Joel Chan
@joelchan86.bsky.social
Professing #hci #creativity #toolsforthought #metascience at UMD iSchool and HCIL
#firstgen #immigrant (to USA from Malaysia) academic
Also following #scicomm #openscience #sts #histsci #histtech
https://joelchan.me
#firstgen #immigrant (to USA from Malaysia) academic
Also following #scicomm #openscience #sts #histsci #histtech
https://joelchan.me
Pairs well with these two essays that rhyme closely with what I'll outline in my upcoming vision essay:
bengoldhaber.substack.com/p/a-knowledg...
www.shishyko.com/essays/publi...
bengoldhaber.substack.com/p/a-knowledg...
www.shishyko.com/essays/publi...
A Knowledge Commons for the 21st Century
Truth-seeking infrastructure at scale
bengoldhaber.substack.com
January 7, 2026 at 6:52 PM
Pairs well with these two essays that rhyme closely with what I'll outline in my upcoming vision essay:
bengoldhaber.substack.com/p/a-knowledg...
www.shishyko.com/essays/publi...
bengoldhaber.substack.com/p/a-knowledg...
www.shishyko.com/essays/publi...
Thinking about analogy of "recycling waste information" re: distributed new knowledge infrastructures for science, from Paul Edwards' classic paper:
journals.sagepub.com/doi/10.1177/...
(h/t @infotainment.bsky.social)
#commonknowledgefutures
journals.sagepub.com/doi/10.1177/...
(h/t @infotainment.bsky.social)
#commonknowledgefutures
January 7, 2026 at 6:51 PM
Thinking about analogy of "recycling waste information" re: distributed new knowledge infrastructures for science, from Paul Edwards' classic paper:
journals.sagepub.com/doi/10.1177/...
(h/t @infotainment.bsky.social)
#commonknowledgefutures
journals.sagepub.com/doi/10.1177/...
(h/t @infotainment.bsky.social)
#commonknowledgefutures
Reposted by Joel Chan
I want to create an incubator that funds atproto/bluesky experimental projects to the tune of $5k-$10k dev grants.
I am sorely tempted to call it “Chemtrails”, because it seeds the ATmosphere
Hit me with your best ideas!
I am sorely tempted to call it “Chemtrails”, because it seeds the ATmosphere
Hit me with your best ideas!
November 15, 2024 at 2:51 AM
I want to create an incubator that funds atproto/bluesky experimental projects to the tune of $5k-$10k dev grants.
I am sorely tempted to call it “Chemtrails”, because it seeds the ATmosphere
Hit me with your best ideas!
I am sorely tempted to call it “Chemtrails”, because it seeds the ATmosphere
Hit me with your best ideas!
Reposted by Joel Chan
bring back MySpace pages
now imagine if the bluesky pds included a tool for easily putting up a site on your handle, and it did this automatically, and it integrated well with leaflet or something
January 6, 2026 at 11:46 PM
bring back MySpace pages
Just learned of it last week from @atproto.science !
January 6, 2026 at 10:06 PM
Just learned of it last week from @atproto.science !
Lovely, thank you for sharing this!
January 6, 2026 at 10:04 PM
Lovely, thank you for sharing this!
Also curious for you to elaborate on the Unix legacy point!
January 6, 2026 at 8:50 PM
Also curious for you to elaborate on the Unix legacy point!
Oh good question. I think it's closer to plain text than, say, the XML under Word, but less plain text than Markdown? Maybe operative dimension is WYSWIG-ness for humans and portability across tools?
January 6, 2026 at 8:50 PM
Oh good question. I think it's closer to plain text than, say, the XML under Word, but less plain text than Markdown? Maybe operative dimension is WYSWIG-ness for humans and portability across tools?
I wonder if we're using diff. words but agree (I'm using dialect ~informally): I see Quarto's Markdown flavor (pandoc) as an extension/enhancement on top of Markdown (similar w/ MystMarkdown et al). And I think I'm fine with calling that a superset of Markdown.
January 6, 2026 at 8:48 PM
I wonder if we're using diff. words but agree (I'm using dialect ~informally): I see Quarto's Markdown flavor (pandoc) as an extension/enhancement on top of Markdown (similar w/ MystMarkdown et al). And I think I'm fine with calling that a superset of Markdown.
Fair points!
Perhaps value is starting in ~plain text is easy to grasp and tolerant of ~mess.
We're then doing the move you describe of supporting incremental formalization on top of plain text for richer structures.
Then delivering value in the plain text interface from the structure.
Perhaps value is starting in ~plain text is easy to grasp and tolerant of ~mess.
We're then doing the move you describe of supporting incremental formalization on top of plain text for richer structures.
Then delivering value in the plain text interface from the structure.
January 6, 2026 at 8:40 PM
Fair points!
Perhaps value is starting in ~plain text is easy to grasp and tolerant of ~mess.
We're then doing the move you describe of supporting incremental formalization on top of plain text for richer structures.
Then delivering value in the plain text interface from the structure.
Perhaps value is starting in ~plain text is easy to grasp and tolerant of ~mess.
We're then doing the move you describe of supporting incremental formalization on top of plain text for richer structures.
Then delivering value in the plain text interface from the structure.
It's tempting me to open the hellsite to see if those images are real. They probably are! But did FT make them with grok for this piece?? So many questions!
January 6, 2026 at 5:18 PM
It's tempting me to open the hellsite to see if those images are real. They probably are! But did FT make them with grok for this piece?? So many questions!
January 6, 2026 at 5:17 PM
in my work i'm seeing non-trivial scientific process-thinking (sensemaking, literature mapping) in markdown-based tools like @obsidian.md
the simple plain-text format is also enabling fun AI-collab workflows (e.g., w/ claude code) for navigating tradeoff of creative chaos and structure
the simple plain-text format is also enabling fun AI-collab workflows (e.g., w/ claude code) for navigating tradeoff of creative chaos and structure
January 6, 2026 at 4:53 PM
in my work i'm seeing non-trivial scientific process-thinking (sensemaking, literature mapping) in markdown-based tools like @obsidian.md
the simple plain-text format is also enabling fun AI-collab workflows (e.g., w/ claude code) for navigating tradeoff of creative chaos and structure
the simple plain-text format is also enabling fun AI-collab workflows (e.g., w/ claude code) for navigating tradeoff of creative chaos and structure
One recent example I liked: neural-reckoning.org/pub_comob_so...
Also some fun experimentation w/ @quarto.org at JOVI: www.journalovi.org/submit.html#...
And @rmarkdown.bsky.social has been pretty important for awhile for reproducible psych!
So, strictly speaking dialects of Markdown?
Also some fun experimentation w/ @quarto.org at JOVI: www.journalovi.org/submit.html#...
And @rmarkdown.bsky.social has been pretty important for awhile for reproducible psych!
So, strictly speaking dialects of Markdown?
Spiking neural network models of interaural time difference extraction via a massively collaborative process
Neuroscientists are increasingly initiating large-scale collaborations which bring together tens to hundreds of researchers. At this scale, such projects can ta...
neural-reckoning.org
January 6, 2026 at 4:51 PM
One recent example I liked: neural-reckoning.org/pub_comob_so...
Also some fun experimentation w/ @quarto.org at JOVI: www.journalovi.org/submit.html#...
And @rmarkdown.bsky.social has been pretty important for awhile for reproducible psych!
So, strictly speaking dialects of Markdown?
Also some fun experimentation w/ @quarto.org at JOVI: www.journalovi.org/submit.html#...
And @rmarkdown.bsky.social has been pretty important for awhile for reproducible psych!
So, strictly speaking dialects of Markdown?
agreed!
how do you see this overlapping if at all with our HCI discourse? i see some crossover in discourse with the AI shaping billions piece in CACM, but am v curious to see ideas like this in interactions/CACM/UIST, plus others like aicoding.leaflet.pub/3majnyfydzs2y
how do you see this overlapping if at all with our HCI discourse? i see some crossover in discourse with the AI shaping billions piece in CACM, but am v curious to see ideas like this in interactions/CACM/UIST, plus others like aicoding.leaflet.pub/3majnyfydzs2y
Regenerative Software - The Phoenix Architecture
aicoding.leaflet.pub
January 6, 2026 at 4:46 PM
agreed!
how do you see this overlapping if at all with our HCI discourse? i see some crossover in discourse with the AI shaping billions piece in CACM, but am v curious to see ideas like this in interactions/CACM/UIST, plus others like aicoding.leaflet.pub/3majnyfydzs2y
how do you see this overlapping if at all with our HCI discourse? i see some crossover in discourse with the AI shaping billions piece in CACM, but am v curious to see ideas like this in interactions/CACM/UIST, plus others like aicoding.leaflet.pub/3majnyfydzs2y
Really do *not* like the Mediated Future vision (which seems to be dominant in the AI scientist visions). Heartened that I can imagine mapping productive+sustainable business models to the other more palatable/resonant visions!
January 6, 2026 at 4:34 PM
Really do *not* like the Mediated Future vision (which seems to be dominant in the AI scientist visions). Heartened that I can imagine mapping productive+sustainable business models to the other more palatable/resonant visions!
Reposted by Joel Chan
1. My new preprint has its own bluesky account. Why? The problems facing social media & scientific publishing are similar: both are dominated by powerful oligopolies. The @atproto.com tech underlying bluesky that aims to solve the social media prob might also help solve the scientific pub prob 🧪 🧵
1. Preprint: Menopause averted a midlife energetic crisis with help from older children and parents: A simulation study. zenodo.org/records/1814...
Menopause is rare, known to occur only in humans and toothed whales: 🧵
Menopause is rare, known to occur only in humans and toothed whales: 🧵
January 6, 2026 at 3:12 PM
1. My new preprint has its own bluesky account. Why? The problems facing social media & scientific publishing are similar: both are dominated by powerful oligopolies. The @atproto.com tech underlying bluesky that aims to solve the social media prob might also help solve the scientific pub prob 🧪 🧵
January 6, 2026 at 4:28 PM
Like, I could see in this example, subbing out "based on" -->
"The catalog UI regenerates aggressively. AI rewrites for users based on their own unique preferences and design iterations. Failures are visible immediately and recoverable by rollback. The blast radius is one user's session."
"The catalog UI regenerates aggressively. AI rewrites for users based on their own unique preferences and design iterations. Failures are visible immediately and recoverable by rollback. The blast radius is one user's session."
January 6, 2026 at 3:58 PM
Like, I could see in this example, subbing out "based on" -->
"The catalog UI regenerates aggressively. AI rewrites for users based on their own unique preferences and design iterations. Failures are visible immediately and recoverable by rollback. The blast radius is one user's session."
"The catalog UI regenerates aggressively. AI rewrites for users based on their own unique preferences and design iterations. Failures are visible immediately and recoverable by rollback. The blast radius is one user's session."
I could see them as complementary, though, in that the "fast pace layers" could just as well be specified by the user in a malleable software sense, but draw on more durable shared "slow pace" layers?
January 6, 2026 at 3:55 PM
I could see them as complementary, though, in that the "fast pace layers" could just as well be specified by the user in a malleable software sense, but draw on more durable shared "slow pace" layers?
Reposted by Joel Chan
On the Slow Death of Scaling by Sarah Hooker
"For the last decade, it has been hard to stray off the beaten path of accepted wisdom that scaling training parameters drives innovation.
However, the relationship between training compute + performance is uncertain + rapidly changing."
"For the last decade, it has been hard to stray off the beaten path of accepted wisdom that scaling training parameters drives innovation.
However, the relationship between training compute + performance is uncertain + rapidly changing."
January 6, 2026 at 2:21 PM
On the Slow Death of Scaling by Sarah Hooker
"For the last decade, it has been hard to stray off the beaten path of accepted wisdom that scaling training parameters drives innovation.
However, the relationship between training compute + performance is uncertain + rapidly changing."
"For the last decade, it has been hard to stray off the beaten path of accepted wisdom that scaling training parameters drives innovation.
However, the relationship between training compute + performance is uncertain + rapidly changing."
right, something like
[AI codegen being sufficiently good] ---enables--> [malleable software as form of end-user programming]
[AI codegen being sufficiently good] ---probably will require for production/complex applications / enables new paradigm of--> [regenerative software]
?
[AI codegen being sufficiently good] ---enables--> [malleable software as form of end-user programming]
[AI codegen being sufficiently good] ---probably will require for production/complex applications / enables new paradigm of--> [regenerative software]
?
January 6, 2026 at 3:51 PM
right, something like
[AI codegen being sufficiently good] ---enables--> [malleable software as form of end-user programming]
[AI codegen being sufficiently good] ---probably will require for production/complex applications / enables new paradigm of--> [regenerative software]
?
[AI codegen being sufficiently good] ---enables--> [malleable software as form of end-user programming]
[AI codegen being sufficiently good] ---probably will require for production/complex applications / enables new paradigm of--> [regenerative software]
?
Will at least share a @leaflet.pub on this! :)
January 6, 2026 at 3:45 PM
Will at least share a @leaflet.pub on this! :)