Aaron Levie, CEO of Box, identifies the context problem as the #1 challenge for AI agents in the enterprise. He draws a distinction between agentic coding — where much of the needed context already exists within the code base and users are technical enough to supply the rest — and the broader world of knowledge work agents, where the context problem becomes significantly harder.
Levie's observation highlights a critical barrier to enterprise AI adoption: while AI coding assistants benefit from structured, accessible context, general-purpose knowledge work agents must navigate fragmented, unstructured, and often siloed enterprise information. This framing underscores why companies building enterprise AI infrastructure need to solve the context delivery problem before agents can be truly effective across business functions.
Alex Lieberman introduces the concept of the Software Factory, an engineering paradigm where building software shifts from individual craft to an industrialized production line. Rather than humans writing every line of code, agents handle writing, reviewing, testing, and deploying software while humans design the system and set guardrails. He cites examples like StrongDM (where code must not be written or reviewed by humans), Ryan Carson's "Code Factory" (shipping over a thousand pull requests via agents), and major companies like Anthropic (90%+ AI-written code) and Google (75% AI-generated code) as proof points.
The article presents a Software Factory Ladder with five levels (0–4), ranging from fully manual "Artisan" work to an "Autonomous Factory" where agents catch bugs, fix them, and ship changes before any human is aware. Lieberman provides diagnostic questions to help leaders assess where their engineering org sits on this ladder and practical guidance for climbing each rung — from simply giving engineers AI tools (Level 0→1) to eliminating the human review bottleneck (Level 2→3) by defining "low-risk" criteria and letting agent-reviewed, test-passing changes auto-merge.
Beyond engineering, Lieberman argues that verifiability is the key criterion determining which business function becomes the next "factory." Functions with automatically checkable outputs, digital workflows, high volume, and reversible mistakes — such as QA, data pipelines, DevOps, and parts of finance — are most ready for industrialization. Strategy, design taste, and relationship-building remain least factory-ready due to their subjective nature. He emphasizes that the real metric isn't "percentage of code written by AI" but how much of the production line runs without a human in it.
Santiago argues that Claude Code and Codex do not replace Copilot and Cursor, pushing back against a narrative he's heard from the vibe-coding community. He draws a distinction between these tools, noting that Claude Code and Codex are agentic coding tools suited for different use cases than the inline coding assistants like Copilot and Cursor. The post appears to be the beginning of a longer thread elaborating on why these tool categories serve fundamentally different purposes in a developer's workflow.