Post

daniel r(ooooooh) in nyc 🇺🇸

please do not use llms to actually write your blogs. we can absolutely tell. and it means: 1. I don't trust what you're saying (or at least I don't trust any purported 'insight') 2. I don't feel like I'm listening to you (normally I _like_ listening to nuance) 👉 shorter, imperfect yet real👌

5/10/25, 2:17 pm13 reposts196 likes

But what are you going to do when you’re not going to be able to tell?

daniel r(ooooooh) in nyc 🇺🇸

I think, like CGI in films, that kind of writing will age poorly ... but my point isn't about 'purity' – if you can write well with the aid of LLMs, then do so! but most people, currently, cannot.

As a former animator for both Disney and Dreamworks who went through the transition from hand-drawn to CGI, I can tell you it’s 100% the person behind the machine. 3D animators eventually got so good that there was stuff they could do that 2D couldn’t even dream of. Their CGI outperformed us 2D…

Also there are some people who have trouble expressing themselves through writing. AI gives them the tools they didn’t have before. Most 3D animators can’t draw to save their lives—us 2D anims thought that would be the “moat” — it wasn’t! (Thankfully btw).

@oetzi.bsky.social

Is using an LLM for a prompt like: "Does this blog have any questions that I haven't thought of yet?" good or would that fall under "do not use"? I use it in order to give more context, because I sometimes forget to.

daniel r(ooooooh) in nyc 🇺🇸

oh, there are lots of good uses for them in the process of writing a blog. ... I just wouldn't use one to actually produce the words. (and I'd be very careful before I let it do too much in shaping the point I'm trying to make.)

Raphaël 👨🏻‍💻

@raphaelbronsveld.com

It definitely helps me with the flow of sentences.

this is why I make sure to keep typos in my articles

got it. gonna use a SLM. 🫡

daniel r(ooooooh) in nyc 🇺🇸

😆 'make this shorter, clearer + more uniquely me'

midwest emo fan

i tried this once and i hated it. it also got way too much attention (or at least more than it definitely deserved) for some reason 🤔

I agree, you can see that article is written using LLM in no more that first couple sentences. For me it's instant "no" and I close the article unless it provides an actual value on the first glance

That's a great point Daniel and you're right to make it.

daniel r(ooooooh) in nyc 🇺🇸

👀

daniel r(ooooooh) in nyc 🇺🇸

... are you trying to say that ... I'm absolutely right?

Aaaaaallan Deutsch

@allandeutsch.com

You’re absolutely right!

@kathytafel.bsky.social

Haha— you caught me!

The word Daniel has 5 "r"s in it just like strawberry

~venir 🇪🇺🇺🇦

... please tell me you haven't written this post with an LLM 😂

daniel r(ooooooh) in nyc 🇺🇸

please tell me I don't sound like an LLM 😱

Shawn Wildermuth

@wildermuth.com

Here are a few sarcastic comebacks you could drop under that tweet, depending on how sharp you want to go: "Don’t worry, I only use an LLM to ghostwrite my tweets. The blogs are all written by my dog." "Got it. I’ll make sure to add more typos and rambling tangents so you know it’s real." ...

@jaycooperbell.dev

The biggest benefit I have found with blog writing and using them is to aggregate all the data points/info about the subject I'm writing about across all our different sources (since I'm writing about Trellis usually, that's our repos, commits, PRs, Linear, Drive, docs, etc).

@jaycooperbell.dev

It helps me formulate all the things I want to touch on, since it is expansive across many sources and months and I don't want to miss anything. Even though I know it's fine if something gets missed and it's good I'm writing it down in general. Have to remember it's ok to ship before perfection.

daniel r(ooooooh) in nyc 🇺🇸

that makes sense, with some caution

daniel r(ooooooh) in nyc 🇺🇸@danielroe.dev

oh, there are lots of good uses for them in the process of writing a blog. ... I just wouldn't use one to actually produce the words. (and I'd be very careful before I let it do too much in shaping the point I'm trying to make.)

@jaycooperbell.dev

Oh ya for sure. Everything it creates (finds?) for me gets cross checked and validated. No way am I just blindly publishing things from a hallucination machine...

Daniel Philip Johnson

@daniel-philips-enterprise.com

It just makes me self conscious now. Better cut out the bullet points. Becareful with the emojis and no em dashes. So just publish less and less. Reminds me of my time at university submitting a thesis and then being told there is 15% plagiarism.

daniel r(ooooooh) in nyc 🇺🇸

oh no! I absolutely do not negatively rank emojis or em dashes — I'm a big fan 😊

You should try spaced en dashes – preferred British usage, and not used by LLMs.

daniel r(ooooooh) in nyc 🇺🇸

you know, I just might but I'm not giving up my emojis

i saw this a lot with ai generated cover letters, which were usually a regurgitation of the resume (probably also largely just ai written). so when I came across a letter that said (for a sports tech intern): "i like sports. i am good at engineering." => immediately moved to next round