary folk who only ever had "time and interest and some facility with words"? I think that would be judgment, lived experience, and taste.