adverbian

No one can convince me that the Crowley and Aziraphale we all know— the real ones, not the strangers of the finale— would ever choose our destruction (or their own, for the matter). They would understand on a personal, molecular level how vital it is to keep going and how important it is to be yourself in spite of the oppressive systems that do their damndest to shape us. They wouldn’t give up— not ever.

So fuck this finale, fuck the idea that it’s too late and fuck anyone who says otherwise, frankly.

I want to write something thoughtful but before I do, I need simply say: fuck that noise.

Good Omens is a queer allegory, and if queer people excel at anything it is existing despite the bigotry in the world. We have always been here. We always will be. Despite what religious and corporate over Lords desire. Queer joy is an act of rebellion in our current world, and is one that should be embraced rather than met with the message of the only way we can find happiness is to exist in a perfect world.

Aziraphale and Crowley both know that, too. I cannot believe our guardian angel and demon could co-exist with us for thousands of years and come to the conclusion that utter destruction is the only way to find your joy.

Enjoy the ending if you want. I am happy for you to do so. But don't go spouting destructive nonsense like this is the only path.

“In the end,” Crowley said, “Sebastian faced his friends, and watched them raise their bows. Draw back the strings. And loose the bolts toward his beautiful body.”

“Crowley,” whispered Aziraphale.

Crowley pushed his dark glasses up, fixing Aziraphale with his gaze. “Do you know what he did then?”

“I was there —“

“He blessed the arrows.” Crowley took hold of his shoulders. “And he ordered the men to shoot again. Again. Again.”

“Again,” echoed Aziraphale. His lips were parted; his chest rose and fell under the rich silk embroidery on his pale blue doublet.

“And do you know what happened after?”

“What else?” the angel breathed.

“He lived,” said Crowley. “The bastard lived. They dragged him down. Threw him in the pit. Left him for dead. But someone found him. Took him under their wing. Someone kind. Someone good. Someone who helped him, just like he’d helped everyone else. And he bloody well lived.”

Aziraphale stood with the light of hope blazing in his eyes, on the precipice of some grand gesture, or some grand folly.

And then he slumped, and said “Not for long.”

“He lived for long enough,” said Crowley, letting go. “Long enough to save some of the people he was protecting. Long enough to tell off the emperor. And long enough, maybe, for a little space of joy before the end.”

They walked together, not quite touching, a while longer.

Aziraphale said, “Of course, Sebastian earned the martyr’s crown, in the end.”

Anyway I was bitching about how hard it is to come up with feasible planets in Star Wars because sometimes you need a new planet from scratch and sometimes you need to know more about a planet than the 'has jungles, is probably a moon technically' than Wookieepedia will give you, and they're like 'oh yeah I can do something about that'.

So they've written (in Matlab but they swear it will run as a .exe as well and I may be conscripted to embed it as a web tool at some point) a star system generator.

You input what you know about the planet (ecosystem, population, sun colour, does it have liquid water, does it have a moon or moons, is it a moon or moons, temperature averages, atmosphere, you get me) and it will give you the... everything else about the star system, in obedience to real-universe physics. And if you input nothing you get a randomly generated star system.

And I’m like oh I know people who will be into this with a vengeance, and they're not on Tumblr, so this is me seeing who exactly would be keen on, and I cannot stress this enough, a real-physics comprehensive star system generator.

We have a meeting scheduled for next Tuesday to discuss my "valid concerns" so now I need all of you to send me anti-AI resources for that meeting because I am extremely bad at defending my viewpoints on the spot. I get flustered and forgetful and avoidant, so I need to have key arguments and evidence to back it up at the ready. Unfortunately, all my searches for "how to defend my anti-ai stance in the workplace" are only returning articles about "How to convince your employees that AI is a good thing"

Relevant information:

They are mostly looking at MS co-pilot and claude ai

They would be unlikely to use it to make any graphics so stolen art is only fuel to the ethical argument, not a main concern

Confidentiality. If your company allows its data to be put into an AI tool, the AI company may keep a copy on its servers. Your company’s data may be used to train the AI model, which means it may be used to answer other people’s questions. That puts your company’s data at risk of being exposed to competitors or hackers.

Reliability. How does your company plan to quality-check the AI outputs? AI will often sound right, but as soon as you start to fact-check, you will find out that it’s factually incorrect. It often makes up numbers, graphs, and sources. Any AI outputs must be independently fact-checked by a human before they can be used to make any real-world decisions. Otherwise, the company is at risk of making bad decisions based on bad information, resulting in business losses and possibly even legal consequences. EY (formerly Ernst & Young) did a survey finding that “Nearly every large company to have introduced AI has incurred some initial financial loss … often due to compliance failures, flawed outputs, bias, or disruptions to sustainability goals.” Your company should consider those possible financial risks.

Customer perception. Your company should know that the public increasingly perceives AI as high-risk and low-value. It may hurt your company’s brand perception. Here is an open-access article that surveyed a nationally-representative sample of 1100 people from Germany — they found that “while many AI applications are viewed as likely to occur, they are often associated with high perceived risk and limited benefit.”

Functionality. Your company needs to check and see whether these AI tools can actually do the things they are imagining, before they commit to paying for a subscription.

For example, I recently tested my employer’s internal AI tool to see whether it could automatically extract tables of data from PDF articles and put them into Excel. It couldn’t “just do it.” All it could do was give me a sample Python script that I needed to modify and run myself. The output of the script still requires a fair amount of manual cleanup.

Also, the sample script it gave me didn’t even work, because it was based on an outdated version of one of the Python libraries. So I had to debug the Python script myself.

EY — who did the study cited in the previous reblog, finding that businesses often incurred financial losses when AI gave them bad data — recently got bitten themselves. They had to retract a report after researchers caught fake data and references made up by AI. That can’t have been good for EY’s reputation. “EY retracts study after researchers discover AI hallucinations” (Financial Times, May 15, 2026)

“Starbucks quietly retired its AI agent just months after deployment after it miscounted coffee shop inventories and slowed down baristas” (Fortune, May 28, 2026)

“Doctors’ AI Systems Are Hallucinating Nonexistent Medical Issues During Appointments With Patients” (Futurism, May 16, 2026)

“AI hallucinations are infiltrating expert work—and entering the permanent body of knowledge” (Forbes, May 24, 2026)

“How AI Hallucinations Are Creating Real Security Risks” (The Hacker News, May 14, 2026)

We have a meeting scheduled for next Tuesday to discuss my "valid concerns" so now I need all of you to send me anti-AI resources for that meeting because I am extremely bad at defending my viewpoints on the spot. I get flustered and forgetful and avoidant, so I need to have key arguments and evidence to back it up at the ready. Unfortunately, all my searches for "how to defend my anti-ai stance in the workplace" are only returning articles about "How to convince your employees that AI is a good thing"

Relevant information:

They are mostly looking at MS co-pilot and claude ai

They would be unlikely to use it to make any graphics so stolen art is only fuel to the ethical argument, not a main concern

Confidentiality. If your company allows its data to be put into an AI tool, the AI company may keep a copy on its servers. Your company’s data may be used to train the AI model, which means it may be used to answer other people’s questions. That puts your company’s data at risk of being exposed to competitors or hackers.

Reliability. How does your company plan to quality-check the AI outputs? AI will often sound right, but as soon as you start to fact-check, you will find out that it’s factually incorrect. It often makes up numbers, graphs, and sources. Any AI outputs must be independently fact-checked by a human before they can be used to make any real-world decisions. Otherwise, the company is at risk of making bad decisions based on bad information, resulting in business losses and possibly even legal consequences. EY (formerly Ernst & Young) did a survey finding that “Nearly every large company to have introduced AI has incurred some initial financial loss … often due to compliance failures, flawed outputs, bias, or disruptions to sustainability goals.” Your company should consider those possible financial risks.

Customer perception. Your company should know that the public increasingly perceives AI as high-risk and low-value. It may hurt your company’s brand perception. Here is an open-access article that surveyed a nationally-representative sample of 1100 people from Germany — they found that “while many AI applications are viewed as likely to occur, they are often associated with high perceived risk and limited benefit.”

Functionality. Your company needs to check and see whether these AI tools can actually do the things they are imagining, before they commit to paying for a subscription.

For example, I recently tested my employer’s internal AI tool to see whether it could automatically extract tables of data from PDF articles and put them into Excel. It couldn’t “just do it.” All it could do was give me a sample Python script that I needed to modify and run myself. The output of the script still requires a fair amount of manual cleanup.

Also, the sample script it gave me didn’t even work, because it was based on an outdated version of one of the Python libraries. So I had to debug the Python script myself.

At Eden's gate a gentle angel

Stood shining with a bowed-down head,

While dark and fierce, the rebel demon

Above the hellish chasm sped.

The spirit born of doubt and scorning

Upon the spirit pure did gaze,

And felt, through unexpected yearning,

A tender warmth for the first days.

"Forgive me," then he softly uttered,

"I saw you, and your light was true;

Not all within this world I hated,

@earlgreydyke exactly.