William Gunn
@metasynthesis.net
Communications and community. More here: http://synthesis.williamgunn.org/about/
Talk to me: https://calendar.app.google/z4KR3xfXTc178es47
Talk to me: https://calendar.app.google/z4KR3xfXTc178es47
An HBR study of AI at work finds people who adopt it are feeling more stressed. I feel like they have the causality backwards: it's the stress-prone people who are first to adopt it, but if you're not a good driver, buying a Ferrari isn't going to make you better. hbr.org/2026/02/ai-d...
February 10, 2026 at 5:17 PM
An HBR study of AI at work finds people who adopt it are feeling more stressed. I feel like they have the causality backwards: it's the stress-prone people who are first to adopt it, but if you're not a good driver, buying a Ferrari isn't going to make you better. hbr.org/2026/02/ai-d...
The monks walking for peace is such a sweet story. Feels like something from the 70s. www.reuters.com/world/us/wal...
Walk for peace: Buddhist monks arrive in Washington after 2,300-mile journey
Draped in burnt-orange robes, two dozen Buddhist monks are due to finish a 2,300-mile "Walk for Peace" in Washington, D.C., on Tuesday, a self-described spiritual journey across nine states that has b...
www.reuters.com
February 10, 2026 at 5:07 PM
The monks walking for peace is such a sweet story. Feels like something from the 70s. www.reuters.com/world/us/wal...
Jevon's Paradox strikes again
Because employees could do more, work began bleeding into lunch breaks and late evenings. The employees' to-do lists expanded to fill every hour that AI freed up, and then kept going.
The first signs of burnout are coming from the people who embrace AI the most | TechCrunch
Because employees could do more, work began bleeding into lunch breaks and late evenings. The employees' to-do lists expanded to fill every hour that AI freed up, and then kept going.
techcrunch.com
February 10, 2026 at 4:00 PM
Jevon's Paradox strikes again
Reposted by William Gunn
📣 Applications for the 23rd Summer Institute on Bounded Rationality are now open!
✨Join us in Berlin @arc-mpib.bsky.social June 08–16, 2026, to explore the topic of “Decision Making in the Age of AI”.
✏️ More details + application form (deadline: March 16): www.mpib-berlin.mpg.de/research/res...
✨Join us in Berlin @arc-mpib.bsky.social June 08–16, 2026, to explore the topic of “Decision Making in the Age of AI”.
✏️ More details + application form (deadline: March 16): www.mpib-berlin.mpg.de/research/res...
February 10, 2026 at 12:42 PM
📣 Applications for the 23rd Summer Institute on Bounded Rationality are now open!
✨Join us in Berlin @arc-mpib.bsky.social June 08–16, 2026, to explore the topic of “Decision Making in the Age of AI”.
✏️ More details + application form (deadline: March 16): www.mpib-berlin.mpg.de/research/res...
✨Join us in Berlin @arc-mpib.bsky.social June 08–16, 2026, to explore the topic of “Decision Making in the Age of AI”.
✏️ More details + application form (deadline: March 16): www.mpib-berlin.mpg.de/research/res...
Clearly this means you should round up your 8 best friends and go to the park
February 10, 2026 at 4:43 AM
Clearly this means you should round up your 8 best friends and go to the park
Hey Frontiers, if you don't want your editors corrupted by bribery, I have a brilliant idea. Pay them more than the bribe. www.researchinformation.info/news/frontie...
Frontiers warns over paid approaches to editors - Research Information
Publisher says financial compensation has been offered in connection with editorial or peer-review activities
www.researchinformation.info
February 9, 2026 at 3:50 PM
Hey Frontiers, if you don't want your editors corrupted by bribery, I have a brilliant idea. Pay them more than the bribe. www.researchinformation.info/news/frontie...
All recommendation systems do this: they show you the best matches first, but then they run out. It's fun to try to notice when that happens. You can try it on any platform.
concept: a feed that's For You for N minutes, after which it shows you increasingly Not For You posts to get you to log off
February 7, 2026 at 7:08 PM
All recommendation systems do this: they show you the best matches first, but then they run out. It's fun to try to notice when that happens. You can try it on any platform.
Reposted by William Gunn
How can we anticipate when AI will be able to do our jobs?
You could try building benchmarks as a leading indicator of automation, but they fail to capture the complexity of real-world tasks.
So Epoch AI researcher Anson Ho argues for an alternative: try automating tasks in your own job. 🧵
You could try building benchmarks as a leading indicator of automation, but they fail to capture the complexity of real-world tasks.
So Epoch AI researcher Anson Ho argues for an alternative: try automating tasks in your own job. 🧵
February 7, 2026 at 6:03 PM
How can we anticipate when AI will be able to do our jobs?
You could try building benchmarks as a leading indicator of automation, but they fail to capture the complexity of real-world tasks.
So Epoch AI researcher Anson Ho argues for an alternative: try automating tasks in your own job. 🧵
You could try building benchmarks as a leading indicator of automation, but they fail to capture the complexity of real-world tasks.
So Epoch AI researcher Anson Ho argues for an alternative: try automating tasks in your own job. 🧵
New polling out of Australia shows AI is a major concern among workers, with 69% wanting better regulation. unionsnsw.org.au/media-releas...
February 6, 2026 at 9:39 PM
New polling out of Australia shows AI is a major concern among workers, with 69% wanting better regulation. unionsnsw.org.au/media-releas...
TakeOverBench.com Mapping the progress to humanity losing control. From @xrobservatory.bsky.social & @pauseai.bsky.social
February 6, 2026 at 5:58 PM
TakeOverBench.com Mapping the progress to humanity losing control. From @xrobservatory.bsky.social & @pauseai.bsky.social
Anthropic found that 4.6 surpasses their tests for level 4 (the highest level) autonomy. Instead of taking that seriously, they decided to chuck the eval and just ask people if they should release it instead. 2nd image is an artists depiction of how that went. www-cdn.anthropic.com/0dd865075ad3...
February 6, 2026 at 5:56 PM
Anthropic found that 4.6 surpasses their tests for level 4 (the highest level) autonomy. Instead of taking that seriously, they decided to chuck the eval and just ask people if they should release it instead. 2nd image is an artists depiction of how that went. www-cdn.anthropic.com/0dd865075ad3...
Reposted by William Gunn
February 5, 2026 at 6:03 PM
There haven't been enough feuds among the big AI labs. This ought to fix it.
Okay well the OpenAI CMO is hopping on now and it is obvious that Anthropic hurt them and they were deeply unprepared
February 5, 2026 at 9:54 PM
There haven't been enough feuds among the big AI labs. This ought to fix it.
Reposted by William Gunn
Tracker for changes to Claude’s constitution: https://claude-soul.watch
February 4, 2026 at 11:09 PM
Tracker for changes to Claude’s constitution: https://claude-soul.watch
Still going exponential. How much time do we have left?
February 5, 2026 at 9:48 PM
Still going exponential. How much time do we have left?
Reposted by William Gunn
AI governance increasingly relies on broken benchmarks. @ankareuel.bsky.social found many can't distinguish signal from noise, lack documentation, and have poor validity. GPQA claims 448 multiple choice questions measure graduate reasoning. It doesn't really. 👇
February 5, 2026 at 4:32 PM
AI governance increasingly relies on broken benchmarks. @ankareuel.bsky.social found many can't distinguish signal from noise, lack documentation, and have poor validity. GPQA claims 448 multiple choice questions measure graduate reasoning. It doesn't really. 👇
Some proper oratory here!
Sir Ian McKellen performing a monologue from Shakespeare’s Sir Thomas More on the Stephen Colbert show. Never have I heard this monologue performed with such a keen sense of prescience. Nor have I ever been in this exact historical moment.TY Sir Ian, for reaching us once again.
#Pinks #ProudBlue
#Pinks #ProudBlue
February 5, 2026 at 5:00 PM
Some proper oratory here!
Reposted by William Gunn
Learn more in our blog post, "The Scientist AI: Safe by Design, by Not Desiring":
lawzero.org/en/unlisted/...
(4/4)
lawzero.org/en/unlisted/...
(4/4)
LawZero | The Scientist AI: Safe by Design, by Not Desiring
Scientific theories aspire to describe what is, as opposed to prescribe what ought to be. At LawZero, we take this idea as a design principle for safe artificial intelligence: that understanding—even ...
lawzero.org
February 5, 2026 at 3:15 PM
Learn more in our blog post, "The Scientist AI: Safe by Design, by Not Desiring":
lawzero.org/en/unlisted/...
(4/4)
lawzero.org/en/unlisted/...
(4/4)
Reposted by William Gunn
At LawZero, we're rethinking the building blocks of frontier AI to create an intelligent machine that is both highly capable and safe-by-design. We’re excited to share our first blog post outlining some of the objectives and core components of our Scientist AI project. 🧵
(1/4)
(1/4)
February 5, 2026 at 3:15 PM
At LawZero, we're rethinking the building blocks of frontier AI to create an intelligent machine that is both highly capable and safe-by-design. We’re excited to share our first blog post outlining some of the objectives and core components of our Scientist AI project. 🧵
(1/4)
(1/4)
Reposted by William Gunn
So many developers have sent me that Anthropic skills/mastery case study that I realized I should ungate what I *already wrote* about this: beginning principles to design workflows that work *with* your mind, not against it, & protect your problem-solving
www.fightforthehuman.com/cognitive-he...
www.fightforthehuman.com/cognitive-he...
Cognitive Helmets for the AI Bicycle: Part 1
I hear people name these three fears: will developers lose their problem-solving skills, learning opportunities, and critical thinking? One science-backed area can help: better metacognitive strategie...
www.fightforthehuman.com
February 4, 2026 at 6:13 PM
So many developers have sent me that Anthropic skills/mastery case study that I realized I should ungate what I *already wrote* about this: beginning principles to design workflows that work *with* your mind, not against it, & protect your problem-solving
www.fightforthehuman.com/cognitive-he...
www.fightforthehuman.com/cognitive-he...
If you care about the truth, you need more than just confirming data to prove a theory. You also need data that rules out other possible hypotheses.
If you just care about clicks, not so much, but be careful about exchanging long-term credibility for short term gains.
If you just care about clicks, not so much, but be careful about exchanging long-term credibility for short term gains.
We tracked public TikToks containing keywords like "ICE," "Alex Pretti," "Renee Good," "Trump," and "Epstein" over time. There was a big drop - but it hit everything, political and non-political alike. "Recipe" and "Oscar" posts fell off too.
February 4, 2026 at 10:45 PM
If you care about the truth, you need more than just confirming data to prove a theory. You also need data that rules out other possible hypotheses.
If you just care about clicks, not so much, but be careful about exchanging long-term credibility for short term gains.
If you just care about clicks, not so much, but be careful about exchanging long-term credibility for short term gains.
Reposted by William Gunn
The institutions that embrace machine-first FAIR will find themselves having more impact for their research and researchers.
More reuse. More trust. More interoperability.
Value, not volume.
www.digital-science.com/blog/2026/02...
More reuse. More trust. More interoperability.
Value, not volume.
www.digital-science.com/blog/2026/02...
Value over Volume: The Next Ten Years of Open Data - Digital Science
Mark Hahnel shares the open data wins of the past decade, the challenges, and the future of data sharing.
www.digital-science.com
February 4, 2026 at 9:00 AM
The institutions that embrace machine-first FAIR will find themselves having more impact for their research and researchers.
More reuse. More trust. More interoperability.
Value, not volume.
www.digital-science.com/blog/2026/02...
More reuse. More trust. More interoperability.
Value, not volume.
www.digital-science.com/blog/2026/02...
Reposted by William Gunn
🚨New WP "@Grok is this true?"
We analyze 1.6M factcheck requests on X (grok & Perplexity)
📌Usage is polarized, Grok users more likely to be Reps
📌BUT Rep posts rated as false more often—even by Grok
📌Bot agreement with factchecks is OK but not great; APIs match fact-checkers
osf.io/preprints/ps...
We analyze 1.6M factcheck requests on X (grok & Perplexity)
📌Usage is polarized, Grok users more likely to be Reps
📌BUT Rep posts rated as false more often—even by Grok
📌Bot agreement with factchecks is OK but not great; APIs match fact-checkers
osf.io/preprints/ps...
February 3, 2026 at 9:55 PM
🚨New WP "@Grok is this true?"
We analyze 1.6M factcheck requests on X (grok & Perplexity)
📌Usage is polarized, Grok users more likely to be Reps
📌BUT Rep posts rated as false more often—even by Grok
📌Bot agreement with factchecks is OK but not great; APIs match fact-checkers
osf.io/preprints/ps...
We analyze 1.6M factcheck requests on X (grok & Perplexity)
📌Usage is polarized, Grok users more likely to be Reps
📌BUT Rep posts rated as false more often—even by Grok
📌Bot agreement with factchecks is OK but not great; APIs match fact-checkers
osf.io/preprints/ps...
Reposted by William Gunn
With all the noise around AI, I hope this Report provides policymakers, researchers, and the public with the reliable evidence they need to make more informed choices. We also have an “Extended Summary for Policymakers”:
internationalaisafetyreport.org/publication/...
(18/19)
internationalaisafetyreport.org/publication/...
(18/19)
2026 Report: Extended Summary for Policymakers
The Extended Summary for Policymakers of the 2026 International AI Safety Report. The second International AI Safety Report, published in February 2026, is the next iteration of the comprehensive revi...
internationalaisafetyreport.org
February 3, 2026 at 1:16 PM
With all the noise around AI, I hope this Report provides policymakers, researchers, and the public with the reliable evidence they need to make more informed choices. We also have an “Extended Summary for Policymakers”:
internationalaisafetyreport.org/publication/...
(18/19)
internationalaisafetyreport.org/publication/...
(18/19)
Reposted by William Gunn
AI Safety Researchers in London 🇬🇧: Attend the London Alignment Workshop, March 2–3! Top ML researchers from industry, academia & government will discuss AI alignment, including model evaluations, interpretability, and robustness. 👇
February 4, 2026 at 9:30 AM
AI Safety Researchers in London 🇬🇧: Attend the London Alignment Workshop, March 2–3! Top ML researchers from industry, academia & government will discuss AI alignment, including model evaluations, interpretability, and robustness. 👇