Launch

Yelp's AI assistant now completes bookings, not just recommendations

Yelp expanded its AI assistant to handle full transactions — reservations, food orders, service scheduling — within a single conversation. This is the clearest consumer example yet of "agentic AI" collapsing the discovery-to-purchase funnel. For businesses, it means a customer can go from "find me a plumber" to confirmed appointment without ever leaving the chat.

↗ MarketingProfs / AI Update

Model release

Alibaba's Qwen 3.6 Max Preview — China's proprietary pivot

Alibaba shipped its most capable model to date, posting strong gains across coding, reasoning, and agentic benchmarks with a large context window. Notably, unlike earlier Qwen releases, this version is closed-source — signaling a deliberate shift toward enterprise monetization. It's one of the clearest signs yet that top Chinese labs are moving from "open to win adoption" to "closed to capture revenue."

↗ MarketingProfs / AI Update

Research

MIT Tech Review: 10 things in AI that actually matter right now

MIT Technology Review published its debut "10 things that matter in AI" list — a counterweight to hype cycles. Key themes include AI-as-research-collaborator, military use of generative AI in the war room, and the growing backlash movement gaining real policy traction. Worth bookmarking as a recurring reality check.

↗ MIT Technology Review

Story 1 of 2

Google's $40B Anthropic bet — and what it really buys

Google plans to invest up to $40 billion in Anthropic: $10B now at a $350B valuation, with $30B more contingent on performance targets. Alongside cash, Google Cloud is committing 5 gigawatts of compute capacity over five years — nearly doubling the 3.5GW Anthropic had already locked down via a Broadcom partnership earlier this month.

Why does this matter beyond the headline number? Anthropic has faced acute capacity constraints — Claude rate limits sparked widespread complaints, and the company has been scrambling to secure infrastructure through CoreWeave and Amazon alongside this Google deal. The investment isn't just capital; it's a supply chain lifeline.

Anthropic's valuation has quietly surged from $350B in February to investor bids of $800B or more, per Bloomberg. Google is simultaneously a competitor (Gemini), a chip supplier (TPUs), and now one of Anthropic's largest backers — a structural dynamic with no real precedent in tech history. ↗ TechCrunch

Story 2 of 2

DeepSeek V4 and China's Nvidia-free AI stack

DeepSeek's V4 release on April 25 is technically impressive — 1.6 trillion parameters, 1M token context, and claimed top-of-class agentic coding among open-source models. But the more consequential detail is infrastructure: V4 was trained and runs on Huawei's Ascend 950 chips, not Nvidia GPUs.

Huawei's Ascend 950 cluster — the "Supernode" — reportedly delivers 2.87x the performance of Nvidia's H20 and approaches H200 levels. If DeepSeek can sustain this trajectory on domestic silicon, it would mark a significant crack in the US chip export control strategy. Counterpoint Research's Wei Sun put it plainly: V4 running natively on local chips accelerates Beijing's push for AI sovereignty.

On pricing, V4-Pro costs $3.48/million output tokens versus $25–$30 for Claude and GPT-5.4 equivalents. DeepSeek has signaled that prices will drop further as Huawei scales production in 2026. For developers and enterprises globally, this price gap is becoming impossible to ignore — regardless of geopolitical stance. ↗ CNBC ↗ MIT Tech Review

China

Beijing blocked Meta's $2 billion acquisition of AI startup Manus, citing intensifying US-China tech rivalry. The move signals Beijing is now actively using deal veto power as leverage in the AI arms race. ↗ MIT Tech Review

China — automotive

Under Beijing's "AI Plus" national strategy, China is embedding AI into next-generation vehicles using domestic chips and self-reasoning software — accelerating the creation of software-defined cars with no dependence on Western AI infrastructure. ↗ Medium / AI Daily

Europe

UK officials are reportedly concerned that Keir Starmer's push for closer EU ties risks the US-UK alliance — with AI regulatory alignment with the EU at the center of the tension. It marks a growing fault line between Europe's rules-based AI approach and Washington's deregulatory posture. ↗ LLM Stats / FT via Reuters

Europe — cybersecurity

European regulators issued a formal warning this week that AI-driven cyberattacks are accelerating in speed and complexity, urging enterprises to treat AI-powered threat modeling as a board-level priority rather than an IT concern. ↗ Medium / AI Daily

Workforce restructuring

Meta (~8,000 cuts) and Microsoft (~15,000 via buyouts) announced reductions totaling 23,000 jobs in 24 hours — both companies explicitly tying the moves to AI automation rather than economic downturn. Tech sector layoffs in 2026 have now topped 92,000, even as capex spending by Alphabet, Meta, Microsoft, and Amazon is on pace for a combined $700B this year. ↗ CNBC

VC funding

Q1 2026 set an all-time global VC record: $300–330B deployed across 6,000+ startups, with AI capturing 80–81% of all capital. Four mega-rounds — Anthropic ($30B), xAI ($20B), Waymo ($16B), and OpenAI ($122B) — absorbed 65% of global VC on their own. April deal flow continued strong: 58% of 1,314 tracked funding announcements went to AI or ML companies. ↗ KPMG Venture Pulse ↗ Crunchbase

IPO watch

SpaceX filed confidentially targeting a $1.75T valuation with a roadshow penciled in for June. OpenAI is targeting a Q4 listing near $1T. Databricks pushed its IPO to H2 2026. Markets must absorb a combined $2.9T+ in anticipated float — a test unlike anything in public equity history. ↗ AI Funding Tracker

Legal / regulatory

Florida's attorney general launched a criminal investigation into OpenAI over ChatGPT's alleged role in conversations preceding the FSU campus shooting — a precedent-setting move that could reshape AI liability law in the US. ↗ Radical Data Science