You Should Know About System One Models

Much of production AI is asking an LLM a yes-or-no question and begging for JSON. TypeSafe's Jev skips the text entirely: hand it state and questions, and get typed answers with probabilities back.

Building AI Analysts for Wall Street: A Q&A with Compound Co-Founder Yijia Liang

Compound is building long-running autonomous analysts for private equity. Its co-founder explains why general-purpose AI falls short in finance and what it takes to earn the trust of high-stakes users.

A/B Testing AI Products Is Weird

Shipping an AI photo-feedback feature at Bumble broke a simple rule: if the test wins, ship it. With generative features, a winning result still leaves questions about output quality, true cost, model choice, and where regulation lets you launch.

Q&A with Al Chen, Software Engineering Leader at Cisco

Al Chen leads customer engineering for Splunk Agent Observability. He shares war stories from deploying AI agents inside major banks: clearing security, compliance, and legal gates, preventing data leaks, and designing agents that fail gracefully.

Why Smaller AI Models Can Win in the Enterprise

Three engineering lessons from pre-training a 1-billion-parameter foundation model on proprietary transaction data for fraud detection, which improved detection accuracy by 24% over previous approaches.

Your Agent Paid. That Does Not Mean It Should Have.

Protocols like x402 let agents pay for APIs over plain HTTP with no human in the loop. Testing the official client against six attacks shows why a successful payment is the start of the security question, and where deterministic checks belong.

ROI That Survives the Second Question

When intelligence becomes a commodity, judgment is the scarce resource. A real AI return accounts for the full cost of building, checking, and correcting the work, and shows where expert judgment was better spent.

Does Your Agent Have ID?

Denmark just saw its first payment made entirely by an AI agent, and six banks including Bank of America and Capital One have published principles for shopping agents. Most of their worries come down to identity: who the agent is, who sent it, and what it's allowed to spend.

Banks Are Renting Their Agents. These Two Built One.

83% of banks using agentic AI get it embedded in third-party software. Francesco Galletta and Felippe Felisola Caso went the other way, and this week they show how to build a financial analyst harness a team can actually trust.

Your Guide to AI Engineer New York 2026

Built for Wall Street, this year's conference brings 1,000+ engineers and 100+ speakers from OpenAI, Anthropic, and Google DeepMind to Two Sigma, Citadel Securities, and JPMorganChase. Here's how the week runs, plus three tips for making the most of it.

A World Worth Understanding

Why The Daily Context exists: the people who make software are endlessly interesting, and their stories deserve telling, especially the ones whose work too often happens out of view.

What a Passing Eval Still Doesn't Tell You

An eval proves something specific under tested conditions. Before handing an agent production authority, ask what its passing eval actually authorizes, and treat the untested behavior as part of the release decision.

Where to Spend Your Week at AI Engineer New York

137 sessions, up to five at once. A day-by-day guide to the workshops, keynotes, and evening events worth building your week around.