Forgetting Well: The Compaction Problem Behind Every Long-Running Agent
Baseten's research team and Harvey are circling the same question: what should an agent keep when it cannot possibly keep everything?
Writing.
No posts in this topic yet.
2026
Agent Memory Engineering: The Discipline Behind How Agents Actually Remember
Nico Bustamante just published the deepest practitioner guide yet on how three production agents actually implement memory—and the lessons translate beyond the CLI into how any team building enterprise agents should be thinking.
The Harness Is What Matters: Anatomy of the Modern AI Agent Stack
Harrison Chase, CEO of LangChain, breaks down the modern AI agent stack—why the harness matters more than the model, and the core primitives that make agents work.
OpenAI Frontier and the Rise of the Business Context Layer
OpenAI just launched Frontier, its enterprise agent platform. The most significant part isn't the agents—it's the Business Context layer underneath them, and what it signals about the future of enterprise AI.
Context Is the Product: Best Practices for Building AI Agents That Actually Work
Everyone can call an LLM API. What separates agents that work from agents that fail is context—clean, normalized, searchable knowledge that turns a generic model into a domain expert.
Before You Deploy: Why AI Evaluations Are the Difference Between the 95% That Fail and the 5% That Succeed
With 95% of AI projects failing and $67.4B lost to hallucinations in 2024, pre-deployment evaluations are the critical differentiator between success and expensive failure.
The Question That Separates AI Agents That Deliver From Those That Don't
Enterprise apps with AI agents will surge from 5% to 40% by 2026. The key differentiator isn't the technology—it's whether you can verify the agent's work. A framework from Anthropic's internal practices.
From 10 SDRs to 1: Inside Vercel's AI Agent Transformation
Vercel reduced their inbound SDR team from 10 to 1 in six weeks—with conversion rates matching their best human performers. Inside the development process, architecture, and GTM transformation playbook.
The Era of Context: Why Your Data Becomes Your Moat in the Age of AI Agents
Box CEO Aaron Levie argues that in a world where everyone has access to the same AI intelligence, context—your proprietary data, processes, and knowledge—becomes the ultimate competitive advantage.
2025
The Complete Guide to Hosting AI Agents in Production
AI agents need more than an API call—they need a persistent environment with tools, state, and security. A first-principles guide to the four hosting patterns that enterprise teams must understand.
The Enterprise AI Playbook: McKinsey's Former AI Chief on What It Takes to Succeed in 2026
Matt Fitzpatrick spent a decade leading AI at McKinsey. Now as CEO of Invisible Technologies, he reveals what separates companies that transform with AI from those stuck in experimentation—and why 2026 is the year that gap becomes insurmountable.
The $37 Billion Inflection Point: What 2025 Enterprise AI Reports Reveal About the Future of Business
Enterprise AI spending surged 22x in two years. New reports from OpenAI, Menlo Ventures, and a16z reveal who's winning, where the money is going, and what separates leaders from laggards.
The Box Blueprint: How to Move from AI Experimentation to Strategic Execution
Box ideated 100+ AI agents, then focused on 15-25 big bets that transformed how 2,800 employees work. Their governance model offers a blueprint for enterprise AI transformation.
From Pilot to Production: A Practical Framework for Enterprise AI Implementation
Most organizations know what AI high performers do. The challenge is execution. A proven 5-phase framework for moving from AI experimentation to enterprise-wide transformation.
The 6% Problem: What McKinsey's 2025 State of AI Report Reveals About Enterprise Transformation
Nearly nine out of ten organizations now use AI. But only 6% are capturing meaningful enterprise value. McKinsey's latest research reveals what separates the leaders from the rest.