AI Today — What's Moving

Learning Objectives

• Analyze real-world AI news through a strategic lens
• Identify market trends: pricing wars, regulation, hardware, and agents
• Connect current events to foundational AI concepts

---

Stories

1. DeepSeek Launches V3.2-Exp with Sparse Attention

DeepSeek released V3.2-Exp, featuring Sparse Attention that slashes API costs significantly. This is a direct signal that an AI inference pricing war has begun.

Implications:

• Open-weight models are driving costs toward zero
• Smaller players can now access frontier-level capabilities
• Incumbents (OpenAI, Anthropic) must justify premium pricing

2. US Senators Propose the AI Risk Evaluation Act

A bipartisan bill requiring mandatory risk assessments for advanced AI models before deployment.

Implications:

• Federal regulation is taking shape, potentially preempting state-level patchwork
• Compliance costs will become a factor in AI development
• May slow down releases but increase trust in deployed systems

3. Google Launches Search Live

Google merged voice, camera, and conversational AI into a single search interface — "Search Live." Point your camera at something, speak a question, and get an AI-generated answer.

Implications:

• Search is no longer text-only — multimodal input is the new norm
• Traditional SEO will need to adapt to voice + visual queries
• AI-native interfaces are becoming the default

4. Humain Horizon Pro — AI-Native Laptop

A dedicated AI laptop with a local model runtime built in, allowing users to run LLMs without cloud dependency.

Implications:

• Hardware is being rebuilt for AI-native workflows
• Local inference eliminates privacy concerns and latency
• May reshape the PC market if adoption scales

5. States Scramble to Regulate AI Therapy Apps

Illinois and Nevada have banned AI therapy apps outright, while other states are developing frameworks for AI mental health tools.

Implications:

• The regulatory landscape is fragmenting (state by state)
• AI in high-stakes domains faces the toughest scrutiny
• Builders must be aware of jurisdiction-specific rules

---

Key Themes

• Cost collapse: Open-weight models are driving inference costs toward zero
• Regulatory fragmentation: Federal vs. state regulation creates compliance complexity
• Hardware evolution: Devices are being rebuilt for AI-native experiences
• Agentic shift: The next paradigm is agents that act, not just chat

---

Discussion Questions

1. How does DeepSeek's pricing strategy affect the broader AI tool landscape?
2. What are the implications of fragmented state-level AI regulation for startups?
3. How does "Search Live" change the way users interact with information?
4. Would you use an AI therapist? What safeguards would you need?