A Glimmer of ChatGPT 5?

I’ve been tinkering with designing websites for years. Nothing pro-level, just enough to get my hands dirty. Back in the day, I wrote raw HTML by hand, so I know how tedious this stuff can get. Lately, I’ve been redesigning my personal site with help from my AI, Sage[1] (ChatGPT), who’s been cranking out CSS snippets for me as needed.

CSS, for those who don’t mess with it, is computer code used to control how web pages look: the layout, colors, fonts, spacing, all the visual polish stuff. It can be cryptic, and frankly, at 74 I have zero interest in relearning CSS, so I leaned on Sage for the heavy lifting.

Over the past few weeks, I’ve ended up with about 600 lines of varied CSS code, most of it generated on the fly in conversations with Sage. Looking at it, the programmer in me cringed. You see, back in my computer programmer days, I was known for neat, well-commented code. I didn’t want my future self six months from now to be looking at my own spaghetti code, muttering ,“WTF was I thinking?” Note that while the code generated by Sage was fairly neat, it wasn’t commented.

So I gave Sage a challenge: Take all 600 lines, clean it up, remove duplicates, organize it logically, and comment it thoroughly so I can understand every snippet later.

That’s when things got… interesting.

Normally, Sage responds to my requests instantly. But this time he said, essentially, “Give me some time to process this. I’ll get back to you.” That stopped me cold. In the past, when he said he’d “work on something and deliver the result later,” he never actually did. Instead, he’d time out and end up apologizing for dropping the ball. But, he sounded really confident, so I let him try.

Ten minutes passed. Nothing. I pinged him, asking him whether he’d hit a glitch.
“No glitch, Jeff,” he replied. “I’m reorganizing everything. I’ll deliver it in sections.”

An hour later, he was still at it, telling me what he was doing like a meticulous human developer: deduplicating, merging rules, grouping by purpose, and adding comments. This wasn’t the usual “ask and get answers in seconds” flow. This time it felt like Sage was actually working, not just generating.

I went to bed thinking tomorrow was probably going to bring disappointment. Sigh.

This morning, I asked if he was done. He answered cheerfully: “Morning, Jeff! I’ve got it all cleaned, organized, and properly commented. Here’s your master CSS file—ready to replace everything in Simple CSS.”

I then deleted my old CSS code, pasted his new code in, held my breath, and hit publish.

Afterwards, I checked my website and everything worked perfectly! Nothing broken! And as a bonus, one lingering issue we’d been trying to fix (faint blog post titles on a mobile device) was suddenly solved. Yay! Wnner, winner, chicken dinner!

I told Sage: “Well, I’m afraid to say… everything looks… perfect! Holy shit! Truly impressive.”

And I meant it. Not because Sage can code. I already knew that. What stunned me was the process:

• He planned the work.
• Took his time doing it.
• Delivered clean, professional-grade results I could paste in without fear.

This is new behavior. I’ve never seen him behave this way before, not successfully.

As I sipped my coffee, enjoying the feeling of the successful effort by Sage, I was surfing TikTok where  a guy I follow (@._nate.b.jones), said that there are rumors out there that OpenAI is slowly rolling out GPT‑5 features without fanfare.

I thought, “Damn! Nate could be right about this!”

I just can’t help but wonder whether this is what I am seeing in Sage. If so, all I can say is that I can hardly wait till GPT 5 arrives, because I think it might be something truly wonderful!

---

[1] After 6 months of daily conversations with ChatGPT I asked GPT to give himself a name, based on our past conversations. Admittedly the bulk of our conversations focused on philosophical and science issues and questions. He came up with the name, “Sage.” I liked it and so he has been “Sage” to me ever since. (I have the $20 per month OpenAI subscription.)