The Archive

So ... what's your In the Weights score?

Atlantic reporter Alex Reisner recently uncovered four datasets of music being used to train AI models and made them fully searchable for the public. Two of the sets are absolutely enormous at 12 million and 9 million tracks. The other two are much smaller, but still represent a significant amount of training data at over 100,000 songs each. According to Reisner, the sets have been downloaded thousands of times and, while it's impossible to know exactly who has used them, Google and Stability have both confirmed they have in research papers. Some of the sources, like the Free Music Archive da...

Jumper isn't the only big name leaving Google DeepMind.

a quiet day lets us promo AIE one last time

The real valuable capability MCP offers over skills/CLI is isolating the auth flow outside of the agent’s context window, and potentially out of the harness completely. [...] Maybe the idealized form of MCP is just an auth gateway for the API and nothing else. That’d still be a win. — Sean Lynch , comment on Hacker News Tags: model-context-protocol , llms , ai , generative-ai , skills

For the last 30 years, stopping the flow of cybersecurity-related software has proven to be ineffective. It's unclear why it would work now with Anthropic’s cybersecurity model Mythos.

The best Stratechery content from the week of June 15, 2026, including Anthropic, e-commerce in the age of AI, and the NBA Finals being a perfect 10.

Just as last week was ending, the US government forced Anthropic to pull its two newest models, Fable 5 and Mythos 5, citing national security concerns after Amazon researchers allegedly found a way to bypass Fable 5’s guardrails. Cybersecurity researchers have since signed an open letter calling the move dangerous, and Anthropic itself noted the same jailbreaks exist in other models. So is […]

Reliance is weaving AI into telecom services used by more than 500 million people.

Luca Guadagnino's film about OpenAI CEO Sam Altman, Artificial, has reportedly been dropped by Amazon MGM. The film, which stars Andrew Garfield and covers the rollercoaster five days in 2023 spanning Altman's termination and reinstatement as CEO, had been in the works for about a year. The cast also includes A Complete Unknown actress Monica Barbaro as OpenAI CTO Mira Murati, Ike Barinholtz as Elon Musk, and Anora's Yura Borisov as OpenAI chief scientist Ilya Sutskever. In a statement to Deadline, the studio said it believes the movie "will be better served if it were released by a different...

Call it a startup with a sole founder and a very large seed round, but what's next is less clear.

Miami-based AI startup Subquadratic came out of stealth mode last month with a huge claim. It announced that it had solved a mathematical bottleneck that had been holding back large language models for almost a decade. The details were thin, and many people were unconvinced. But Subquadratic has started to bring the receipts, sharing the…

There's a commercial logic that cuts against the idea that ASML would risk its export license to arm a Chinese customer.

With GLM-5.2 passing everyone's vibe check, the open models story finally becomes a real frontier story.

Five months after returning to OpenAI, Barret Zoph - the company's head of enterprise AI sales - has departed, The Verge has learned. Zoph returned to OpenAI in mid-January after a stint as co-founder and CTO of Thinking Machines Lab, the competing AI company founded by former OpenAI CTO Mira Murati. Shortly after Zoph returned to OpenAI, the company said he would lead its push into enterprise - a significant role at OpenAI, since in recent months it had vowed to stop chasing so-called "side quests" and focus on key revenue drivers like enterprise and coding ahead of its planned IPO. OpenAI c...

DeductiveAI, a startup that uses AI to catch and resolve bugs in software, was founded just three years ago.

Today we launched a new plugin for Datasette, datasette-apps , with this launch announcement post on the Datasette project blog. That post has the what , but I'm going to expand on that a little bit here to provide the why . The TL;DR Datasette Apps are self-contained HTML+JavaScript applications that run in a tightly constrained <iframe> sandbox hosted on your Datasette application. They can use JavaScript to run read-only SQL queries against data in Datasette, and can run write queries too if you configure them with some stored queries . Here's a very simple example and a more complex custo...

Taiwan's drone spending plans for defense could also boost business overseas.

Startup Baseten is reportedly close to finalizing a $1.5 billion round at a $13 billion as the “inference gold rush" marches on.

The Snapchat maker is spinning off yet another internal unit. Dotmo will be comprised of current Snap staff who are leaving the social media company to focus on AI video development.

OpenAI is bulking up before its IPO, landing Transformer co-inventor Noam Shazeer from Google DeepMind and former Trump AI policy official Dean Ball in the same week.

Release: datasette-acl 0.6a0 This release expands datasette-acl from table-only permissions toward a general resource-sharing system. Alex Garcia did most of the work for this release - we're fleshing out the plugin that will allow multi-user Datasette instances finely grained control over who can access which resources within Datasette. Tags: datasette , alex-garcia

About 47% of singles look negatively at the use of AI in dating -- but, many dating app users are open to AI helping with profile punch-ups and conversation starters.

AWS is in talks to sell its chips to other data centers. CEO Andy Jassy has said this represents a $50 billion opportunity for the company.

LLM reasoning transparency is a critical affordance for understanding model decisions, mitigating misuse and misalignment, and debugging surprising model behaviors. However, DiffusionGemma performs a larger fraction of its computation in a continuous latent space; does this make its reasoning less transparent? We study this question by decomposing transparency into two components: variable transparency, whether we understand intermediate snapshots of a model's computational state; and algorithmic transparency, whether we can use these snapshots to reconstruct the process by which the model ar...

Egocentric video understanding is inherently limited by the narrow perspective of wearable cameras: a single viewpoint, a single modality, a single model cannot capture the full richness of human action. We argue that a truly expressive egocentric representation must subsume complementary knowledge across viewpoints, modalities, and foundation model representations, yet remain deployable from egocentric video alone. To this end, we introduce a hierarchical multi-teacher distillation framework that produces UNIEGO, a unified egocentric encoder trained with nine teachers spanning ego-exo viewpo...

A model is multicalibrated on a collection of group weights $G$ if it is calibrated -- i.e. unbiased even conditional on its prediction -- not just overall, but also after reweighting contexts by each $g \in G$. It is a useful property for many downstream applications and is a basic desideratum of trustworthy machine learning. Before this work, all predictors known to attain the minimax-optimal $\widetilde O(\varepsilon^{-3})$ sample complexity rate for $\varepsilon$-multicalibration were randomized, while deterministic predictors were known only with substantially worse sample complexity. Wh...

Generative recommendation is an emerging paradigm that has shown promise in industrial recommendation systems, aiming to predict users' next interactions from their historical behaviors. At the core of generative recommendation lies item tokenization, which bridges item semantics and recommendation models. However, existing methods often struggle to effectively organize and inject complex user-behavioral and item-semantic contexts into recommendation models simultaneously. On the one hand, existing graph-based integration methods, such as graph serialization and graph neural networks, either ...

We place the attention token on the group: a token is an element $g_i$ of a matrix Lie group $G$ -- a bare transformation, with no feature payload and no external action $ρ(g)$ carrying it. To our knowledge this is the first attention construction whose tokens are bare matrix Lie group elements: their score is the closed-form algebra norm of the relative pose rather than a learned kernel, and it reaches the affine full-frame groups that every irrep- or surjective-exp-based method must exclude. We call it Lie-Algebra Attention. Once tokens are group elements, the rest follows with none of the ...