Where are all the banger blog posts?

You can see the impact of LLMs everywhere except the banger blog post statistics

Dec 20, 2024

Every day, an army of thought leaders espouse how LLMs are changing their life and disrupting everything. Unfortunately for us appreciators of the humble blog posts, this hasn’t meant more banger blog posts.

I would consider a banger a creation that is particularly viral, unique, engaging, and novel. Neither these AI thought leaders nor the general banger writing population have become insanely prolific thanks to LLMs. They haven’t turned everyone into Michael Crichton:

Michael Crichton spent four years at Harvard Medical School, and what does he have to show for it? Seven novels, a movie, several more manuscripts and screenplays and a fast‐paced, provocative book of non‐fiction on the state of the American hospital, which was published last week.

On the other hand, LLMs have both:

Made it significantly easier to produce a large number of words about a subject quickly.
Increased the rate of bangers in other areas, like software development.

You would think writing, which is core to what LLMs do, would be impacted more. Why hasn’t it?

LLMs are having a huge impact on coding

This lack of written bangers reveals itself when compared to coding. There are many more people saying “I built this in 3 hours” than “I wrote this in 3 hours.” Some examples:

The world's shortest hackathon (2 hours) hosted by Vercel and Nvidia
Everything I built with Claude Artifacts this week
Recreating complex canvas animations in 15 minutes.
Pixel perfectly rebuilding a UI that 800+ engineers couldn’t nail.
People recreating entire UIs and apps with v0, Bolt.new, Claude Artifacts, aider.chat.

This personal site that looks like a receipt is a perfect example of a banger coding project. Its viral, unique, engaging, and novel. Thanks to LLMs, it was built in a few hours.

Beyond bangers, millions of people have used LLMs to become better developers. GitHub found that developers complete task 55% faster when they use LLMs. This means a lot more code and projects being written and completed, and a lot more potential for bangers.

Why aren’t LLMs increasing the number of banger posts?

When compared with coding, writing just hasn’t kept up. There’s many potential reasons why:

1. Writing is sand, code is Lego

Writing is like sand. It starts as nothing and will crumble without support and careful planning. Code is like Lego, it gives you the structure and pieces you need to make your vision a reality.

Because of the structure and reusability of code, your vision of a banger is a lot closer to reality for coding projects than it is for writing ones. The LLMs has a better understanding of the pieces available to make that vision real. Reusing writing to the same degree would be considered plagiarism.

The reusability and structure also means there is less to go wrong. Although the quality of code may vary as much for code as it does for writing, end users don’t see the code, they only see the experience it produces. For writing, the words are the experience. Poor quality is much more obvious.

Ultimately, LLMs can produce both passable writing and code, but passable code is much more likely to create a banger experience than passable writing.

2. Writers haven’t figured out how to use LLMs

There is a cultural layer of using LLMs writers haven’t figured out. There are plenty of good blog posts about how developers are using LLMs, but none (I can remember) about writers using them. None of the writers I admire have written about how it has changed the way they do things.

For my workflow, LLMs basically function as a superpowered Google search. I have a question or an idea I want to explore and use the LLM to explore it. Going further, I find they work well to interact and ask questions of documents (or code) using projects. It has made me a better writer, but it hasn’t helped me produce 10x more bangers.

This is similar to someone I consider to be a counter point to this whole post: Dwarkesh Podcast.1 He leverages LLMs heavily and built one of the best and most important podcasts that churns out banger after banger.

While asking Claude questions, Dwarkesh uses the LLM as a sounding board to validate and refine his grasp of his guest’s field of expertise. This back-and-forth with Claude helps him discern if he’s “found a blind spot in their thinking” or if he’s simply “confused by their ideas.”

Why haven’t we heard similar things from writers? Probably because the truly LLM-empowered writer doesn’t exist yet.

3. The tools aren’t there yet

Tools often act as productized versions of workflows people have figured out. Because writers haven’t fully figured out how to use LLMs to produce great writing, there’s little for LLMs to productize.

There are many tools for writers, like Type.ai and Notion AI, but these rarely are part of the workflow of banger producing writers. These tools seem to be used more for simple tasks like formatting, grammar-checking, and synonyms. At best, most writers seem to just use ChatGPT and Claude.

Developers are constantly shaping their tools to improve their effectiveness. This is something we’ve seen yet again with all of the LLM-powered, coding-related products. LLMs quickly became integrated with the products, like IDEs, they were already using. Unfortunately, this is not the same (yet) for writers.

As proof of this, look at the testimonials of writing and coding related tools. I rarely see praise for writing related tools on social, and the testimonials on their site are from people and companies I’ve never heard of. A tool like Cursor, on the other hand, regularly gets praised for both helping non-technical people code and seasoned engineers be more productive. Their site is filled with people and companies I’ve actually heard of.

4. LLMs make you smarter but that doesn’t make you a better writer

LLMs can teach you anything. They are like a tutor with infinite patience that can help you understand hard concepts, ideas, books, languages, and innovations. Knowing these things does not necessarily translate into your ability to write about them well.

Have you ever read an academic paper? These are written by the smartest people in the world about the most important ideas in their lives. Theoretically, these should be bangers, but a vast majority are boring, dense, and basically unreadable unless it is your job. This shows that being smart is not all matters for banger production.

The idea that getting smarter does not lead to more bangers is also a bearish one for LLM progress. A vague “smartness” seems to be the main focus when it comes to LLMs. Sadly for me, there’s no LLM developer optimizing for a model that writes banger blog posts.

Maybe Facebook, TikTok, or X (sigh) will figure this out. After all, creating an algorithm that optimizes for engagement is core to each of their businesses. They are in a position to combine this engagement focus with LLMs to create my imagined banger blog post wireheading machine.

5. Distribution is ruined (and so is my banger radar)

The last potential answer is that the world is producing more banger blog posts, I’m just not seeing them. This could be caused by LLMs or just be coincidental.

Twitter used to be the go-to place where I saw posts and could judge if they were bangers, but the algorithm changes have ruined this. X now downranks links, especially to Substack, where many bangers live (and has made a large impact to the banger blog post numbers 🙏). Maybe I’m nostalgic for a time where a blog post would take over Twitter for days at a time, but that’s not realistic anymore.

Even if Twitter’s algorithm didn’t change, everyone isn’t in one place anymore. They’re on Substack, Beehiiv, Ghost, Mastodon, Bluesky, Threads, TikTok, and many more. A banger doesn’t capture everyone’s attention like it used to.

The spirit of writing will always matter

Like Homer invoking the muse, there is some creative spirit of bangers that LLMs have not captured. The areas LLMs aren’t good at might reveal where this creative spirit lies: topic selection, audience understanding, context, sources, and structure. Avoiding being automated away means tapping into your creative spirit and improving your skills in these areas.

There might come a day where LLMs do manage to create banger after banger, but that will still require someone to curate and lead them (taste matters, as tech bros like to say). This same creative spirit will still matter in this future.