Practical AI and machine learning engineering. LLM inference, tokenization, RAG pipelines, model deployment, vector databases, and the infrastructure behind modern AI applications.
54 articles
TDD doesn't work on non-deterministic LLM outputs. Eval-driven development is the analog: paired prompt-expectation sets, rule-based or LLM-judge scoring, run as a regression suite on every change. Tools (promptfoo, Braintrust, OpenAI Evals), real customer-support example, and what evals catch that code review misses.
A deep dive into running large language models on CPUs. Includes performance benchmarks, limitations, and optimization strategies.
A practitioner comparison of LangGraph, CrewAI, and AutoGen -- benchmarks on research, code gen, and data analysis agents with code examples, token efficiency, and production guidance.
A practical guide to the Model Context Protocol (MCP) -- Anthropic's open standard for connecting AI models to external tools and data. Architecture, TypeScript server examples, and security best practices.
VRAM requirements, tokens/sec benchmarks, and cost-per-token analysis for every GPU worth considering in 2026. Covers RTX 4060 through H100, multi-GPU setups, Apple Silicon, and budget tiers from $260 to $22,000.
The marketing says self-evolving; the engineering says adapter tuning + adaptive tool selection + memory consolidation on a static base. Honest breakdown of what M2.7 ships, when it wins, and when it doesn't.
Per-token pricing, caching credits, batch discounts, and hidden costs across OpenAI, Anthropic, Google, and open-source LLM providers. Includes four real workload simulations and cost optimization strategies.
Gemini 3.1 Pro tops the LM Council April 2026 board on GPQA Diamond and ARC-AGI-2 at 50% lower cost — but Opus 4.7 still leads on coding. The honest task-by-task decision guide.
Moonshot's Kimi K2.6 hits ~74% SWE-Bench Pro at $0.30 per typical run — 17-25x cheaper than Opus 4.7. Real benchmarks, where it falls short, and the two-tier routing pattern teams use in production.
New articles delivered to your inbox. No spam.