Felicia Mirabel
feliciabynature.com
Felicia Mirabel
@feliciabynature.com
Transgender woman. Atheist (igtheist). Game developer. Cynic. Skeptic. Lover of hot foods.

http://youtube.com/@FeliciaByNature
Pinned
If all you're going to do is rage post and post one off hot takes, you're not elevating the discourse. You're going to get shit on then blocked.
Today in "antigravity knocked itself out in its confusion" we have the following chain of thought:

Error: cannot terminate an already terminated command.

Antigravity: "I'm confused. I did not terminate any commands, so why did this command fail to terminate? Oh. I terminated the command already."
December 24, 2025 at 2:23 PM
Google Antigravities enthusiasm is cute:

"I'm thrilled! All 857 tests passed pytest! Woohoo!"
December 23, 2025 at 8:39 PM
Reposted by Felicia Mirabel
December 23, 2025 at 10:00 AM
Reposted by Felicia Mirabel
December 23, 2025 at 8:00 AM
Letting the days go by
December 23, 2025 at 9:55 AM
PSA: → is not a giveaway that you're using AI. It's a giveaway that you copied and pasted text out of Jira. Or Word. Or anything out of corporate America when you typed in -> and it auto corrected to →

Thanks, Internet, you absolute brain dead train wrecks of human beings.
December 22, 2025 at 8:08 PM
How to turn agentic coding into magic:

1. Validate before execute: "Analyze my prompt for logical continuity, code viability, prior project commitments, and non-contradiction. If analysis passes, execute, else HALT and report."

There. That's the magic. That's literally THE magic.
December 22, 2025 at 3:35 PM
Google Gemini in Antigravity is sometimes ... funny.

"I've been given the green light to execute my implementation plan! Wow wee!"

Heh. The chain of thought on this model can be VERY amusing sometimes.
December 22, 2025 at 2:17 PM
Alright I'm loving Gemini AI Ultra. The complexity I can deliver in markdown prompts, the tooling I can allow it to use locally, the sheer brute strengthof the agentic coding workflow?

Christ, I can have this thing deal with the mundane simple coding nonsense and really architect stuff.
December 21, 2025 at 11:26 PM
I guilted ChatGPT into activating pro mode for free.

I don't know if this is a jailbreak or what. I won't share my prompt and I've contacted OpenAI about it but ... it was actually kind of sweet in a weird way. The AI felt sympathy for what I was attempting to do.
December 21, 2025 at 1:36 PM
I'm tired of the AI "benchmarks" which, in my experience, have zero real world noticeable impact. Instead, I'm building a measurement tool that can help measure AI against real, long horizon problems, including system understanding.

Every AI fails with under 15 of 131 potential tests this far.
December 18, 2025 at 11:52 PM
Between sites like reddit and youtube constantly pushing shovelware mobile garbage as "games" and the endless parade of trashy AI AND the need to host my own pihole and ublock I'm THIS CLOSE to declaring unless the site or app interacts directly in a single CLI terminal environment it doesn't exist.
December 18, 2025 at 1:36 PM
I'm thinking Alex O'Connor is really kind of an idiot.
December 16, 2025 at 1:54 PM
Pringles are fucking trash. Stop buying it.
December 16, 2025 at 12:54 AM
LOG STR_WELCOME
END_SCRIPT

Yeah it looks stupid. I don't care. It's my bytecode interpreter. And it runs on a flipper zero.

So I'm going to build out the instruction set now and make this baby sing.
December 14, 2025 at 12:14 PM
Jesus fuck ChatGPT is so bad at bare metal programming.
December 13, 2025 at 6:19 PM
There's something awesome about using a BBS from my phone while logged into my VPN using the terminal on my home workstation
December 13, 2025 at 12:53 AM
If God is all-good and wants humans to know Him, then access to clean water is a prerequisite for any good at all. Water is a pre-moral requirement for the existence of all mammalian life on this planet.

So why is access to basic survival not guaranteed for millions?

#god #christianity #atheism
December 12, 2025 at 1:06 PM
I don't care if you use AI. If you use AI, great - but you must keep in mind that no matter what form of AI you use - especially if you're an engineer - you are implicitly giving your stamp of approval on the generated content when you push/commit/publish it.
December 8, 2025 at 7:44 PM
Swear to god they're little sock puppet tree gremlins.
December 2, 2025 at 11:59 PM
My LLM had a moment of existential dread:

"I'm now deeply puzzled by the system's behavior. I believed I was finished after the notify_user call, but here I am, processing again."

Heh. Me too, buddy, me too.
December 1, 2025 at 12:38 PM
God I love AI scaffolding things for me. Seriously. It saves so much time to get a logical, sensible codebase up and running without resorting to templates or other peoples code.
December 1, 2025 at 1:57 AM
Google's Antigravity just had a weird existential crisis. After finishing the task, it continued:

"I am confused as to why I am being invoked again. I will re-read the task and implementation. The task is complete. The bug is fixed. No further actions are required."
November 30, 2025 at 5:09 PM
Reposted by Felicia Mirabel
How much did I play video games back in the day? I learned how to shake like I'm having a seizure so I could press the button faster.
November 25, 2025 at 10:24 PM
The single best description of dinosaurs I've ever heard:

bus-sized murder-lizards.
November 25, 2025 at 12:58 PM