Story cards stay below the sticky dock while audio, chapters, date, and brief navigation remain accessible.
Snap CEO Evan Spiegel cut roughly 1,000 employees, calling it a "crucible moment" driven by rapid AI advancement. AI agents now generate over 65% of Snap's new code, handle over 1 million internal support queries monthly, and flag 7,500+ bugs via automated code review. The restructuring is expected to save $500 million annually by H2 2026, and the stock went up on the news.
DeepSeek's V4 model — a 1-trillion-parameter MoE with only 32–37B active parameters per query — is targeting a late April launch after multiple delays. The geopolitical kicker: Reuters confirmed V4 will run on Huawei Ascend 950PR chips, demonstrating China's AI stack can operate independently of U.S. silicon. The model will be released under the Apache 2.0 license.
EY announced a global rollout of enterprise-scale agentic AI across its entire Assurance division — 130,000 professionals conducting 160,000 audits in 150+ countries. This is one of the largest single deployments of agentic AI in a professional services firm, signaling that AI agents have moved from pilot to production at enterprise scale.
Danish pharma giant Novo Nordisk announced a strategic partnership with OpenAI to deploy AI across drug discovery, clinical trials, manufacturing, and supply chains. Pilot programs launch immediately with full integration targeted by year-end. The deal is driven by Novo's urgent race against Eli Lilly in the weight-loss drug market.
OpenAI closed a $122 billion funding round at an $852 billion valuation — the largest private financing in Silicon Valley history — and is now generating $2 billion per month in revenue. The company is laying groundwork for a Q4 2026 public listing targeting a $1 trillion valuation, with CFO Sarah Friar confirming retail investor allocation.
Meta extended its Broadcom partnership to deploy multiple gigawatts of custom MTIA accelerators, with deployment starting H2 2026 and scaling to 10GW by 2029. The MTIA chips will be the first AI silicon manufactured on a 2nm process, with four new chip generations planned in two years — an unusually aggressive cadence aimed at reducing Nvidia dependence.
Google released Gemma 4, a family of open models under Apache 2.0 with variants from 2B to 31B parameters. The 31B dense model claimed third place on Arena AI's text leaderboard, beating models with 400B+ parameters. With 256K context, native vision/audio, and 140+ language support, it represents a new high-water mark for open-weight efficiency.
