Novo Nordisk Deploys GPT to Slash Timelines
Novo Nordisk announced a company‑wide partnership with OpenAI to embed the new GPT‑Rosalind life‑sciences model across its entire drug‑development pipeline. The AI system will be used for multi‑omics target identification, hypothesis generation, experimental design, and even manufacturing and supply‑chain operations, with pilots already running and full rollout planned by the end of 2026. The move is a direct response to a looming $230 billion U.S. patent cliff that threatens the revenue of its flagship obesity and diabetes drugs. Executives framed the deployment as an existential step to stay ahead of rivals such as Eli Lilly.
Google Accelerates Gemma 4 by 3x with Multi-Token Prediction
On May 5, Google launched Multi‑Token Prediction (MTP) drafter models for the entire Gemma 4 family, including the 31 B dense and 26 B Mixture‑of‑Experts variants. The drafters use speculative decoding to deliver up to three‑fold faster inference on consumer GPUs and Apple Silicon...
Google Cloud Hits $20B Revenue in Q1 with 63%
Google Cloud posted $20.03 billion in Q1 2026 revenue, a 63% year‑over‑year increase driven by surging demand for its Gemini Enterprise AI stack. Generative AI product usage jumped nearly 800% YoY, while token processing volume rose 60% quarter‑over‑quarter. The cloud backlog more...
Meta’s AI Ad Engine Delivers 33% Revenue Growth
Meta reported Q1 2026 revenue of $56.3 billion, a 33% year‑over‑year jump driven almost entirely by AI‑enhanced ad ranking models. The upgrades—doubling user‑sequence length, accelerating same‑day post indexing, and richer content embeddings—lifted ad impressions 19% and raised average CPM 12%, adding roughly...
Tencent Open-Sources Hy3 at 74.4%
Tencent unveiled the Hy3 preview on April 23, a 295‑billion‑parameter mixture‑of‑experts (MoE) model that activates only 21 billion parameters per token. On the SWE‑bench Verified benchmark it achieved a 74.4% score, a 40‑point jump from its Hy2 predecessor and approaching the performance...
Google Launches Gemini Enterprise Agent Platform
Google announced the Gemini Enterprise Agent Platform at Cloud Next 2026, retiring the Vertex AI brand. The new platform unifies model training, inference, orchestration, memory, observability and governance, targeting enterprise‑grade AI agents. In Q1 2026 it logged 40% quarter‑over‑quarter growth, reaching...
Claude Opus 4.7 Fuels Design Revolution
Anthropic unveiled Claude Design in a research preview on April 17, a visual creation tool built on the Claude Opus 4.7 model. The service lets users generate slide decks, one‑pagers, and branded assets from natural‑language prompts, with exports to PDF,...
Ising Models Redefine Quantum Error Correction
NVIDIA unveiled Ising, the first open‑source AI model suite built for quantum computing, on April 14‑15, 2026. The family includes a 35‑billion‑parameter vision‑language model that shrinks quantum‑processor calibration from days to hours, and a 3‑D CNN decoder that speeds error‑correction...
Open-Sourced a 24/7 AI Research Lab
Andrej Karpathy released the 630‑line autoresearch script on GitHub, enabling an AI agent to run roughly 700 autonomous ML experiments over two days. The loop cut the “Time‑to‑GPT‑2” benchmark from 2.02 to 1.80 hours, an 11% efficiency gain on a...