1. NVIDIA releases Nemotron-Labs-3-Puzzle-75B-A9B: compressed hybrid MoE LLM with 2.03× server throughput — Efficiency-focused model matching user throughput while doubling server throughput for inference at scale. Marktechpost · Jul 9
2. Martin Fowler: Viability of Local Models for Coding — Thoughtworks engineer Birgitta Böckeler returns to local models after years of disappointment, examining when they're actually viable for real development work. Martin Fowler · Jul 8
3. Databricks benchmarks coding agents on multi-million-line production codebase — Internal evaluation of AI coding agents against real-world enterprise code, not toy benchmarks. Databricks · Jul 8
4. Indeed: AI and Job Postings — From Destruction to Creation? — Agentic AI may be flipping the relationship between AI exposure and job posting growth; software dev postings are rebounding. Indeed Hiring Lab · Jul 8
5. Meta open-sources Astryx: customizable, agent-ready design system — Component library built with AI coding agents as first-class consumers. GitHub · Jul 9
6. LLM-as-a-Verifier: general-purpose verification framework (arXiv 2607.05391) — Using LLMs as the verification layer for autonomous agent outputs, replacing bespoke verification systems. arXiv · Jul 9
7. Ember: lightweight headless browser for AI agents (17 MB idle) — Purpose-built browser for agents that doesn't consume gigabytes of RAM. GitHub · Jul 9

---

Try This Next

• 🔧 Read Martin Fowler's local models piece — the best practical guide to running coding models on your own machine
• 🎥 Test Ember if your agents need web browsing with low overhead
• 📝 Watch the Indeed Hiring Lab data — if agentic AI is a net job creator for developers, that changes the adoption narrative