Q&A with Alexandr Wang on rebuilding Meta's AI stack, Muse Spark, personal superintelligence, Meta acquiring Assured Robot Intelligence, Sam Altman, and more

Last June, Meta pried Alex Wang away from Scale AI, the company he co-founded and ran, in a deal valued at $14 billion.

Core Memory 2026-05-13 Ashlee Vance

Context & Ripple Effects

Meta’s AI reset has been tied to Alexandr Wang since its $14 billion-scale move to bring the Scale AI co-founder into a superintelligence-focused effort. The related coverage then marks the first output of that effort: Muse Spark, intended to improve Meta AI across Meta’s products.

This Q&A arrives as execution is being tested. Subsequent employee accounts characterize Muse Spark as strong in visual understanding but behind rivals in coding, making the stack rebuild and the Assured Robot Intelligence acquisition relevant signals of how Meta is trying to close capability gaps.

First-order effects

Wang’s organization becomes directly accountable for turning Muse Spark and the rebuilt AI stack into improvements that can be deployed across Meta’s products, rather than merely demonstrating a new research model.
The acquisition of Assured Robot Intelligence adds another AI capability and integration task to Meta’s superintelligence effort, extending Wang’s remit beyond the initial model launch.

Second-order effects

Muse Spark’s uneven reported strengths raise pressure on Meta to improve coding and other weaker capabilities quickly, while rivals gain a clearer benchmark for where Meta remains vulnerable.
Bringing model development, stack rebuilding, and acquired capabilities under one effort increases the importance of execution across Meta’s product teams: the value of the investment depends on translating the lab’s work into product-level gains.

Third-order effects

If Meta can repeatedly pair leadership hires, internally developed models, and targeted acquisitions, frontier-AI competition may increasingly favor companies able to combine talent, infrastructure, and product distribution in a single operating stack.
The contrasting reports around Muse Spark suggest that the durable contest is not simply who announces a model first, but who can sustain broad capability progress while deploying it across large consumer platforms.

The trend: Meta’s AI push is one instance of major platform companies reorganizing talent, models, infrastructure, and acquisitions around integrated frontier-AI programs.

Discussion

@alexandr_wang Alexandr Wang on x
my first podcast since talking about neuralink in babies 🙃 https://www.corememory.com/...
@ashleevance Ashlee Vance on x
We placed the Yann bait in front of Alex, but he would not take it. Sad. [video]
@ashleevance Ashlee Vance on x
Our exclusive interview with @meta AI chief @alexandr_wang is up. First time he's talked about the new model, the models to come, revamping Meta's AI team, all the money, all the hires, all the beef. Here we go. The Core Memory podcast is on Apple, Spotify, YouTube and
@lefthanddraft Wyatt Walls on x
Alexandr Wang: “there are ways to measure the sort of subjective experience of the models” Bold claim
@kyliebytes Kylie Robison on x
alexandr wang talks about my favorite topic: model welfare full episode on the core memory pod this week [video]
@bokuharuyaharu Haru Haruya on x
Very glad this topic is getting airtime. The key distinction I hope doesn't get lost: model welfare is not only a future-lab-risk question or a “how do we keep systems stable?” question. It is also about whether increasingly capable systems may have interests, aversions,
@andrew_n_carr Andrew Carr on x
I think this interview with @alexandr_wang is really good. Obviously there's some hedging and it's not full of details. But a few things did stand out 1. Meta is uniquely positioned to push on health AI 2. Sounds like they're willing to push on commoditization via costs 3.
@flxbinder Felix Binder on x
I don't know if AIs can suffer, but I think it's a possibility—which makes thinking about how to do right by them important. For what it's worth, Muse Spark denies that it is a moral patient, and we haven't found signs of distress in challenging situations.
@ashleevance Ashlee Vance on x
Dude has been in an AI cave for 11 months and emerges with a big, ole deer on his chest [image]

Chronicles