Topic
Large pre-trained models built for non-text data such as time series, satellite imagery, and medical scans, reusable across tasks with little retraining.
Anonymous early Anthropic backer expects shares to price near $1.8 trillion and trade at $2.3 to $2.4 trillion, Fortune reported. That would clear SpaceX's $2.1 trillion opening-day close by $0.2 trillion to $0.3 trillion, but it is one holder's forecast and the S-1 is unpublished.
Reality
Mistral released Timestral, four CPU-capable forecasting models, with weights licensed only to firms at or under $20M in monthly revenue. The license text, as DEV Community quotes it, tests the preceding month, so eligibility can change monthly.
Reality
Starling will train a transactional foundation model on more than 15 billion anonymised transactions in Civo's UK sovereign cloud, with NVIDIA support. Production is due in early 2027, fraud detection first. The blueprint and chips are NVIDIA's, so the proprietary part is the data.
Perspective Coverage
5 publishersReality
Mark Zuckerberg said Meta committed $100 billion to $200 billion to AI infrastructure before knowing whether the technology would work. Investors are underwriting it through 2026 capex guidance of $130 billion to $145 billion, up from $72 billion.
Perspective Coverage
5 publishersReality
Meta says its central GEM ads model now learns from every ad surface and has delivered an 18.9% conversion gain. The post says a model that size is too costly to serve live, so the gains pass to other models through post-training.
Reality
AMIFound, a single model, outperformed 38 baselines at restoring degraded medical images across seven modalities, according to its authors. The downstream results are what make a shared restoration step ahead of clinical AI plausible.
Reality
A Fortune column says SaaS incumbents answered collapsed multiples with record debt and buybacks presented as AI strategy. It quantifies the de-rating precisely and leaves the debt unsized at the five companies it names.
Reality
Existing 2D/3D registration models align well for some patients and fail for others. MIT's xvr, out today in Nature, fits itself to one patient in about five minutes, then matches that patient's X-rays to their 3D scan in seconds.
Reality
Exein's valuation is thirty times what it was two years ago on technology already inside more than two billion devices, and the foundation model that telemetry is meant to train is not due until the first quarter of 2027.
Reality
TranscriptFormer learned gene expression the way a language model learns text, across 12 species from single-celled yeast to humans. Its cross-species matches arrive as hypotheses a bench experiment still has to test.
Reality
Chen and colleagues report that a single text-prompted model scored highest overall across 16 pathology test sets, measured against baselines their code list names. The abstract publishes the counts; the margins are left out.
Reality
Google's new forecaster is 330m parameters trained on a trillion time points. In zero-shot multivariate mode it anticipates a 20% promotion lift that a univariate model misses. That is most of what an in-house demand model does.
Reality
Google's new 330 million parameter forecaster takes covariates and related series zero-shot in one forward pass, which puts a lot of bespoke per-SKU plumbing up for review, assuming the accuracy transfers to your own series.
Reality