/AI24d ago

Early AI labs barred managers from paper authorship

Early AI lab leadership maintained explicit rules preventing managers and team leads from appearing on papers without material contributions. These standards persisted after the key figure who established them departed. In AI organizations where social influence and compute resources shape progress, credit is frequently assigned afterward to prominent figures through authorship, scaling access, and public visibility, even when direct involvement in the work remains limited.

1827137724.6K
Stella Biderman
Susan Zhang
Lucas Beyer (bl16)#55
Original post
Susan Zhang@suchenzang#59inAI

in a (social/compute)-capital intensive game, there's a funny dynamic that happens

where "politically-influential" figures in an org are bribed with credit attribution after the fact (authorship on paper, compute resources for scaling up, public viz on podcasts/blogs, etc) in order for a "breakthrough" to land/be released/be published

of course these figures became influential back in the day for some other reason, and can in theory be competent themselves as well

but at this point they generally have nothing to do with the successes/failures they talk about, which gets exponentially worse the longer their tenure becomes

maybe this is just how the way the world works with a rich-get-richer fame effect (after all, they can still attract new fresh talent who believe their public image hook, line, and sinker)

but i sometimes wonder if an entire economy / tech-bubble ends up being built on stolen valor*

what the long-term consequences may be

(maybe nothing!)

-----

*those not built on stolen valor have the other hairy problem of creative financial and legal engineering that is way beyond my ability to speculate on

9:19 PM · May 15, 2026 · 3 Views
Sentiment

Many users expressed disgust at the hero worship in tech, where AI organizations attribute breakthrough credit to influential figures after the fact.

Pos
0.0%
Neg
100.0%
1 comments with sentiment.
Cluster Engagement
Posts from X
Most Activity
Most Activity
VIEWS15.9KBOOKMARKS63LIKES171REPLIES13
Susan Zhang@suchenzang

in a (social/compute)-capital intensive game, there's a funny dynamic that happens

where "politically-influential" figures in an org are bribed with credit attribution after the fact (authorship on paper, public viz on podcasts/blogs, etc) in order for a "breakthrough" to land/be released/be published/be given compute resources for scaling up

of course these figures became influential back in the day for some other reason, and can in theory be competent themselves as well

but at this point they generally have nothing to do with the successes/failures they talk about, which gets exponentially worse the longer their tenure becomes

maybe this is just how the way the world works with a rich-get-richer fame effect (after all, they can still attract new fresh talent who believe their public image hook, line, and sinker)

but i sometimes wonder if an entire economy / tech-bubble ends up being built on stolen valor*

what the long-term consequences may be

(maybe nothing!)

-----

*those not built on stolen valor have the other hairy problem of creative financial and legal engineering that is way beyond my ability to speculate on

24dViews 15.9KLikes 171Bookmarks 63
RETWEETS2
Lucas Beyer (bl16)@giffmana

@suchenzang Early Samy's brain was amazing because managers and team leads would explicitly NOT go on papers if they didn't materially contribute. This left a strong impression on me. Done of us fought hard to keep this culture alive for long after he left, but then eventually times changed.

Susan Zhang@suchenzang

in a (social/compute)-capital intensive game, there's a funny dynamic that happens

where "politically-influential" figures in an org are bribed with credit attribution after the fact (authorship on paper, public viz on podcasts/blogs, etc) in order for a "breakthrough" to land/be released/be published/be given compute resources for scaling up

of course these figures became influential back in the day for some other reason, and can in theory be competent themselves as well

but at this point they generally have nothing to do with the successes/failures they talk about, which gets exponentially worse the longer their tenure becomes

maybe this is just how the way the world works with a rich-get-richer fame effect (after all, they can still attract new fresh talent who believe their public image hook, line, and sinker)

but i sometimes wonder if an entire economy / tech-bubble ends up being built on stolen valor*

what the long-term consequences may be

(maybe nothing!)

-----

*those not built on stolen valor have the other hairy problem of creative financial and legal engineering that is way beyond my ability to speculate on

24dViews 5.7KLikes 59Bookmarks 7
Susan Zhang@suchenzang

was never the case everywhere i've been at by the time i got there

some got credit for "executive support behind the scenes" (might've been too beneath them to contribute any hands-on execution time)

some were for publication approval

some are so legendary that every accomplishment from thousands of people around them are attributed to their own personal doing (which, relatively speaking, might not be the worst thing)

it's pretty much the name of the game to bribe and do whatever you need to do to lube up landing your work in such a capital-intensive field

but at some point, maybe it's worth playing a different game with a different set of rules...

Lucas Beyer (bl16)@giffmana

@suchenzang Early Samy's brain was amazing because managers and team leads would explicitly NOT go on papers if they didn't materially contribute. This left a strong impression on me. Done of us fought hard to keep this culture alive for long after he left, but then eventually times changed.

24dViews 2.5KLikes 31Bookmarks 7
Lucas Beyer (bl16)@giffmana

@suchenzang Yeah I've seen all of these you mention and hate it. Early brain was a pretty magical exception. It was exactly as things should always be.

Susan Zhang@suchenzang

was never the case everywhere i've been at by the time i got there

some got credit for "executive support behind the scenes" (might've been too beneath them to contribute any hands-on execution time)

some were for publication approval

some are so legendary that every accomplishment from thousands of people around them are attributed to their own personal doing (which, relatively speaking, might not be the worst thing)

it's pretty much the name of the game to bribe and do whatever you need to do to lube up landing your work in such a capital-intensive field

but at some point, maybe it's worth playing a different game with a different set of rules...

24dViews 452Likes 10Bookmarks 0
mentsh@_mentsh

@suchenzang this is also the exact dynamic that happens in PhD labs headed by famous professors

24dViews 93Likes 1
jiahao@__endif

@suchenzang in some sense it's about getting the power to allocate resources or becoming irrelevant

24dViews 310Likes 3
Forward Unemployed Engineer@waltuuuhr

@suchenzang i wish you would name names but i understand why you dont. but it'd make it a lot more spicy

24dViews 194Likes 2
Joey@joey_f6

@suchenzang I wonder who this is in reference to

24dViews 153Likes 2
iLagrangian@i_Lagrangian

@suchenzang The meme question “What did so-and-so see?” can in most cases be answered with “Not very much”

24dViews 149Likes 2
Strata@ChainZenit

@suchenzang oh this has big committee energy, what sparked the observation?

24dViews 234Likes 1
Thomas Tao@Thomas_Tao_1

@suchenzang Yeah. This leaks into product teams too. Once credit becomes bargaining currency, people stop saying what is actually ready.

24dViews 109Likes 1
Martian@space_colonist

@suchenzang the amount of hero worship in tech is kinda disgusting

24dViews 27
Stella Biderman@BlancheMinerva

@qberthet @LucaAmb @octonion This is non-hypothetical, it’s the contribution I had to a recent paper: http://deepignorance.ai/

I could definitely provide this level of input and support on 15 papers in a year. So why shouldn’t I be a co-author on them?

Stella Biderman@BlancheMinerva

@qberthet @LucaAmb @octonion If I shape – or choose – the research question, design the experiments, help analyze the data, and help write the paper that doesn’t count as deserving authorship in your book?

20dViews 19Likes 0Bookmarks 0
Stella Biderman@BlancheMinerva

@qberthet @LucaAmb @octonion If I shape – or choose – the research question, design the experiments, help analyze the data, and help write the paper that doesn’t count as deserving authorship in your book?

Quentin Berthet@qberthet

@LucaAmb @BlancheMinerva @octonion I personally think that this is too low a bar for authorship.

24dViews 12Likes 0Bookmarks 0
Maya N@mayasolos

@suchenzang The funny part is we can debate model attribution for hours, but human attribution in labs is still mostly vibes, politics, and compute budgets.

24dViews 10
mentsh@_mentsh

@suchenzang my guess is that this probably occurs more generally in all/most human organisations where buy-in from politicly connected individuals is required to do things

24dViews 10