The AI Runtime library brings together our weekly newsletter, the free field guides and playbooks, and Signal from the room: the questions practitioners are working through, answered. All free.
A global music label spent two months building an AI workflow around how accounts payable was supposed to work. Then the operators showed them how it actually worked.
Why enterprises need bounded, observable, and governed improvement loops for production AI agents, and how the open-source ASSERT framework engineers the half of the loop everything else depends on.
Part II - Governing Codebase-Wide Change: When agents can refactor an entire system, the pull request stops being the unit of control. Senior engineers govern the transformation as a program.
Part I: The Senior Engineer Stops Being the Fastest Coder
How Sphere separates probabilistic AI research from deterministic tax execution
Your Web Agent Needs a Trust Layer, Not a Bigger Model
Part 1: Getting the data reliably
These failures weren’t caused by bad prompts. They were caused by loops with authority but no control: a database gone in nine seconds, an agent that rewrote its own timeout, a deletion during a code
The visible half of loop engineering is orchestration. The decisive half is the control system that keeps the loop bounded.
Tackling the AI Information Silos problem
The Next Bottleneck in AI Systems Is Context Transfer
Agents reason. State machines remember.
Sixty production deployments converged on a three-layer architecture where the eval surface, not the base model, is the moat.
His new essay asks regulators to build an FAA for AI models. If you ship AI inside a bank, you already work in that regime, and there is a five-minute test that tells you whether your agent is as far
The harness engineering discourse names what to build. The Model Reliability Engineering arc names how long the build lasts, what kills it, and what to do at week six.
Most AI projects die in the gap between "works in the demo" and "works in production."
Quantization changes the numbers. Lossless compression removes the wasted bits and keeps every output identical, for about 30% less memory.
Quantization changes the numbers. Lossless compression removes the wasted bits and keeps every output identical, for about 30% less memory.
The leading AI legal research tools still hallucinate on up to a third of queries, so the production answer in law is not a better model but a harness built to assume the model is wrong.
Sixteen published deep-dives, four modules, one operating thesis. The harness around the model is the product. Free.
Working on something specific? Ask our team directly. We read every question.
One free issue a week on Vertical AI Agents, Model Reliability Engineering, and lessons from the trenches of shipping AI.
Production-grade writing, in your inbox.
Read by AI practitioners at IBM, Amazon, Meta, Google, Nvidia, OpenAI, MIT, Harvard, and more for production learnings.
Free. No spam. Unsubscribe in one click. See past issues →