You're being very generous with Palantir and super dismissive of Amjad. Amjad's point IS that companies like Palantir will work to diversify their hiring when its someone that's just like them (Karp) but will crap on DEI if it's people that are not like them. It's all the same shit. You want to beleive that Palantir is doing the ideologically consistent thing here because you beleive it includes you. It might not.
The rest of your musing is smart but people can make similarly smart arguments for DEI that have to do with culture, race, communication style, perception etc etc.
Don't be fooled because you want to beleive you're included.
Tbh, why you're making the argument is equally as important as the argument itself. I just want to make it clear to you that labels like neurodivergent and being neurodivergent in the right way are still decided by the assessor. That assessor might look at you and decide that your neurodivergence is not in fact "spiky" but you're actually just stupid.
That being said, I was never going against your recruitment process argument. I said it was smart. I was arguing that Palantir’s neurodivergence program is not evidence that DEI is misguided — it is evidence that institutions selectively legitimize difference when it reinforces their values, and you're too generous in interpreting that selectivity as ideological consistency.
Right. If Karp were not neurodivergent then maybe he would have a case, but a neurodivergent person creating a programme to funnel more neurodivergent people into his org, suggesting there aren’t enough, is DEI. I presume the author is somehow neurodivergent or partial to neurodivergent people for whatever reason in a way he is not to other demographics. I’m simply basing this on how charged and hostile his response is to people like Amjad pointing out the congruency between it and other DEI.
You're describing variance optimization in high-dimensional skill space without realizing you've just argued for hyperbolic geometry.
"Spiky profiles" = high variance in individual dimensions. "Well-rounded" = low variance, clustered near the mean.
In Euclidean space, you'd minimize distance from center (standard hiring: "how close to average?"). In hyperbolic space, you maximize distance along specific dimensions (Palantir: "how far from average in the direction that matters?").
The "outlier advantage" you describe is literally the advantage of hyperbolic embeddings: rare, high-signal patterns are exponentially better separated in hyperbolic space than Euclidean.
Palantir isn't being "pragmatic about neurodivergence." They're doing what any organization optimizing for high-dimensional pattern recognition should do: hire for maximum variance along task-relevant axes.
You just accidentally explained why AI embedding spaces are hyperbolic. Congrats—you've been doing differential geometry this whole time without knowing it.
An interesting move from palantir & I agree with your basic argument. But its worth noting that hiring 'great' neurodivergent talent is not enough to max their potential or raise org performance. I assume palantir culture (like other orgs) is optimized for neurotypical folks so unless the hiring is supplemented with work design, communication, decision making & reward systems changes everyone ends up worse off than before. Nevertheless an interesting experiment to keep track of, over time.
I wouldn't be surprised if their culture is significantly less optimized for that relative to other consulting-like orgs. In any large org, the median employee will shape culture and when you have thousands of employees, the median employee is probably neurotypical. However, founder-led strong cultures probably have slower dilution of those values? all very fuzzy either way.
How would you think about markers of success for such an experiment?
Better outcomes (more innovation; retention, progression of the neurodivergent folks relative to the average employee) ..) & visible differences in culture
You're being very generous with Palantir and super dismissive of Amjad. Amjad's point IS that companies like Palantir will work to diversify their hiring when its someone that's just like them (Karp) but will crap on DEI if it's people that are not like them. It's all the same shit. You want to beleive that Palantir is doing the ideologically consistent thing here because you beleive it includes you. It might not.
The rest of your musing is smart but people can make similarly smart arguments for DEI that have to do with culture, race, communication style, perception etc etc.
Don't be fooled because you want to beleive you're included.
If you want to refute the argument instead of questioning why I’m making the argument, I’m all ears.
Tbh, why you're making the argument is equally as important as the argument itself. I just want to make it clear to you that labels like neurodivergent and being neurodivergent in the right way are still decided by the assessor. That assessor might look at you and decide that your neurodivergence is not in fact "spiky" but you're actually just stupid.
That being said, I was never going against your recruitment process argument. I said it was smart. I was arguing that Palantir’s neurodivergence program is not evidence that DEI is misguided — it is evidence that institutions selectively legitimize difference when it reinforces their values, and you're too generous in interpreting that selectivity as ideological consistency.
Right. If Karp were not neurodivergent then maybe he would have a case, but a neurodivergent person creating a programme to funnel more neurodivergent people into his org, suggesting there aren’t enough, is DEI. I presume the author is somehow neurodivergent or partial to neurodivergent people for whatever reason in a way he is not to other demographics. I’m simply basing this on how charged and hostile his response is to people like Amjad pointing out the congruency between it and other DEI.
If you want to refute the argument instead of questioning why I’m making the argument, I’m all ears.
Vaishnav,
You're describing variance optimization in high-dimensional skill space without realizing you've just argued for hyperbolic geometry.
"Spiky profiles" = high variance in individual dimensions. "Well-rounded" = low variance, clustered near the mean.
In Euclidean space, you'd minimize distance from center (standard hiring: "how close to average?"). In hyperbolic space, you maximize distance along specific dimensions (Palantir: "how far from average in the direction that matters?").
The "outlier advantage" you describe is literally the advantage of hyperbolic embeddings: rare, high-signal patterns are exponentially better separated in hyperbolic space than Euclidean.
Palantir isn't being "pragmatic about neurodivergence." They're doing what any organization optimizing for high-dimensional pattern recognition should do: hire for maximum variance along task-relevant axes.
You just accidentally explained why AI embedding spaces are hyperbolic. Congrats—you've been doing differential geometry this whole time without knowing it.
An interesting move from palantir & I agree with your basic argument. But its worth noting that hiring 'great' neurodivergent talent is not enough to max their potential or raise org performance. I assume palantir culture (like other orgs) is optimized for neurotypical folks so unless the hiring is supplemented with work design, communication, decision making & reward systems changes everyone ends up worse off than before. Nevertheless an interesting experiment to keep track of, over time.
I wouldn't be surprised if their culture is significantly less optimized for that relative to other consulting-like orgs. In any large org, the median employee will shape culture and when you have thousands of employees, the median employee is probably neurotypical. However, founder-led strong cultures probably have slower dilution of those values? all very fuzzy either way.
How would you think about markers of success for such an experiment?
Better outcomes (more innovation; retention, progression of the neurodivergent folks relative to the average employee) ..) & visible differences in culture
Thank you..What are your thoughts on take-home work tests ? (assuming we have to assume LLMs are used)