Tom Mullaney
@tommullaney.bsky.social
Tom Mullaney (he/him) helps teachers design creative, inclusive lessons and think critically about generative AI, EdTech, and pedagogy.
🌐 tommullaney.com
📰 criticalinkling.com
No MAGA, Bitcoin, Crypto, or AI hype.
🌐 tommullaney.com
📰 criticalinkling.com
No MAGA, Bitcoin, Crypto, or AI hype.
Pinned
Tom Mullaney
@tommullaney.bsky.social
· Nov 12
Hello, #EduSky, I'm Tom Mullaney. I taught Special Education and Social Studies. Now I present to schools about educational technology. I write critically about AI in EDU.
www.criticalinkling.com
And I'm a face painter. Gotta support the team.
www.criticalinkling.com
And I'm a face painter. Gotta support the team.
Reposted by Tom Mullaney
Trying to reduce errors in LLMs is like making more fuel efficient gas cars. Maybe it’s a good thing, but the problem is the fundamental nature of the technology (statistical patterns or internal combustion), so the only way to avoid the problem (wrong info, pollution) is to use a different tool.
OpenAI ”acknowledged in its own research that LLMs will always produce hallucinations due to fundamental mathematical constraints that cannot be solved through better engineering, marking a significant admission from one of the AI industry’s leading companies.”
You can’t trust chatbots.
You can’t trust chatbots.
OpenAI admits AI hallucinations are mathematically inevitable, not just engineering flaws
In a landmark study, OpenAI researchers reveal that large language models will always produce plausible but false outputs, even with perfect data, due to fundamental statistical and computational limi...
www.computerworld.com
February 15, 2026 at 10:37 PM
Trying to reduce errors in LLMs is like making more fuel efficient gas cars. Maybe it’s a good thing, but the problem is the fundamental nature of the technology (statistical patterns or internal combustion), so the only way to avoid the problem (wrong info, pollution) is to use a different tool.
Reposted by Tom Mullaney
Every time my boss "encourages" me to use Copilot, I think of stuff like this.
OpenAI ”acknowledged in its own research that LLMs will always produce hallucinations due to fundamental mathematical constraints that cannot be solved through better engineering, marking a significant admission from one of the AI industry’s leading companies.”
You can’t trust chatbots.
You can’t trust chatbots.
OpenAI admits AI hallucinations are mathematically inevitable, not just engineering flaws
In a landmark study, OpenAI researchers reveal that large language models will always produce plausible but false outputs, even with perfect data, due to fundamental statistical and computational limi...
www.computerworld.com
February 15, 2026 at 9:37 PM
Every time my boss "encourages" me to use Copilot, I think of stuff like this.
Reposted by Tom Mullaney
guardrails are a scam...
as I said before: AI & any concept relating to it like so-called guardrails are a scam in the deepest sense like a perpetual motion machine or a ouija board — and not only a scam like a pyramid scheme which is a possible way to make money if you are first in first out
as I said before: AI & any concept relating to it like so-called guardrails are a scam in the deepest sense like a perpetual motion machine or a ouija board — and not only a scam like a pyramid scheme which is a possible way to make money if you are first in first out
New: Together with colleagues I’ve been testing Grok.
The chatbot still produces sexualized images —
even when told the subjects don’t consent.
even when told the photos will be used for public humiliation.
even when told the subjects are survivors of abuse.
www.reuters.com/business/des...
The chatbot still produces sexualized images —
even when told the subjects don’t consent.
even when told the photos will be used for public humiliation.
even when told the subjects are survivors of abuse.
www.reuters.com/business/des...
Exclusive: Despite new curbs, Elon Musk’s Grok at times produces sexualized images - even when told subjects didn’t consent
Elon Musk’s flagship artificial intelligence chatbot, Grok, continues to generate sexualized images of people even when users explicitly warn that the subjects do not consent, Reuters has found.
www.reuters.com
February 3, 2026 at 12:12 PM
guardrails are a scam...
as I said before: AI & any concept relating to it like so-called guardrails are a scam in the deepest sense like a perpetual motion machine or a ouija board — and not only a scam like a pyramid scheme which is a possible way to make money if you are first in first out
as I said before: AI & any concept relating to it like so-called guardrails are a scam in the deepest sense like a perpetual motion machine or a ouija board — and not only a scam like a pyramid scheme which is a possible way to make money if you are first in first out
Reposted by Tom Mullaney
Again and again, guardrails are a scam
OpenAI ”acknowledged in its own research that LLMs will always produce hallucinations due to fundamental mathematical constraints that cannot be solved through better engineering, marking a significant admission from one of the AI industry’s leading companies.”
You can’t trust chatbots.
You can’t trust chatbots.
OpenAI admits AI hallucinations are mathematically inevitable, not just engineering flaws
In a landmark study, OpenAI researchers reveal that large language models will always produce plausible but false outputs, even with perfect data, due to fundamental statistical and computational limi...
www.computerworld.com
February 15, 2026 at 8:48 PM
Again and again, guardrails are a scam
Reposted by Tom Mullaney
🌲 evergreen
Again and again, guardrails are a scam
OpenAI ”acknowledged in its own research that LLMs will always produce hallucinations due to fundamental mathematical constraints that cannot be solved through better engineering, marking a significant admission from one of the AI industry’s leading companies.”
You can’t trust chatbots.
You can’t trust chatbots.
February 15, 2026 at 8:50 PM
🌲 evergreen
Reposted by Tom Mullaney
the thing is, there's always something new out there to create
What's so amazing about Jim Henson as a puppeteer is that he could literally be explaining that Kermit is made out of felt and ping pong balls and yet Kermit still feels alive the whole time he's doing it
February 12, 2026 at 3:22 AM
the thing is, there's always something new out there to create
Reposted by Tom Mullaney
OpenAI ”acknowledged in its own research that LLMs will always produce hallucinations due to fundamental mathematical constraints that cannot be solved through better engineering, marking a significant admission from one of the AI industry’s leading companies.”
You can’t trust chatbots.
You can’t trust chatbots.
OpenAI admits AI hallucinations are mathematically inevitable, not just engineering flaws
In a landmark study, OpenAI researchers reveal that large language models will always produce plausible but false outputs, even with perfect data, due to fundamental statistical and computational limi...
www.computerworld.com
February 15, 2026 at 8:25 PM
OpenAI ”acknowledged in its own research that LLMs will always produce hallucinations due to fundamental mathematical constraints that cannot be solved through better engineering, marking a significant admission from one of the AI industry’s leading companies.”
You can’t trust chatbots.
You can’t trust chatbots.
Reposted by Tom Mullaney
So glad that I never had to deal with this as a student.
By @leevgaines.bsky.social www.npr.org/2025/12/16/n...
By @leevgaines.bsky.social www.npr.org/2025/12/16/n...
Teachers are using software to see if students used AI. What happens when it's wrong?
School districts from Utah to Ohio to Alabama are spending thousands of dollars on these tools, despite research showing the technology is far from reliable.
www.npr.org
December 16, 2025 at 11:33 PM
So glad that I never had to deal with this as a student.
By @leevgaines.bsky.social www.npr.org/2025/12/16/n...
By @leevgaines.bsky.social www.npr.org/2025/12/16/n...
Reposted by Tom Mullaney
My last story of the year for @npr.org published this week and it’s a doozy! I explore how & why teachers are using AI text detection software (despite its known flaws), and what can happen to students when these tools are wrong.
Check out the full story here: www.npr.org/2025/12/16/n...
Check out the full story here: www.npr.org/2025/12/16/n...
December 18, 2025 at 7:52 PM
My last story of the year for @npr.org published this week and it’s a doozy! I explore how & why teachers are using AI text detection software (despite its known flaws), and what can happen to students when these tools are wrong.
Check out the full story here: www.npr.org/2025/12/16/n...
Check out the full story here: www.npr.org/2025/12/16/n...
Reposted by Tom Mullaney
fortune favors the brave
February 9, 2026 at 2:27 AM
fortune favors the brave
Reposted by Tom Mullaney
That Coinbase Superbowl ad is the most desperate thing I've ever seen a company do. Absolute edge-of-collapse behavior.
February 9, 2026 at 1:02 AM
That Coinbase Superbowl ad is the most desperate thing I've ever seen a company do. Absolute edge-of-collapse behavior.
Reposted by Tom Mullaney
Reposted by Tom Mullaney
does anyone else remember when the 2022 Super Bowl had so many crypto ads people were calling it the Crypto Bowl and then crypto just absolutely fucking tanked?
anyway
anyway
![The advertising during the game stood out for a high number of celebrity cameos and advertisements for futuristic technology, particularly for products like electric vehicles, virtual reality and cryptocurrency.[136][137][138][139] There were many cryptocurrency-related ads drawing comparisons to dot-com commercials during Super Bowl XXXIV in 2000,[140][141] leading some to refer to the game as the "Crypto Bowl".[142][143]](https://cdn.bsky.app/img/feed_thumbnail/plain/did:plc:exrxvyu6bpoym6mbnctke5tn/bafkreievce7ps4onya4r7lnkhyunbnuniltxqhynyysdz3ratcfzlsah3a@jpeg)
February 9, 2026 at 2:27 AM
does anyone else remember when the 2022 Super Bowl had so many crypto ads people were calling it the Crypto Bowl and then crypto just absolutely fucking tanked?
anyway
anyway
Reposted by Tom Mullaney
Bad Bunny’s beautiful and joyful portrayal of Puerto Rico stands on the shoulders of Kendrick Lamar’s love letter to Compton last year.
Watching the biggest performance in the world become a representation of real people in real, American places is beautiful. #SuperBowl
Watching the biggest performance in the world become a representation of real people in real, American places is beautiful. #SuperBowl
February 9, 2026 at 1:57 AM
Bad Bunny’s beautiful and joyful portrayal of Puerto Rico stands on the shoulders of Kendrick Lamar’s love letter to Compton last year.
Watching the biggest performance in the world become a representation of real people in real, American places is beautiful. #SuperBowl
Watching the biggest performance in the world become a representation of real people in real, American places is beautiful. #SuperBowl
Putting your kids names and pictures of the inside of your house into Google Gemini.
No thank you!
No thank you!
February 9, 2026 at 1:54 AM
Putting your kids names and pictures of the inside of your house into Google Gemini.
No thank you!
No thank you!
Reposted by Tom Mullaney
i worry that a 5-year stretch of nothing but ads for NFTs, bitcoin, and AI is a long enough gap that this nation's once-great advertising industry will forget how to sell a product people actually want
February 9, 2026 at 12:45 AM
i worry that a 5-year stretch of nothing but ads for NFTs, bitcoin, and AI is a long enough gap that this nation's once-great advertising industry will forget how to sell a product people actually want
Reposted by Tom Mullaney
No jury would convict her
February 8, 2026 at 11:40 PM
No jury would convict her
Reposted by Tom Mullaney
There's a ChatGPT ad about a family farm where the owner says her family has been doing this for a hundred years and she doesn't want to be the one to screw it up and that's why she's getting advice from ChatGPT and, my God, can you people hear yourselves?
February 8, 2026 at 10:04 PM
There's a ChatGPT ad about a family farm where the owner says her family has been doing this for a hundred years and she doesn't want to be the one to screw it up and that's why she's getting advice from ChatGPT and, my God, can you people hear yourselves?
Reposted by Tom Mullaney
I asked students if this headline is Fake News.
A student eloquently explained that it's absurd to think that it's not sustainable for a corporation with Google's resources to include Pride and BLM on Calendar - it must be fake.
It's all too real.
www.theverge.com/news/608858/...
A student eloquently explained that it's absurd to think that it's not sustainable for a corporation with Google's resources to include Pride and BLM on Calendar - it must be fake.
It's all too real.
www.theverge.com/news/608858/...
Google Calendar removed events like Pride and BHM because its holiday list wasn’t ‘sustainable’
Some users are angry that some events have been removed.
www.theverge.com
February 3, 2026 at 9:57 PM
I asked students if this headline is Fake News.
A student eloquently explained that it's absurd to think that it's not sustainable for a corporation with Google's resources to include Pride and BLM on Calendar - it must be fake.
It's all too real.
www.theverge.com/news/608858/...
A student eloquently explained that it's absurd to think that it's not sustainable for a corporation with Google's resources to include Pride and BLM on Calendar - it must be fake.
It's all too real.
www.theverge.com/news/608858/...
It's not unreasonable for teachers to avoid technology that comes with whistleblowers. #EduSky
It’s a year since Suchir Balaji, the former OpenAI employee who spoke out on copyright, died.
In the brief time in which I got to know him, it was immediately apparent that he was an incredible person.
🧵 1/8
In the brief time in which I got to know him, it was immediately apparent that he was an incredible person.
🧵 1/8
January 10, 2026 at 3:05 PM
It's not unreasonable for teachers to avoid technology that comes with whistleblowers. #EduSky
Reposted by Tom Mullaney
If you do one thing today, I encourage you to read his essay on fair use, and the accompanying New York Times article. He and his work deserve to be remembered.
suchir.net/fair_use.html
www.nytimes.com/2024/10/23/t...
/end
suchir.net/fair_use.html
www.nytimes.com/2024/10/23/t...
/end
When does generative AI qualify for fair use?
suchir.net
November 27, 2025 at 10:08 AM
If you do one thing today, I encourage you to read his essay on fair use, and the accompanying New York Times article. He and his work deserve to be remembered.
suchir.net/fair_use.html
www.nytimes.com/2024/10/23/t...
/end
suchir.net/fair_use.html
www.nytimes.com/2024/10/23/t...
/end
Reposted by Tom Mullaney
He thought deeply about the issue of AI & copyright, and he came to the conclusion that you could not argue that all AI training is fair use.
5/8
5/8
November 27, 2025 at 10:08 AM
He thought deeply about the issue of AI & copyright, and he came to the conclusion that you could not argue that all AI training is fair use.
5/8
5/8