Sithamet
@sithamet.bsky.social
Writer (who writes novels) looking to publish his genre-slippery sci-fi thriller. Building AI-based solutions in EdTech. Consultant: AI integration, knowledge management & automation. Check my novel: psychothrone.com/silence-plan
Pinned
Sithamet
@sithamet.bsky.social
· Nov 13
Hi, so here I am at Bsky because Twitter has no reach for newbies.
Will post about building with AI in EdTech, struggles with traditional publishing of an untraditional novel, and stuff.
Have nothing to sell yet (but my unpublished novel).
Check it out btw: psychothrone.com/silence-plan
Will post about building with AI in EdTech, struggles with traditional publishing of an untraditional novel, and stuff.
Have nothing to sell yet (but my unpublished novel).
Check it out btw: psychothrone.com/silence-plan
AI art is actually punk if you think about it. Everyone hates it, it's a counter culture movement, it will be the future. Nothing gets more punk than AI art
December 18, 2024 at 12:46 PM
AI art is actually punk if you think about it. Everyone hates it, it's a counter culture movement, it will be the future. Nothing gets more punk than AI art
Completely insensitive to what the majority of the US undergrad is, which is a grind of useless essays that do not really make your brain work.
November 29, 2024 at 10:07 PM
Completely insensitive to what the majority of the US undergrad is, which is a grind of useless essays that do not really make your brain work.
Want to self-improve on Bluesky?
Follow people trolling who will require you to get smarter.
Follow people trolling who will require you to get smarter.
November 28, 2024 at 9:33 PM
Want to self-improve on Bluesky?
Follow people trolling who will require you to get smarter.
Follow people trolling who will require you to get smarter.
When I first heard about cogsec (cognitive security), it was very schizo for me. Now, I think it's the new literacy.
November 28, 2024 at 6:57 PM
When I first heard about cogsec (cognitive security), it was very schizo for me. Now, I think it's the new literacy.
Deepfakes, rogue manipulative AIs, and NSWF AIs must be allowed and legal.
Bad actors will have them anyway.
Instead, teach people cognitive security, digital literacy, and healthy mistrust.
Rules for pedestrians changed when cars appeared. Now, the rules of being online are up to change.
Bad actors will have them anyway.
Instead, teach people cognitive security, digital literacy, and healthy mistrust.
Rules for pedestrians changed when cars appeared. Now, the rules of being online are up to change.
November 28, 2024 at 6:56 PM
Deepfakes, rogue manipulative AIs, and NSWF AIs must be allowed and legal.
Bad actors will have them anyway.
Instead, teach people cognitive security, digital literacy, and healthy mistrust.
Rules for pedestrians changed when cars appeared. Now, the rules of being online are up to change.
Bad actors will have them anyway.
Instead, teach people cognitive security, digital literacy, and healthy mistrust.
Rules for pedestrians changed when cars appeared. Now, the rules of being online are up to change.
Reposted by Sithamet
I am queer, but Veilguard's writing is really bad because it creates queer characters whose whole identity is based on gender and sexual orientation issues. It's caricaturic and discredits the ideas of equity, making it cultish and uncool.
And that's me, the writer. Normies must be losing shit.
And that's me, the writer. Normies must be losing shit.
November 28, 2024 at 3:01 PM
I am queer, but Veilguard's writing is really bad because it creates queer characters whose whole identity is based on gender and sexual orientation issues. It's caricaturic and discredits the ideas of equity, making it cultish and uncool.
And that's me, the writer. Normies must be losing shit.
And that's me, the writer. Normies must be losing shit.
It's kinda crazy that people compare the liberal ethics of Claude and the history rewriting of Chinese models as completely equivalent cultural things. No.
In the AI field, you aren't fighting for culture or social justice, people; the skin in the game is your very existence now.
In the AI field, you aren't fighting for culture or social justice, people; the skin in the game is your very existence now.
November 28, 2024 at 12:55 PM
It's kinda crazy that people compare the liberal ethics of Claude and the history rewriting of Chinese models as completely equivalent cultural things. No.
In the AI field, you aren't fighting for culture or social justice, people; the skin in the game is your very existence now.
In the AI field, you aren't fighting for culture or social justice, people; the skin in the game is your very existence now.
Honest question to seasoned writers: it is normal that I am most productive and creative past midnight?
I want to live forever and get my deep sleep and circadian rhythms and stuff. But the sheer pleasure of creative power flowing through my veins is enough for me to risk my very life.
I want to live forever and get my deep sleep and circadian rhythms and stuff. But the sheer pleasure of creative power flowing through my veins is enough for me to risk my very life.
November 26, 2024 at 11:29 PM
Honest question to seasoned writers: it is normal that I am most productive and creative past midnight?
I want to live forever and get my deep sleep and circadian rhythms and stuff. But the sheer pleasure of creative power flowing through my veins is enough for me to risk my very life.
I want to live forever and get my deep sleep and circadian rhythms and stuff. But the sheer pleasure of creative power flowing through my veins is enough for me to risk my very life.
If Bluesky is where you learn, capturing sources while scrolling is crucial. Don’t mess it up.
November 26, 2024 at 10:46 PM
If Bluesky is where you learn, capturing sources while scrolling is crucial. Don’t mess it up.
When using new Claude's custom styles, always check the style prompt manually after generation.
In my tests, it hallucinated <userExamples>, leading to an output that does not really match the example style.
Replace it with an actual example of your text, and you'll get better results.
In my tests, it hallucinated <userExamples>, leading to an output that does not really match the example style.
Replace it with an actual example of your text, and you'll get better results.
November 26, 2024 at 9:01 PM
When using new Claude's custom styles, always check the style prompt manually after generation.
In my tests, it hallucinated <userExamples>, leading to an output that does not really match the example style.
Replace it with an actual example of your text, and you'll get better results.
In my tests, it hallucinated <userExamples>, leading to an output that does not really match the example style.
Replace it with an actual example of your text, and you'll get better results.
I mean I should be humble, but it's literally me.
November 26, 2024 at 9:00 PM
I mean I should be humble, but it's literally me.
Being a sci-fi writer prepares you better for the future than any job or degree you may possibly have. AI engineering, including.
November 26, 2024 at 12:44 PM
Being a sci-fi writer prepares you better for the future than any job or degree you may possibly have. AI engineering, including.
The cybersecurity situation is going to be insane soon. Thanks to AI, of course. 🧵
November 26, 2024 at 12:42 PM
The cybersecurity situation is going to be insane soon. Thanks to AI, of course. 🧵
I feel that the advantage that LLMs give you will soon close.
Yeah, now you are a closet superhero with great powers, while people around you IRL walk powerless.
Soon, Corpos will make using LLMs mandatory for employees to increase productivity. 🧵
Yeah, now you are a closet superhero with great powers, while people around you IRL walk powerless.
Soon, Corpos will make using LLMs mandatory for employees to increase productivity. 🧵
November 25, 2024 at 8:20 PM
I feel that the advantage that LLMs give you will soon close.
Yeah, now you are a closet superhero with great powers, while people around you IRL walk powerless.
Soon, Corpos will make using LLMs mandatory for employees to increase productivity. 🧵
Yeah, now you are a closet superhero with great powers, while people around you IRL walk powerless.
Soon, Corpos will make using LLMs mandatory for employees to increase productivity. 🧵
4chan and imageboards are really useful for broadening your horizons. You cannot google the story of the Chinese student roasted and eaten like in this pic, no matter how hard you try.
November 25, 2024 at 7:26 PM
4chan and imageboards are really useful for broadening your horizons. You cannot google the story of the Chinese student roasted and eaten like in this pic, no matter how hard you try.
A non-blue checkmark is getting reach with obviously fake info (hating the West is a popular narrative in Russia). Some guy praises Must for truth right after.
And algo pushes it right on me all of a sudden, instead of AI posts (all I'm interested in on X).
And algo pushes it right on me all of a sudden, instead of AI posts (all I'm interested in on X).
November 25, 2024 at 3:27 PM
A non-blue checkmark is getting reach with obviously fake info (hating the West is a popular narrative in Russia). Some guy praises Must for truth right after.
And algo pushes it right on me all of a sudden, instead of AI posts (all I'm interested in on X).
And algo pushes it right on me all of a sudden, instead of AI posts (all I'm interested in on X).
Casually writing detective scenes while people are shooting drones down outside; man can get used to anything.
November 24, 2024 at 9:05 PM
Casually writing detective scenes while people are shooting drones down outside; man can get used to anything.
I should be doing Google's long-context competition, but instead, I am writing gay mage porn now. Guess it explains a lot about where I am now in life atm (0 regrets).
November 23, 2024 at 8:43 AM
I should be doing Google's long-context competition, but instead, I am writing gay mage porn now. Guess it explains a lot about where I am now in life atm (0 regrets).
Reposted by Sithamet
Looking for crime, mystery and thriller book reviews and recommendations? Here's a starter pack of blog bloggers: go.bsky.app/QgFVnrR
November 17, 2024 at 12:11 PM
Looking for crime, mystery and thriller book reviews and recommendations? Here's a starter pack of blog bloggers: go.bsky.app/QgFVnrR
Given all Gemini's struggles, it really screams to me that Grok 3 will be underwhelming (if it ever sees the light), and Musk will be so unhappy. See a lot of Grok-pushing on Twitter lately, which is likely preparation for the coming release. 🧵
November 18, 2024 at 5:42 PM
Given all Gemini's struggles, it really screams to me that Grok 3 will be underwhelming (if it ever sees the light), and Musk will be so unhappy. See a lot of Grok-pushing on Twitter lately, which is likely preparation for the coming release. 🧵
Hey, @bsky.app, can we please have bookmarks? Really like all those Starter Lists I see, but I would rather not follow all and everyone immediately, but rather save for later and proceed with caution.
November 18, 2024 at 9:30 AM
Hey, @bsky.app, can we please have bookmarks? Really like all those Starter Lists I see, but I would rather not follow all and everyone immediately, but rather save for later and proceed with caution.
AI won't ever write for you. But don't overlook AIs as a work tool, though.
Without AI's help in coding, I'd never have been able to recover, sort, label, and organize 400k words of the novel written in Discord (yeah, my co-author and I write in private Discord servers).
Without AI's help in coding, I'd never have been able to recover, sort, label, and organize 400k words of the novel written in Discord (yeah, my co-author and I write in private Discord servers).
November 18, 2024 at 9:23 AM
AI won't ever write for you. But don't overlook AIs as a work tool, though.
Without AI's help in coding, I'd never have been able to recover, sort, label, and organize 400k words of the novel written in Discord (yeah, my co-author and I write in private Discord servers).
Without AI's help in coding, I'd never have been able to recover, sort, label, and organize 400k words of the novel written in Discord (yeah, my co-author and I write in private Discord servers).
I can reliably say that writing (real creative fiction writing) is the last industry to be slain by AI.
Why? Because alignment and helpfulness training AIs undergo before going public significantly limits the scope and variability of the language they can produce... 🧵
Why? Because alignment and helpfulness training AIs undergo before going public significantly limits the scope and variability of the language they can produce... 🧵
November 18, 2024 at 9:21 AM
I can reliably say that writing (real creative fiction writing) is the last industry to be slain by AI.
Why? Because alignment and helpfulness training AIs undergo before going public significantly limits the scope and variability of the language they can produce... 🧵
Why? Because alignment and helpfulness training AIs undergo before going public significantly limits the scope and variability of the language they can produce... 🧵
The reason why I won't be sad if my product flops is because I build a thing I really need and can't have it on the market. So, at least, I'll have the best LLM-powered translation tool on the market to keep pushing my novel and writing abroad.
November 17, 2024 at 9:35 AM
The reason why I won't be sad if my product flops is because I build a thing I really need and can't have it on the market. So, at least, I'll have the best LLM-powered translation tool on the market to keep pushing my novel and writing abroad.
At this point, no AI benchmark matches the vibe and actual usefulness of the model.
It's so bad that I rewrote some evals to serve as internal benchmarks with which I "welcome" new models.
It's so bad that I rewrote some evals to serve as internal benchmarks with which I "welcome" new models.
November 14, 2024 at 1:13 PM
At this point, no AI benchmark matches the vibe and actual usefulness of the model.
It's so bad that I rewrote some evals to serve as internal benchmarks with which I "welcome" new models.
It's so bad that I rewrote some evals to serve as internal benchmarks with which I "welcome" new models.