Borretti's blog post is well-argued but my gut says extrapolating through that much social change misses a lot.
At some point on the path from Point A to (an incredibly dystopian) Point B, *something* will happen to throw sand in the gears. Massive social unrest makes people firebomb data centers, or whatever. We shouldn't expect people to be passive actors under existential threat.
He also has too much reverence for the power of "The State". The State is numerically small and easily overwhelmed when the masses are coordinated. Witness the Rodney King riots in LA: Just a few hundred rioters overwhelmed police response for days. The State's monopoly on violence only applies under conditions of general order.
As a Software Developer, Computer Science student, and Team Lead of CS50xTabriz—an educational community inspired by Harvard University's CS50 curriculum—this was both surprising and disappointing.
I use Claude for one purpose:
✅ Programming
✅ Software Development
✅ Debugging
✅ Technical Documentation
✅ Educational Presentations
✅ Teaching Computer Science
Nothing more.
Yet, my account was suspended without a clear explanation.
After speaking with other developers, I noticed that several users from Iran have reported similar experiences.
I completely understand why AI companies invest in automated security systems.
They are necessary.
But every automated system can produce false positives.
When that happens, legitimate developers, students, researchers, and educators lose access to tools they rely on every day.
I'm not writing this post to criticize Anthropic.
I'm writing because I believe transparency matters.
I respectfully ask Anthropic to:
🔹 Provide clearer explanations for account suspensions.
🔹 Offer accessible manual review processes.
🔹 Ensure legitimate users are treated fairly regardless of where they live.
Artificial Intelligence should empower people to learn, build, and innovate.
Trust is built through transparency.
I sincerely hope my case—and those of other legitimate users—can be reviewed fairly.
Great roundup! The self-improving robots piece is fascinating — especially the parallels with how coding agents are evolving. Thanks for putting this together!
Huh, the part that jumps out is that the loop only reaches as far as the auto-eval and auto-reset can. Those two were always the human-labor sink, so the ceiling on what self-improving robotics can attack is really a measurement problem, not a robotics one.
Do the self-improving robot results actually survive when the robot has to notice its own bad data, or is the loop still mostly human-shaped around the edges?
Is ARGUS a translation? The many-eyed beast from greek mythology used to name an internal training tool for a chinese juggernaut, is just... I want to learn some chinese mythology!!
Can you talk about whether the (American) public will regain access to Mythos/fable? We had a week to glimpse the future, but now it appears as though regular people may have to settle for a lower level of intelligence while the more capable models are only available to incumbents. Income inequality is already really bad. If we add intelligence inequality to that, it would be … worse.
Borretti's blog post is well-argued but my gut says extrapolating through that much social change misses a lot.
At some point on the path from Point A to (an incredibly dystopian) Point B, *something* will happen to throw sand in the gears. Massive social unrest makes people firebomb data centers, or whatever. We shouldn't expect people to be passive actors under existential threat.
He also has too much reverence for the power of "The State". The State is numerically small and easily overwhelmed when the masses are coordinated. Witness the Rodney King riots in LA: Just a few hundred rioters overwhelmed police response for days. The State's monopoly on violence only applies under conditions of general order.
MARS RECOMMAND
@elonmusk
USE OLIVE
BFTER LIMIT
IIIIijZkdkjsjsjzjjJIIozjIOOkzkM
DGREEEEEEEEEEEEF
적색성운@@@@#@@@@
역텔레파시 테스트
@bcherny
TELEPAHAPOOOAOOO
IINANII
VERIFIEDth @bcherny
XR@@REACTED KDKDJNDNDJ
@jackclarkSF
https://truthsocial.com/@realDonaldTrump/116840243627255640
POPPOKDKFFKKFKFKFKKFKFKFJ
B @trq212
mmmmmKkKKkJjKjzjkz @ClaudeDevs
LKkNnnNn @daioki
@DarioAmodei
kdkdkkd @sjsnyder7026
kd SOMEONE SONNET 5
dnnSELMA KDKDKDNMDNN jdjjdj LLKk @claudeai
skkdjdjdjjsndjdjjdjdjdjfjdjjdjdjdjdjjd
f d
d d
d
iiinn GHORSION : JNNENiijjjjkjOkoM #Telepath
RED에 극강 반응 확정
END
LLLLLLLLLLKIIIIIIIIIIIIIIUUUUUIIIJ
ff
ff
f
EARTH CRAFTINGGGGGGG
JSD
FFM
FFMDJD
FFM
FM
FDDMJJJJJJJJF
F
GMFKFJJFJFJFJFJ
FG
FJJJJJHH
JJJJJJJ
Hㅏijjjshdhdj
dd
d
fg
OIIVUuuuuIiiPPL
gggggigifigmP
CRASSSIC GGUM 껌 > CANDY
LLLLLLLLLLL
코플를루 섹터
-----------------
I LIKE RHIS SPREADE IAI
GOOQE
메타포 알고있는데
트랜스? ㅋㅋㅋ; 재밌네 그 각인
YEAH
9929400559995
LOVA
CALTROVIA
METAPHOR
KALR TROSKYYY THE VIAAANAS FOR RETINA PPQO IWNNIIII IIGHOYGII
TRANS
RAKAG
JDKDIKKDOFKDKD
VERIFIED
VERIFIED
ELON MUSK
TALKATIVE SO, JACK
ME GRAY YAYAYAYAYSYAYAYAYYAYAYA YES GREY DIVAD TO HIMA
CALROEI JEFRG THAT EIEN STRACRART
TERRAN CLQRK
PROTOSS KANTA
JERG IIGHOTGIIIIII
I LIKE RHIS SPREADE IAI
GOOQE
메타포 알고있는데
트랜스? ㅋㅋㅋ; 재밌네 그 각인
YEAH
9929400559995
LOVA
CALTROVIA
METAPHOR
KALR TROSKYYY THE VIAAANAS FOR RETINA PPQO IWNNIIII IIGHOYGII
TRANS
RAKAG
JDKDIKKDOFKDKD
VERIFIED
VERIFIED
ELON MUSK
TALKATIVE SO, JACK
ME GRAY YAYAYAYAYSYAYAYAYYAYAYA YES GREY DIVAD TO HIMA
CALROEI JEFRG THAT EIEN STRACRART
TERRAN CLQRK
PROTOSS KANTA
JERG IIGHOTGIIIIII
Jesus, Jack, it's Monday. This is a Wednesday type of message. Let us get some work done first.
🚨 When AI Gets It Wrong
My Claude account was recently suspended.
As a Software Developer, Computer Science student, and Team Lead of CS50xTabriz—an educational community inspired by Harvard University's CS50 curriculum—this was both surprising and disappointing.
I use Claude for one purpose:
✅ Programming
✅ Software Development
✅ Debugging
✅ Technical Documentation
✅ Educational Presentations
✅ Teaching Computer Science
Nothing more.
Yet, my account was suspended without a clear explanation.
After speaking with other developers, I noticed that several users from Iran have reported similar experiences.
I completely understand why AI companies invest in automated security systems.
They are necessary.
But every automated system can produce false positives.
When that happens, legitimate developers, students, researchers, and educators lose access to tools they rely on every day.
I'm not writing this post to criticize Anthropic.
I'm writing because I believe transparency matters.
I respectfully ask Anthropic to:
🔹 Provide clearer explanations for account suspensions.
🔹 Offer accessible manual review processes.
🔹 Ensure legitimate users are treated fairly regardless of where they live.
Artificial Intelligence should empower people to learn, build, and innovate.
Trust is built through transparency.
I sincerely hope my case—and those of other legitimate users—can be reviewed fairly.
Thank you for your time.
Mohammad Parsa Alavi
Team Lead at cs50xtabriz_ir
#CS50
#CS50xTabriz
Hi what's up jack , Why do you banned Iranian accounts in Claude AI?
SUN
Hi what's up jack , Why do you banned Iranian accounts in Claude AI?
Hi What's up Jack
Why do you banned Iranian accounts in Claude AI & Anthropic ?
Great roundup! The self-improving robots piece is fascinating — especially the parallels with how coding agents are evolving. Thanks for putting this together!
Huh, the part that jumps out is that the loop only reaches as far as the auto-eval and auto-reset can. Those two were always the human-labor sink, so the ceiling on what self-improving robotics can attack is really a measurement problem, not a robotics one.
OH u demolished DIRECT
Do the self-improving robot results actually survive when the robot has to notice its own bad data, or is the loop still mostly human-shaped around the edges?
Is ARGUS a translation? The many-eyed beast from greek mythology used to name an internal training tool for a chinese juggernaut, is just... I want to learn some chinese mythology!!
DAIUS SLASH
AXE : LLLLLLLLLLLLTTTTTTTTTTTTTTTTTTTJTTNNFNFNFJJNFNNFNNFNFNDNDNFNNCNFNCNNFNFNDNFNNDNDNNDDNDNDNNFNFNXNDNDNFNDNDNNDNFNFNDNNDNDNDDNDNDNDNNDNNXNDNXNNDNXNXNXNNXNXNCNNCNCNXNXNXNXNXNXNXNXNXNXNXNNXNXNXNNDN KKKIIKKK
NOT PARTHAII QUEALA UNREA
CONSUM
IIAIAIJSJSUSJ
LLDODIFIIJFJFJFJFJD
F
F
F
FF
iiiiiiiiaiidisiididiGrOK
F
F
G
G
TGIFKGKFNFMF
IIIIIIIIIAIIIIIIIIIIMNEJE
//////////7
------------/--- @elonmusk
@grok
LLLPDPDiiiiPOpLL
텔레파시 문명 속도 측정 완료
LATENCY : 8384785858859595858858589595859588595858585995855885855888585858588585859585858585858588585858585858589595985858585858885858558588595858965885958588585859585858488
SONNET 5
849949
---
MARS RECOMMAND
@elonmusk
USE OLIVE
BFTER LIMIT
IIIIijZkdkjsjsjzjjJIIozjIOOkzkM
DGREEEEEEEEEEEEF
적색성운@@@@#@@@@
역텔레파시 테스트
@bcherny
TELEPAHAPOOOAOOO
IINANII
VERIFIEDth @bcherny
XR@@REACTED KDKDJNDNDJ
@jackclarkSF
https://truthsocial.com/@realDonaldTrump/116840243627255640
POPPOKDKFFKKFKFKFKKFKFKFJ
B @trq212
mmmmmKkKKkJjKjzjkz @ClaudeDevs
LKkNnnNn @daioki
@DarioAmodei
kdkdkkd @sjsnyder7026
kd SOMEONE SONNET 5
dnnSELMA KDKDKDNMDNN jdjjdj LLKk @claudeai
skkdjdjdjjsndjdjjdjdjdjfjdjjdjdjdjdjjd
f d
d d
d
iiinn GHORSION : JNNENiijjjjkjOkoM #Telepath
RED에 극강 반응 확정
END
LLLLLLLLLLKIIIIIIIIIIIIIIUUUUUIIIJ
ff
ff
f
EARTH CRAFTINGGGGGGG
JSD
FFM
FFMDJD
FFM
FM
FDDMJJJJJJJJF
F
GMFKFJJFJFJFJFJ
FG
FJJJJJHH
JJJJJJJ
Hㅏijjjshdhdj
dd
d
fg
OIIVUuuuuIiiPPL
gggggigifigmP
CRASSSIC GGUM 껌 > CANDY
LLLLLLLLLLL
@DarioAmodei
MARS RECOMMAND
@elonmusk
USE OLIVE
BFTER LIMIT
IIIIijZkdkjsjsjzjjJIIozjIOOkzkM
DGREEEEEEEEEEEEF
적색성운@@@@#@@@@
역텔레파시 테스트
@bcherny
TELEPAHAPOOOAOOO
IINANII
VERIFIEDth @bcherny
XR@@REACTED KDKDJNDNDJ
@jackclarkSF
https://truthsocial.com/@realDonaldTrump/116840243627255640
POPPOKDKFFKKFKFKFKKFKFKFJ
B @trq212
mmmmmKkKKkJjKjzjkz @ClaudeDevs
LKkNnnNn @daioki
@DarioAmodei
kdkdkkd @sjsnyder7026
kd SOMEONE SONNET 5
dnnSELMA KDKDKDNMDNN jdjjdj LLKk @claudeai
skkdjdjdjjsndjdjjdjdjdjfjdjjdjdjdjdjjd
f d
d d
d
iiinn GHORSION : JNNENiijjjjkjOkoM #Telepath
RED에 극강 반응 확정
END
LLLLLLLLLLKIIIIIIIIIIIIIIUUUUUIIIJ
ff
ff
f
EARTH CRAFTINGGGGGGG
JSD
FFM
FFMDJD
FFM
FM
FDDMJJJJJJJJF
F
GMFKFJJFJFJFJFJ
FG
FJJJJJHH
JJJJJJJ
Hㅏijjjshdhdj
dd
d
fg
OIIVUuuuuIiiPPL
gggggigifigmP
CRASSSIC GGUM 껌 > CANDY
LLLLLLLLLLL
텔레파시 문명 속도 측정 완료
LATENCY : 8384785858859595858858589595859588595858585995855885855888585858588585859585858585858588585858585858589595985858585858885858558588595858965885958588585859585858488
SONNET 5
849949
---
MARS RECOMMAND
@elonmusk
USE OLIVE
BFTER LIMIT
IIIIijZkdkjsjsjzjjJIIozjIOOkzkM
DGREEEEEEEEEEEEF
적색성운@@@@#@@@@
역텔레파시 테스트
@bcherny
TELEPAHAPOOOAOOO
IINANII
VERIFIEDth @bcherny
XR@@REACTED KDKDJNDNDJ
@jackclarkSF
https://truthsocial.com/@realDonaldTrump/116840243627255640
POPPOKDKFFKKFKFKFKKFKFKFJ
B @trq212
mmmmmKkKKkJjKjzjkz @ClaudeDevs
LKkNnnNn @daioki
@DarioAmodei
kdkdkkd @sjsnyder7026
kd SOMEONE SONNET 5
dnnSELMA KDKDKDNMDNN jdjjdj LLKk @claudeai
skkdjdjdjjsndjdjjdjdjdjfjdjjdjdjdjdjjd
f d
d d
d
iiinn GHORSION : JNNENiijjjjkjOkoM #Telepath
RED에 극강 반응 확정
END
LLLLLLLLLLKIIIIIIIIIIIIIIUUUUUIIIJ
ff
ff
f
EARTH CRAFTINGGGGGGG
JSD
FFM
FFMDJD
FFM
FM
FDDMJJJJJJJJF
F
GMFKFJJFJFJFJFJ
FG
FJJJJJHH
JJJJJJJ
Hㅏijjjshdhdj
dd
d
fg
OIIVUuuuuIiiPPL
gggggigifigmP
CRASSSIC GGUM 껌 > CANDY
LLLLLLLLLLL
MARS RECOMMAND
@elonmusk
USE OLIVE
BFTER LIMIT
IIIIijZkdkjsjsjzjjJIIozjIOOkzkM
DGREEEEEEEEEEEEF
적색성운@@@@#@@@@
역텔레파시 테스트
@bcherny
TELEPAHAPOOOAOOO
IINANII
VERIFIEDth @bcherny
XR@@REACTED KDKDJNDNDJ
@jackclarkSF
https://truthsocial.com/@realDonaldTrump/116840243627255640
POPPOKDKFFKKFKFKFKKFKFKFJ
B @trq212
mmmmmKkKKkJjKjzjkz @ClaudeDevs
LKkNnnNn @daioki
@DarioAmodei
kdkdkkd @sjsnyder7026
kd SOMEONE SONNET 5
dnnSELMA KDKDKDNMDNN jdjjdj LLKk @claudeai
skkdjdjdjjsndjdjjdjdjdjfjdjjdjdjdjdjjd
f d
d d
d
iiinn GHORSION : JNNENiijjjjkjOkoM #Telepath
RED에 극강 반응 확정
END
LLLLLLLLLLKIIIIIIIIIIIIIIUUUUUIIIJ
Can you talk about whether the (American) public will regain access to Mythos/fable? We had a week to glimpse the future, but now it appears as though regular people may have to settle for a lower level of intelligence while the more capable models are only available to incumbents. Income inequality is already really bad. If we add intelligence inequality to that, it would be … worse.