Tuesday, February 17, 2026

Start of archive·May 16

Top story

Platform WatchPragmatic Engineer

How Codex is built

Understanding how leading AI coding agents are architected gives technical leaders insights into building internal AI development tools and evaluating vendor capabilities.

Exclusive deepdive into OpenAI's Codex team methodology

The briefs

As AI moves beyond conversational interfaces into autonomous agents, leaders need new frameworks for selecting the right AI tools for specific business functions rather than defaulting to general-purpose chatbots.

AI evolution from chatbots to agentic workflows

Anthropic's new Sonnet 4.6 is live as a product integration on Vercel's AI Gateway, bringing near-Opus capability and extended thinking to developers building agents. This is infrastructure-layer adoption, not a standalone model release—the story is availability and tooling, not benchmark claims.

Claude Sonnet 4.6 available on AI Gateway

Vercel is shipping AI-powered autonomous code repair into the developer workflow—turning build failures from manual debugging into one-click fixes. This is app-layer AI moving from chatbot to embedded agent.

Vercel Agent now scans deployments for build errors automatically

OpenAI and Paradigm are establishing a new benchmark for evaluating AI agents on high-stakes security tasks—detecting, patching, and exploiting smart contract vulnerabilities. This signals a major shift in how AI capabilities are measured beyond traditional NLP/reasoning tasks and into specialized domain expertise (blockchain security), suggesting agents are moving from theoretical to production-critical applications.

EVMbench evaluates AI agents on smart contract vulnerability detection, patching, and exploitation

Recraft V4's availability on AI Gateway lowers friction for teams integrating specialized image generation into design workflows. The dual-tier model (standard + Pro) targets both iteration speed and print-ready asset production, expanding the AI-native design toolchain.

Recraft V4 now available on Vercel AI Gateway

Vercel's platform is lowering the barrier to deploying autonomous AI agents at enterprise scale. This is a real-world case study of how infrastructure-as-a-service is accelerating AI product velocity for lean teams.

Stably: 6-person team

Amazon Scholar's decade-old machine learning work on predictive forecasting has proven durable enough to win institutional recognition, signaling that foundational ML research on real-world prediction problems maintains long-term value and relevance to enterprise AI applications.

2014 paper on forecasting civil unrest in Latin America

Ramp deployed Vercel's AI SDK for semantic search across internal assets while maintaining production stability at scale. Shows how infrastructure + AI tooling enables both reliability and capability in modern fintech platforms.

100% uptime maintained during 100x traffic surges

The Pragmatic Summit recordings contain actionable insights from engineering leaders that could inform AI implementation strategies and technical decision-making for founders and CTOs.

All summit sessions now available with Q&A