OpenAI Distillation Campaign: How CoT Extraction Works and How to Defend
OpenAI disrupted a coordinated chain-of-thought distillation campaign. How distillation attacks work, how to detect them, and API defenses.
Cohere Embed 5 Pro Explained: Architecture, Benchmarks, Pricing for RAG
Cohere Embed 5 Pro explained: what is confirmed on architecture, retrieval quality vs speed, pricing, and how to migrate a RAG pipeline from older embeddings.
KServe vs Ray Serve vs BentoML: Model Serving on Kubernetes ADR
Compare KServe, Ray Serve and BentoML for model serving on Kubernetes: autoscaling, GPU, LLM support, canaries, operations. Decision record for 2026.
AI Agent Evaluation Harness: Trajectory Evals in Production
How to evaluate tool-using AI agents: trajectory vs outcome evals, simulated environments, LLM judges, regression gates in CI, with a reference architecture.
Hybrid Search and Reranking for Production RAG: An ADR
Hybrid BM25 plus dense retrieval with reranking: architecture, fusion methods, latency and cost, and when to skip reranking. A 2026 decision record.
OpenAI Low-Cost Model Launch and the Shelved Astra Upgrade, Explained
OpenAI launched a low-cost model on Sept 30, 2026, one day after shelving an Astra upgrade. Pricing, positioning, what it signals about frontier economics.
Google Gemini 4 Argon Explained: Benchmarks, Access, What We Know
Google announced Gemini 4 Argon on Sept 30, 2026 after cancelling Gemini 3.5 Pro. What is confirmed, sourced benchmark claims, access limits and how it compares.
Decision Models with Typed Outputs: Jev 1.13 and Solar Decide Explained
A new model category returns typed decisions with probabilities instead of prose. How Jev 1.13 and Solar Decide work, the pricing model and where they replace LLM classifiers.
ElevenLabs Eleven v4: Expressive TTS Architecture and Voice-Agent Latency
ElevenLabs Eleven v4 and Turbo (Sept 28, 2026): 90+ languages, instant voice clones. How modern expressive TTS works and how to budget voice-agent latency.
Qwen3.8-Omni-Flash Explained: Native Omnimodal, 1M Context, Pricing
Alibaba Qwen3.8-Omni-Flash (Sept 18, 2026) takes text, image, audio and video natively with 1M context at $0.15/$0.47 per M tokens. Architecture and cost analysis.