Google AI Calls Local Shops, China Accelerates, Tiny Teams Playbook

Beijing is pouring state money into every layer of the AI stack, from chips to data centers, to turn China into an AI superpower. Its decade-long industrial policy has already helped companies like DeepSeek and Alibaba release open-source models that rank among the world’s best.

Since 2014, the government has funneled nearly $100 billion into a semiconductor fund and in April pledged another $8.5 billion for early-stage AI start-ups. Local authorities offer subsidies, real estate and incubators such as Hangzhou’s Dream Town, while state labs collaborate with Alibaba, ByteDance and others to pool talent and data.

This state-backed, open-source push gives Chinese firms an alternative path around U.S. chip restrictions and positions them to influence global AI standards. OpenAI has publicly warned that Chinese models could shut American providers out of foreign markets, framing the contest as democratic versus authoritarian AI.

Anthropic launches Claude for Financial Services, a tailored version of its Claude for Enterprise platform aimed at regulated finance firms. The move marks the company’s first industry-specific offering.

The package adds higher rate limits, a prompt library, and pre-built MCP connectors to data sources like FactSet, PitchBook, S&P Capital IQ, and Morningstar. Existing Claude for Enterprise customers are not forced to migrate but can switch to access the new features.

The article notes regulated financial institutions have delayed AI adoption due to data-exposure and KYC worries. By bundling vetted data connectors and safety controls, Anthropic presents a single configurable platform that directly addresses those fears.

Google rolled out a free AI-powered calling feature in Search that phones local businesses to check prices, confirm availability and make bookings. The tool is available in 45 U.S. states, excluding Indiana, Louisiana, Minnesota, Montana and Nebraska.

Subscribers to the Google AI Pro and AI Ultra plans receive higher usage limits, and every call is monitored and recorded for quality control. Businesses can opt out, yet the system otherwise canvasses multiple providers and returns detailed options to the user.

The move deepens Google’s effort to keep search traffic from drifting to AI chatbots like ChatGPT and Perplexity, both of which are building browser-based assistants. Google also pushed Gemini 2.5 Pro and the new Deep Search capability to paying users, bundling advanced reasoning and multi-step research inside its core product.

Saudi Arabia has opened registration for its One Million Saudis in AI (SAMAI) program, targeting the upskilling of one million citizens. The initiative is led by the Saudi Data and Artificial Intelligence Authority with the Ministry of Education and the Ministry of Human Resources and Social Development.

Participants will receive training in data science, machine learning, practical AI applications, and ethical practices, including hands-on experience with modern tools. Open to all ages and professions, SAMAI—first announced at the 2023 Global AI Summit in Riyadh—is described as the largest program of its kind.

The move aligns with Vision 2030’s focus on human capital as the Kingdom’s prime asset. By scaling AI literacy nationwide, Saudi Arabia underscores its ambition to cement a leadership position in the global tech landscape.

Latent Space creator Shawn “swyx” Wang released the Tiny Teams Playbook after surveying standout micro-teams at the AI Engineers World’s Fair. The report distills lessons from seven companies that together run on just 100 employees yet generate $200 million in annual recurring revenue.

The playbook details hiring strategies such as paid work trials, top-of-market salaries and product-led recruiting, plus operating norms like near-zero meetings and AI “chief of staff” tooling. It spotlights lean outfits including Gamma, Gumloop and Bolt.new, some of which hit eight-figure ARR or serve tens of millions of users with fewer than 30 staff.

Wang asserts that AI-augmented tiny teams move faster because inter-human trust and communication, not headcount, limit execution speed. He frames the model as the next evolution of the org chart in a world progressing from Level 2 to Level 3 AGI.