Marco
@mcognetta.bsky.social
Language and keyboard stuff at Google + PhD student at Tokyo Institute of Technology.
I like computers and Korean and computers-and-Korean and high school CS education.
Georgia Tech → 연세대학교 → 東京工業大学.
https://theoreticallygoodwithcomputers.com/
I like computers and Korean and computers-and-Korean and high school CS education.
Georgia Tech → 연세대학교 → 東京工業大学.
https://theoreticallygoodwithcomputers.com/
Pinned
Marco
@mcognetta.bsky.social
· Jan 5
A lot of you followed me due to #NLP, but I like to post about #chess (especially computer chess), #programming (especially puzzles, code golf, etc), and machine learning.
And some less technical stuff like #Korean, #Esperanto, and #trains (mostly in Japan, just due to proximity).
And some less technical stuff like #Korean, #Esperanto, and #trains (mostly in Japan, just due to proximity).
Unusual hanja formatting in this passage I read.
I usually see hanja in brackets/parentheses or as sub/superscript or at least bolded, so seeing it just sort of nakedly attached to the hangul is a bit jarring. Especially because there is a space between 아편 and 전쟁 but not in 阿片戰爭.
#한국어 #한자
I usually see hanja in brackets/parentheses or as sub/superscript or at least bolded, so seeing it just sort of nakedly attached to the hangul is a bit jarring. Especially because there is a space between 아편 and 전쟁 but not in 阿片戰爭.
#한국어 #한자
February 13, 2026 at 10:12 PM
Reposted by Marco
I downloaded something like 300GB of open models and wrote a bunch of map-reduce style processing scripts to make this graph.
It's plotting the distribution of weight values across a variety of popular open models, to show that models are almost entirely made up of small floats.
It's plotting the distribution of weight values across a variety of popular open models, to show that models are almost entirely made up of small floats.
February 13, 2026 at 11:56 AM
I downloaded something like 300GB of open models and wrote a bunch of map-reduce style processing scripts to make this graph.
It's plotting the distribution of weight values across a variety of popular open models, to show that models are almost entirely made up of small floats.
It's plotting the distribution of weight values across a variety of popular open models, to show that models are almost entirely made up of small floats.
I need this but for SCVs.
peon-ping — Stop babysitting your terminal
Warcraft III Peon voice lines as Claude Code notifications. Never miss when Claude needs you.
peon-ping.vercel.app
February 12, 2026 at 7:26 AM
I need this but for SCVs.
Reposted by Marco
I'm looking for 5-10 #chess players to test out a tool I'm building. Preferably who play on @lichess.org and are 1200+ in rapid or blitz.
And if you coach chess at all, I'd be extra grateful to have you test it!
NOTE: it is _NOT_ an "LLM chess coach" tool, I promise!
🙏
And if you coach chess at all, I'd be extra grateful to have you test it!
NOTE: it is _NOT_ an "LLM chess coach" tool, I promise!
🙏
February 11, 2026 at 10:37 PM
I'm looking for 5-10 #chess players to test out a tool I'm building. Preferably who play on @lichess.org and are 1200+ in rapid or blitz.
And if you coach chess at all, I'd be extra grateful to have you test it!
NOTE: it is _NOT_ an "LLM chess coach" tool, I promise!
🙏
And if you coach chess at all, I'd be extra grateful to have you test it!
NOTE: it is _NOT_ an "LLM chess coach" tool, I promise!
🙏
I'm looking for 5-10 #chess players to test out a tool I'm building. Preferably who play on @lichess.org and are 1200+ in rapid or blitz.
And if you coach chess at all, I'd be extra grateful to have you test it!
NOTE: it is _NOT_ an "LLM chess coach" tool, I promise!
🙏
And if you coach chess at all, I'd be extra grateful to have you test it!
NOTE: it is _NOT_ an "LLM chess coach" tool, I promise!
🙏
February 11, 2026 at 10:37 PM
I'm looking for 5-10 #chess players to test out a tool I'm building. Preferably who play on @lichess.org and are 1200+ in rapid or blitz.
And if you coach chess at all, I'd be extra grateful to have you test it!
NOTE: it is _NOT_ an "LLM chess coach" tool, I promise!
🙏
And if you coach chess at all, I'd be extra grateful to have you test it!
NOTE: it is _NOT_ an "LLM chess coach" tool, I promise!
🙏
Reposted by Marco
If you think labeling text spans with LLMs is easy, you probably have not tried it yourself (we have! 🙃).
Any method you can think of – be it tagging, matching, or indexing – has flaws.
In our new preprint, we tested them all 💪We also proposed how to improve one of them.
arxiv.org/abs/2601.16946
Any method you can think of – be it tagging, matching, or indexing – has flaws.
In our new preprint, we tested them all 💪We also proposed how to improve one of them.
arxiv.org/abs/2601.16946
January 29, 2026 at 2:20 PM
If you think labeling text spans with LLMs is easy, you probably have not tried it yourself (we have! 🙃).
Any method you can think of – be it tagging, matching, or indexing – has flaws.
In our new preprint, we tested them all 💪We also proposed how to improve one of them.
arxiv.org/abs/2601.16946
Any method you can think of – be it tagging, matching, or indexing – has flaws.
In our new preprint, we tested them all 💪We also proposed how to improve one of them.
arxiv.org/abs/2601.16946
I propose modern frontier models should be classified as "Really Quite Big Language Models" (RQBLM).
Let's just follow radio astronomy in their naming schemes.
Let's just follow radio astronomy in their naming schemes.
People often joke that the smallest LLMs today should be called "small language models", but the GPT-2 tech report uses the phrase "large language model" and GPT-2 variants were 117M-1.5B parameters so anything >=117M is canonically large. cdn.openai.com/better-langu...
February 10, 2026 at 10:09 PM
I propose modern frontier models should be classified as "Really Quite Big Language Models" (RQBLM).
Let's just follow radio astronomy in their naming schemes.
Let's just follow radio astronomy in their naming schemes.
Reposted by Marco
This week on Overcommitted, we got to sit down with Bluesky's favorite tech blogger @samwho.dev and it did not disappoint!
Sam makes some of the coolest tech content on the internet, and if you haven't heard from him yet, you should! Full episode out now: overcommitted.dev/interactive-...
Sam makes some of the coolest tech content on the internet, and if you haven't heard from him yet, you should! Full episode out now: overcommitted.dev/interactive-...
February 10, 2026 at 5:54 PM
This week on Overcommitted, we got to sit down with Bluesky's favorite tech blogger @samwho.dev and it did not disappoint!
Sam makes some of the coolest tech content on the internet, and if you haven't heard from him yet, you should! Full episode out now: overcommitted.dev/interactive-...
Sam makes some of the coolest tech content on the internet, and if you haven't heard from him yet, you should! Full episode out now: overcommitted.dev/interactive-...
My work rotation today. Unbelievably good.
Thai Psych, Molam (หมอลำ), Luk Thung & Soul [Vinyl Studio Session] with Diana Ratsamee
YouTube video by Humano Studios
www.youtube.com
February 9, 2026 at 9:14 PM
My work rotation today. Unbelievably good.
@aclmeeting.bsky.social @aclrollingreview.bsky.social what is the right way to send a complaint about another review from a paper I am reviewing? One in my batch is absolutely terrible, and I am not confident in leaving it to the AC/etc to catch and properly handle it.
February 8, 2026 at 8:13 PM
@aclmeeting.bsky.social @aclrollingreview.bsky.social what is the right way to send a complaint about another review from a paper I am reviewing? One in my batch is absolutely terrible, and I am not confident in leaving it to the AC/etc to catch and properly handle it.
This is exactly what I want my desk space to look like.
February 8, 2026 at 5:47 AM
This is exactly what I want my desk space to look like.
@lichess.org announced a partial 8 piece tablebase that takes up 63TB on disk. They couldn't get a network to help them transfer it, so they just shipped it via plane on hard drives.
"Never underestimate the bandwidth of a station wagon full of tapes hurtling down the highway." -Andy Tanenbaum
"Never underestimate the bandwidth of a station wagon full of tapes hurtling down the highway." -Andy Tanenbaum
Op1 - Partial 8-piece tablebase available
63 TiB of chess knowledge sent across the Atlantic and now available on the Lichess analysis board
lichess.org
February 7, 2026 at 6:12 AM
@lichess.org announced a partial 8 piece tablebase that takes up 63TB on disk. They couldn't get a network to help them transfer it, so they just shipped it via plane on hard drives.
"Never underestimate the bandwidth of a station wagon full of tapes hurtling down the highway." -Andy Tanenbaum
"Never underestimate the bandwidth of a station wagon full of tapes hurtling down the highway." -Andy Tanenbaum
Reposted by Marco
📣 FLaNN 2026 at Yale 🍮
Invited talks+posters (non-archival): expressivity, computation, and learning in neural nets/LLMs
Speakers: Pablo Barceló, David Chiang, Will Merrill, Naomi Saphra, Gail Weiss
Abstracts due Feb 12, 2026
Details: flann.cs.yale.edu
Invited talks+posters (non-archival): expressivity, computation, and learning in neural nets/LLMs
Speakers: Pablo Barceló, David Chiang, Will Merrill, Naomi Saphra, Gail Weiss
Abstracts due Feb 12, 2026
Details: flann.cs.yale.edu
February 4, 2026 at 3:24 PM
📣 FLaNN 2026 at Yale 🍮
Invited talks+posters (non-archival): expressivity, computation, and learning in neural nets/LLMs
Speakers: Pablo Barceló, David Chiang, Will Merrill, Naomi Saphra, Gail Weiss
Abstracts due Feb 12, 2026
Details: flann.cs.yale.edu
Invited talks+posters (non-archival): expressivity, computation, and learning in neural nets/LLMs
Speakers: Pablo Barceló, David Chiang, Will Merrill, Naomi Saphra, Gail Weiss
Abstracts due Feb 12, 2026
Details: flann.cs.yale.edu
Reposted by Marco
The pressure is on across Poker, Werewolf and Chess as we head into the final hour of the Game Arena. Who will emerge as the ultimate champion?
Join @gmhikaru.bsky.social and Nick Schulman now! 👇
www.youtube.com/watch?v=vzMj...
Join @gmhikaru.bsky.social and Nick Schulman now! 👇
www.youtube.com/watch?v=vzMj...
Kaggle Poker / Chess / Werewolf Game Arena Day 3 w Nick Schulman and Hikaru #ad
YouTube video by GMHikaru
www.youtube.com
February 4, 2026 at 6:32 PM
The pressure is on across Poker, Werewolf and Chess as we head into the final hour of the Game Arena. Who will emerge as the ultimate champion?
Join @gmhikaru.bsky.social and Nick Schulman now! 👇
www.youtube.com/watch?v=vzMj...
Join @gmhikaru.bsky.social and Nick Schulman now! 👇
www.youtube.com/watch?v=vzMj...
Reposted by Marco
"Lowering the activation energy" is a great concept
My latest use for Claude Cowork:
I posted photos of a dozen things I wanted to get rid of and asked it to help make a plan of what to donate or trash and where, based on local organizations' rules and hours.
Claude lowers the activation energy for stuff like this that I would otherwise put off.
I posted photos of a dozen things I wanted to get rid of and asked it to help make a plan of what to donate or trash and where, based on local organizations' rules and hours.
Claude lowers the activation energy for stuff like this that I would otherwise put off.
January 31, 2026 at 1:23 PM
"Lowering the activation energy" is a great concept
Reposted by Marco
BTW I have not reposted this in a while and there are plenty of new AI people around. Ping to join!
I did a starter pack of people in New York (City) working on ML/AI. Please distribute and feel free to self nominate!
go.bsky.app/BoEtagz
go.bsky.app/BoEtagz
January 31, 2026 at 4:00 AM
BTW I have not reposted this in a while and there are plenty of new AI people around. Ping to join!
Reposted by Marco
Some opinions about the Anthropic paper which I haven’t read:
1. It confirms all my priors
2. My interlocutors now look foolish
Probably don’t need to read it I guess
1. It confirms all my priors
2. My interlocutors now look foolish
Probably don’t need to read it I guess
January 31, 2026 at 3:53 AM
Some opinions about the Anthropic paper which I haven’t read:
1. It confirms all my priors
2. My interlocutors now look foolish
Probably don’t need to read it I guess
1. It confirms all my priors
2. My interlocutors now look foolish
Probably don’t need to read it I guess
Reposted by Marco
Wow. I set a goal of 50 new paid subscribers for Citation Needed, and you all helped me hit it in just over 24 hours. Thank you so much.
(If you were still hoping to get on board, subscriptions are, of course, still open.)
(If you were still hoping to get on board, subscriptions are, of course, still open.)
I’ve launched a Citation Needed membership drive to celebrate publishing my 100th recap issue!
Citation Needed critically covers cryptocurrency, the crypto industry, and its influence on policy. Paid subscriptions support much more than just the newsletter — here’s what goes into this work:
Citation Needed critically covers cryptocurrency, the crypto industry, and its influence on policy. Paid subscriptions support much more than just the newsletter — here’s what goes into this work:
Citation Needed membership drive
Celebrating 100 recap issues and sustaining critical independent coverage.
www.citationneeded.news
January 31, 2026 at 1:09 AM
Wow. I set a goal of 50 new paid subscribers for Citation Needed, and you all helped me hit it in just over 24 hours. Thank you so much.
(If you were still hoping to get on board, subscriptions are, of course, still open.)
(If you were still hoping to get on board, subscriptions are, of course, still open.)
CFP for the First Workshop on Formal Languages and Neural Networks!
"We welcome posters dicussing the formal expressivity, computational properties, and learning behavior of neural networks!"
Call for posters: flann.cs.yale.edu/cfp.html
Deadline: February 12, 2026
@pentagonalize.bsky.social
"We welcome posters dicussing the formal expressivity, computational properties, and learning behavior of neural networks!"
Call for posters: flann.cs.yale.edu/cfp.html
Deadline: February 12, 2026
@pentagonalize.bsky.social
FLaNN Workshop 2026
flann.cs.yale.edu
January 30, 2026 at 10:28 PM
CFP for the First Workshop on Formal Languages and Neural Networks!
"We welcome posters dicussing the formal expressivity, computational properties, and learning behavior of neural networks!"
Call for posters: flann.cs.yale.edu/cfp.html
Deadline: February 12, 2026
@pentagonalize.bsky.social
"We welcome posters dicussing the formal expressivity, computational properties, and learning behavior of neural networks!"
Call for posters: flann.cs.yale.edu/cfp.html
Deadline: February 12, 2026
@pentagonalize.bsky.social
Caltrain has this problem also. My wifi tends to work for 1-2 stops and then completely cuts out and no amount of power cycling, etc. can restore it.
Like I get that I can just tether, but I shouldn't have to, especially in Silicon Valley.
Korea is really unrivaled here.
Like I get that I can just tether, but I shouldn't have to, especially in Silicon Valley.
Korea is really unrivaled here.
I am convinced the GDP of Japan would rise significantly if the Shinkansen wifi was even moderately usable.
It is the perfect place to hack, other than that you can't access the internet at all >95% of the time.
It is the perfect place to hack, other than that you can't access the internet at all >95% of the time.
January 30, 2026 at 8:19 PM
Caltrain has this problem also. My wifi tends to work for 1-2 stops and then completely cuts out and no amount of power cycling, etc. can restore it.
Like I get that I can just tether, but I shouldn't have to, especially in Silicon Valley.
Korea is really unrivaled here.
Like I get that I can just tether, but I shouldn't have to, especially in Silicon Valley.
Korea is really unrivaled here.
I've been blessed with an influx of new followers (thanks!).
Since you are here, here are a few of my favorite posts.
Since you are here, here are a few of my favorite posts.
A lot of you followed me due to #NLP, but I like to post about #chess (especially computer chess), #programming (especially puzzles, code golf, etc), and machine learning.
And some less technical stuff like #Korean, #Esperanto, and #trains (mostly in Japan, just due to proximity).
And some less technical stuff like #Korean, #Esperanto, and #trains (mostly in Japan, just due to proximity).
January 29, 2026 at 6:00 AM
I've been blessed with an influx of new followers (thanks!).
Since you are here, here are a few of my favorite posts.
Since you are here, here are a few of my favorite posts.
Reposted by Marco
Hello!
This is a reminder that @cornelltech.bsky.social runs a Red Team Clinic that provides a *free* safety consultation to nonprofits / public sector orgs that are developing a public-facing AI tool and want to stress-test it for possible abuse vectors.
Applications welcome on a rolling basis:
This is a reminder that @cornelltech.bsky.social runs a Red Team Clinic that provides a *free* safety consultation to nonprofits / public sector orgs that are developing a public-facing AI tool and want to stress-test it for possible abuse vectors.
Applications welcome on a rolling basis:
‘Red team’ students stress-test NYC health department’s AI | Cornell Chronicle
People usually strive to be their true, authentic selves, but this fall, five master’s students at Cornell Tech adopted not only alter egos but also “bad intent,” in an effort to make AI safer for hea...
news.cornell.edu
January 28, 2026 at 9:33 PM
Hello!
This is a reminder that @cornelltech.bsky.social runs a Red Team Clinic that provides a *free* safety consultation to nonprofits / public sector orgs that are developing a public-facing AI tool and want to stress-test it for possible abuse vectors.
Applications welcome on a rolling basis:
This is a reminder that @cornelltech.bsky.social runs a Red Team Clinic that provides a *free* safety consultation to nonprofits / public sector orgs that are developing a public-facing AI tool and want to stress-test it for possible abuse vectors.
Applications welcome on a rolling basis: