Coding

Tencent's Hunyuan team dropped a 770B-parameter Mixture-of-Experts model under Apache 2.0 with 1M-token context. At $0.834/M input tokens, it undercuts every proprietary frontier model while matching or exceeding their academic accuracy.

Nvidia paused its $36B AI Compute Partnership program after employees warned it could invite antitrust scrutiny. The program guaranteed GPU rentals to cloud providers in exchange for 50% of revenue above a base rate — giving Nvidia unprecedented control over its customers' pricing.

vLLM 0.28.0 ships 584 commits from 270 contributors with Decode Context Parallel, fused MLA kernels for Kimi-K3, DFlash2 speculative decoding, and tiered KV cache offloading. The context-length tax — where long contexts killed throughput — is over.

OpenAI flagged its Astra model as approaching the 'critical' cybersecurity capability threshold — the point where it could independently discover and exploit zero-day vulnerabilities. Here is what this means for every enterprise agent builder.

Stanford HAI's 2026 AI Index reveals transformative shifts: 88% organizational adoption, 77.3% real-world agent success, and $252B global investment. Here are the 12 metrics every AI architect must track.

SWE-bench Verified hit 96% in 2026, up from 60% in 2025. When benchmarks saturate, they stop measuring progress. Here is what the AI evaluation landscape looks like post-saturation.

Prompt caching transforms multi-turn agent economics. We benchmark how GPT, Claude, and Gemini cached prefix pricing reduces agent session costs by 68% while maintaining identical output quality.

The Agent-to-Agent protocol enables AI agents to discover and communicate across frameworks. We benchmark Google ADK A2A against LangGraph cross-agent messaging on latency, throughput, and reliability.

NVIDIA Blackwell Ultra GB300 delivers 10x agent inference throughput over H200. We benchmark real-world multi-agent workloads measuring tokens per second, latency, and cost per million tokens.

NVIDIA reported $96.2B in Q2 2026 revenue, more than doubling year-over-year. Profit doubled to $59.7B. Data center revenue hit $89B. Jensen Huang forecasts 70% sales growth next year. This analysis covers what the numbers mean for AI infrastructure, agent builders, and the compute market.

AutoGen entered maintenance mode in October 2025. Microsoft Agent Framework 1.0 reached GA in April 2026. This migration guide covers every breaking change, provides side-by-side code comparisons, and includes a production deployment checklist for teams moving from AutoGen to MAF 1.0.

OWASP's 2026 report puts prompt injection at the center of agentic AI risk, citing CVEs, supply chain breaches, and tighter regulations. The lethal trifecta—direct injection, indirect injection, and jailbreaking—represents the three attack vectors that every production AI system must defend against.