Opinionaited
01 / 05
EXIT
Diagram showing the four rungs of AI developer tooling abstraction
THE LADDER

From six layers of scaffolding to zero code

Google's ADK requires six abstraction layers. OpenAI's agent framework uses four. Anthropic's Claude Agent SDK stripped it to two. Claude Code eliminated code entirely. OpenClaw made it configurable via JSON and markdown. Each step assumes greater model competence.

Editor's Insight

Opus 4.6 regressed on SWE-bench but jumped on agentic benchmarks. It got worse at being a code-completion engine and better at being a colleague. Average turn duration doubled from 25 to 45 minutes.

Timeline of Anthropic's product releases from January to March 2026
THE BLITZ

Ten major features in ten weeks

Opus 4.6, Agent Teams, Auto-Memory, Code Review, App Store, Academy, Xcode MCP, OAuth expansion — shipped in rapid succession. Companies running Claude Code at scale processed 334 million tokens per day. One enterprise saw 80% PM adoption within six weeks.

Diagram showing platform ecosystem shift after the cease-and-desist
THE FUMBLE

A cease-and-desist that rewrote the platform map

OpenClaw traffic as a percentage of Claude Code volume went from 5.5% on January 26 to 28.9% by March 9. Claude Code's biggest npm download spike coincided with OpenClaw's ACP integration. Anthropic's legal action pushed the fastest-growing open platform layer out of their orbit.

Editor's Insight

OpenAI hired Steinberger on February 14. Codex had full OpenClaw compatibility within weeks. Jensen Huang credited OpenClaw at GTC. The narrative shifted from 'Claude extension' to 'model-agnostic standard.'

The domain gradient from formal to situated knowledge domains
THE GRADIENT

Why coding worked first — and what comes next

Coding has compilers and test suites for verification. Legal and finance have rules but require interpretation. Operations and field work rely on tacit, contextual knowledge. Only 3% of enterprises report high trust in AI agents. The unlock for categories 2 and 3 is a codeless platform that domain experts can configure.

Editor's Insight

89% of enterprises cite data quality problems rooted in uncaptured domain knowledge. The supply chain manager who knows Tuesday deliveries from Supplier X are always 15% short — that knowledge isn't in any database.

Strategic map showing value migration from model layer to platform layer
THE PARADOX

Anthropic proved its own thesis, then handed the conclusion away

OpenClaw now has 321,000 GitHub stars, 1,239 contributors, and 2.32 million npm downloads per week. The open-platform narrative belongs to OpenAI. The enterprise domain-expert layer — the one that matters for the next $100B of AI revenue — is developing outside Anthropic's ecosystem.