Insights on AI, Cloud
& Modern Engineering
We write about AI agents, cloud architecture, cost optimization, and the tools we use every day to build software.
Amazon Quick Desktop App: The AI Assistant That Works Across All Your Apps
Amazon just launched a native desktop app for Quick-a proactive AI assistant that connects to 40+ apps (Google Workspace, Microsoft 365, Salesforce, Zoom), builds a personal knowledge graph, and starts free with no AWS account required.
Cursor SDK Developer Guide: Build Programmatic AI Agents with TypeScript
Cursor just released the @cursor/sdk - a TypeScript package that gives you programmatic access to the same agent runtime powering the Cursor IDE. We cover installation, execution modes, Composer 2 pricing, the full harness (MCP, skills, hooks, subagents), real-world use cases, and how it compares to Claude Code SDK and OpenAI Codex.
Claude Mythos vs GPT-5.5: Benchmarks, Pricing & Which Model Wins
Claude Mythos scores 93.9% on SWE-bench but is locked behind Project Glasswing. GPT-5.5 scores 88.7% and is live today. We compare every benchmark, break down pricing, and give you a practical decision framework for choosing the right model.
GPT-5.5 for Enterprise: Automating Knowledge Work, Computer Use & Multi-App Workflows
GPT-5.5 scores 84.9% on GDPval (44 knowledge-work occupations), 78.7% on OSWorld-Verified (autonomous desktop operation), and 98.0% on Tau2-bench Telecom. We break down what these benchmarks mean for enterprise automation, the super-app vision, and how to plan your migration from GPT-5.4.
GPT-5.5 Codex Integration: Building Autonomous Coding Agents with Spud
GPT-5-Codex pairs GPT-5.5's retrained base with specialized coding optimization. We cover the Codex SDK, Dynamic Reasoning Time (up to 7+ hours), SWE-Bench Pro (58.6%), Terminal-Bench 2.0 (82.7%), multi-agent teams, and cost optimization strategies for autonomous coding workflows.
GPT-5.5 Omnimodal API Guide: Building Apps with Native Text, Image, Audio & Video
GPT-5.5 is natively omnimodal - trained from scratch with text, images, audio, and video in a single system. We cover the API architecture, gpt-image-2 ($5-$30/1M tokens), gpt-realtime-1.5 ($32-$64/1M tokens), cross-modal workflows, and production integration patterns.
GPT-5.5 Safety & Security: Cybersecurity Classification, Red Teaming & Production Guardrails
OpenAI classifies GPT-5.5 as 'High' cybersecurity risk and delayed API access for safety. We cover the risk classification, red teaming from 200 partners, stricter classifiers, production guardrails, SOC 2/HIPAA compliance, and defense-in-depth architecture patterns.
GPT-5.5 vs Gemini 3.1 Pro vs Claude Mythos: Three-Way Frontier Model Comparison
Three frontier models, three different strengths. GPT-5.5 leads agentic workflows (84.9% GDPval, 78.7% OSWorld). Gemini 3.1 Pro leads reasoning (77.1% ARC-AGI-2, 94.3% GPQA). Claude Mythos leads coding (93.9% SWE-bench). We compare benchmarks, pricing, and build a multi-model routing strategy.
DeepSeek V4-Pro vs V4-Flash: Benchmarks, Pricing & Which Model to Choose
DeepSeek shipped two V4 variants on April 23, 2026: V4-Pro (1.6T params, 49B active) and V4-Flash (284B params, 13B active). We compare benchmarks, pricing, reasoning modes, and real-world use cases to help you pick the right one.
DeepSeek V4 vs Claude Opus 4.7 vs GPT-5.5: Frontier Model Showdown
Three frontier models launched in the same week of April 2026. We compare DeepSeek V4-Pro, Claude Opus 4.7, and GPT-5.5 across coding, reasoning, agentic tasks, pricing, and licensing to help you build a multi-model strategy.
Self-Hosting DeepSeek V4: vLLM Setup, Hardware Requirements & Deployment Guide
DeepSeek V4 ships under MIT license with open weights. We cover hardware requirements for V4-Pro (862GB) and V4-Flash (158GB), vLLM deployment, quantization options, expert parallelism, and cost analysis for self-hosted inference.
DeepSeek V4 for AI Agents: Function Calling, MCP Integration & Agentic Workflows
DeepSeek V4 ships with native function calling (128 parallel calls), pre-tuned adapters for Claude Code and OpenCode, and MCPAtlas scores rivaling Opus 4.6. We cover agentic architecture, tool use patterns, and production deployment.
Ship Better Engineering, Every Week
Practical writing on AI agents, cloud architecture, and product teardowns. Read by builders at startups and Fortune 500s.
- New deep-dives on AI agents and cloud architecture
- Engineering teardowns of shipped products
- No spam, unsubscribe in one click
We respect your inbox. Read our privacy policy.