How I Blog with Bots (But You Can Still Blame Me) 😅

The cover image is a result of me getting very impatient with ChatGPT after the third nearly perfect cover of some random stranger that was supposed to be me. So, I’m yelling, “that’s definitely not my face!” and ChatGPT insists theres a resemblance (there’s not). It didn’t take much for me to give up and upload a real picture, as a guide.

🦄 Spoiler: The very next one was a total disaster, but the last attempt? Not half bad! 🤣

At some point during the day yesterday, I was working on a project and talking to a friend about the whole AI “thing” (that’s the new technical word for it, by the way 🤩). Then out of nowhere, it dawned on me… I’m pretty positive that I don’t have a single post with the sort of footer/disclaimer that I’ve been preaching about for weeks, to anyone who would listen to me. 🤦‍♀️

You guys have got off easy (so far) – but some of the folks at work, I’m sure, were thrilled about the weekend! Don’t worry, guys, me too… me too. 😮

So, since I’ve made a point to go back and add a disclaimer to every post thus far, I decided to take a quick minute to explain what that really means – in case anyone is curious.

All the Ways I Let Robots Help 📡

GitHub Copilot

You probably guessed already – this is my primary AI for all things coding, but since their premium limits went into permanent effect on June 18th, I’ve been doing good to break even just with work. I’ll share what part’s have actually been working at some point, but so far the answer is “not a lot”.

I have some ideas, so it’s not a lost cause yet! For now though, I’m hovering right around 67% usage and work might actually cut me off if I don’t make it to the end of the month 👿

Codex

Since I already have an OpenAI subscription, I’ve been tinkering with Codex as a potential solution to the lack of premium requests 😱😭 We’ll see how it goes.

I’m trying to give it a fair chance… The good news in all this? You remember in my nightmare story I mentioned doing some pretty out-there things to get my hands on the enterprise version of the Copilot license? Well, apparently, if you exhaust 100% of your premium requests in less than 4 days AND make sure your boss is instantly aware of the problem – you know, so they understand the true gravity of the matter in real-time (which is around 7 PM on a Saturday) – work is suddenly more willing to share 😇

ChatGPT

I’ve been a long-time user for quite some time, but I wanted to call this guy out – and more specifically, exactly how ChatGPT is involved in all my posts. While Copilot is the topic of most everything I write, ChatGPT is usually busy behind the scenes.

🦄 Don’t worry, I’m sure the obsession will wear off eventually… maybe… oh hell, probably not. Never mind! 😅

Gemini

Gemini and I aren’t talking much right now – which has everything to do with the beta install I put on my Android before fully realizing what Google had actually done with all the hands-free tools the non-generative assistant used to handle (spoiler: not a damn thing is the answer 😡). So, for now, Gemini’s riding the bench.

I’m sure that won’t last long. Google might be trailing OpenAI and Anthropic in the AI game, but I’ve never known them to stay at the back of the pack. 🛸

The (Messy) Human-AI Creative Process 🕹️

Cover art

That one is easy to guess. I’ve almost perfected a prompt that will keep the primary image content vertically “squished” towards the center, so the aspect ratio will match suggested size guidelines after I crop it.

Today’s image is a test in full-size, but it was to good to mess with much, especially considering all the personalities it gave me along the way. 🥸

• Seems like either all of the training data was a perfect square OR more likely it’s been trained to output specific sizes and my attempts at a compromise violate that plan.

• I have a designated project for the banner images and roughly half the time I give it a prompt and just use whatever it spits out at me. The other half, it’s either not at all usable or it’s so terrible I really don’t want to share it.

It will take some work for me to go through those, but yes – I at least save them. 🔒

• Gimp isn’t AI, but it’s my go-to for anything image related. Up until a little bit ago I thought I was saving all images with meta tags that included the “Generated by/Edited with” comment, but either it’s just different on this computer or they’re not really being saved like I thought.

I’m gonna work on that before next post though – those should definitely be there! 🏁

Blog posts 💬

I’m not about to let AI write the whole thing for me (where’s the fun in that?), but I do rely on it a lot for drafts and edits. Just in case anyone’s wondering how the whole process actually happens, here it is – step by step:

ONE. Idea

Sometimes I’ll literally have no idea what I’m going to talk about until I actually start writing. Other times, I’ve been saving ideas for weeks! There’s really no rhyme or reason to this, so I just go with it.

Seems like now that I’m actually in a groove, these “off-the-usual-topic” (read Copilot) posts are more frequent, too. 🏎️

TWO. Draft

IS 100% ME – sometimes I’m lazy and will dictate to ChatGPT (occasionally Gemini, but we’re on a hiatus right now anyway), but I have an instruction that says specifically do not change anything until I tell you I’m ready to start editing.

So for the majority of the week (and more-than-occasionally late the night before 🤐) I’m working on a relatively cohesive unit to cover a specific something (“topic” is usually a stretch, at this point). Typically, the draft version is roughly twice as long as the final version and my posts aren’t exactly short to begin with (I know).

THREE. Edits

So this is where it gets a little tricky. I have a custom set of instructions for the blog posts, in which the majority of it is focused on two major things, in order:

• Scan the entire post and make changes wherever needed to remove unintentional bias, and for every change you make, summarize it with examples and show me why your version is better suited.

I’m getting better as we go, but it’s really just a safety net so I don’t accidentally put my foot in my mouth after taking the scenic route to my point (it’s a talent). 📸

• Keep my original attitude and quirks as much as possible, but not to the extent that it contradicts the first point.

I already trained it a long time ago to censor me. It doesn’t even bring that part up anymore, most of the time anyway. 😶

FOUR. Negotiation

I really don’t have a better word for what happens during this phase, because that’s exactly what it feels like. ChatGPT took my nice story and turned it into something that I only vaguely recognize, so I’ll argue that it’s TOO changed.

ChatGPT argues with me that not everybody understands what a “holler” is and that it’s probably already too far off-topic to explain whatever contextual point I wanted to make anyway.

It’s usually accurate, too. 😤

FIVE. Final polish

After we have a final working version (cause I’m still changing it randomly), the quizzing starts for all those words I want to say, but really just cannot put my finger on or I need to convey some sort of emotion or attitude (and I’m getting sick of seeing the same emojis all the time).

This is where the content of the blog conversation in ChatGPT pivots and I’ll have it show me a random 10 emojis that could work for a particular phrase or scenario – I’ll pick from there. Same with section titles and quotes, too.

💿 When it comes to titles, quotes, and references – it’s rarely as simple as picking one from a list. Most of the time, I’ll pick parts of at least three and then feed them back to ChatGPT to make coherent again.

SIX. Final review

At this point it’s all loaded into Dev, but I’m still reviewing repeatedly until I can make it from top to bottom without the urge to change something immediately (which can sometimes take a couple days – like this post ⚡️).

There’s typically a whole lot of page swapping (preview > edit > preview) until I’m finally happy with the overall version. From there, most posts get scheduled for Wednesday mornings.

🪴 Why? No reason, really – it’s more random chance. I’ve tried to stay on at least a weekly schedule since I joined. If I schedule the big posts, I know I won’t accidentally miss a day. 😇

TL;DR: The Heart of the Matter 🫀

If you’ve made it this far, now you know:

• Yes, I use AI – a LOT
• No, it doesn’t do all the work for me
• Yes, I argue with it (and sometimes lose…)
• No, it hasn’t taken over my personality
• Yes, the cover art is AI-generated
  • unless it’s truly hideous – those are locked in the vault!

The real secret is that all of this is still me! I’m just occasionally powered up by caffeine and robots. ☕️🤖

Chime In 🎤

Got tips, tricks, or your own tales of AI chaos? I’d love to hear from you – add your two cents below!

🛡️ I Used AI, But Blame Me

Everything I share here is my own perspective – created with the help of AI tools (GitHub Copilot, ChatGPT, and their friends), but always with a human (me) in the loop. I do my best to catch accidental bias and fact-check, but if you ever spot something odd, let me know! AI isn’t perfect, and neither am I.

TL; DR: I let the bots help. But if you spot something weird, blame the author, not the algorithm. 🙋‍♀️