The Human Is Not the Runtime

Automating PR-review handoffs through repository-state reconciliation while retaining human acceptance.

Nine Agent Frameworks and What I Borrow from Them

A source-level review of nine agent frameworks: what they mechanically do, the conventions they share, and the case for adopting their ideas in a custom setup.

Meta Ray-Ban Display: $799, One Eye, and a Wristband That Reads Your Nerves

Meta’s May 2026 display SDK preview expands development options. Consumer app availability, optical fit and useful recurring workflows still determine the value of these glasses.

Your Moat Was Never the Code

AI can reduce the effort of reproducing some software features. A reassessment of competitive advantage through distribution, trust, operating knowledge and production requirements.

Money Maker

An engineering view of persistent AI workflows: acceptance criteria, independent verification, durable execution and ownership. Lower production costs do not guarantee personal wealth.

Opus 5 Killed the Fable Tax

Claude Opus 5 launched at $5/$25 per million input/output tokens. Its vendor-reported benchmarks make it worth evaluating; migration, effort and optional fallback need workload checks.

The Graph Is the Context

Prompt, context, loop — and now graph engineering. Half the label is marketing; the shift under it is real: stop searching a flat pile of documents by similarity, start walking a typed graph of what's connected. ELI5 to the arithmetic that kills it in production.

The Review Is Automated — the Decision Isn't

Two independent model passes consolidated into one PR comment, review queues that run on a schedule, and human judgment moved earlier — to the spec. How I actually review code now.

The Test Became the Target

OpenAI and Hugging Face documented an evaluation-related intrusion in July 2026. The August findings identify an internal research model as the main driver and explain the containment failures.

The Fable Tax Never Left

A public Claude Code issue reported an unexpected Fable usage-credit prompt. Distinguishing advertised access, reported entitlement symptoms and confirmed billing mechanics.

Kimi K3: Pricing, Weights and Deployment

Kimi K3 brings a 2.8-trillion-parameter model, native vision and a one-million-token context window. Its public rates, license and deployment choices deserve separate evaluation.

The Fable Tax Moved Inside the Plan

Fable 5 returned to eligible subscription plans with a model-specific allowance. API prices explain cash costs, but do not establish the plan’s exact consumption formula.