AI News
Our editorial picks on the AI stories that matter for Builders.
Meta's Muse Glimmer Puts a 30B Agent Model on Your Laptop
Meta released Muse Glimmer, a 30B agent model, under Apache 2.0. The number that matters is not 30B — it is the 24-32 GB memory envelope, and what it changes for builders in India and the UK.
Sequoia's $1B Nuclear Bet on AI Data Centre Power
Valar Atomics raised $1B led by Sequoia plus a $200M credit facility, weeks after its Ward 250 reactor went critical and powered a Blackwell GPU. What is proven, financed and unproven.
AMD Buys Taalas, the Chip That Hardwires One Model in Silicon
AMD has agreed to acquire Toronto's Taalas, which builds inference chips hard-wired for a single model. Terms were not disclosed. The bet underneath it, and what it means for builders.
Gemini 3.6 Flash Hit GA in July: Google's Subagent-Tier Play
Gemini 3.6 Flash and 3.5 Flash-Lite hit general availability in July with cheaper output and stronger agentic planning. Where the new Flash tier fits in your subagent stack — with verified pricing.
Indian AI Startup Funding Up 4x in H1 2026 — Still a Rounding Error?
Inc42 counts $676M across a record 57 deals for Indian AI-native startups in H1 2026 — over 4x year-on-year. London raised ~$12B in seven months. The honest maths on the gap, for AI builders.
Claude Code Turns Auto Mode On by Default: What Changes 14 August
From 14 August 2026, Claude Code stops asking permission at every step on Pro, Max and Team plans. The new blast-radius rules, the opt-out, and what UK and Indian teams should do this week.