Writing
What we’ve learned.
Field guides to how AI software works, and our calls on where it’s going.
Latest
Designing an Always-On Agent
How to build an agent like Muse, GrokBot and Instinct
November 5, 2026 · 8 min read
Morning! Leave by 8:15.
You’re at gate B12.
Welcome to Austin!
Your Uber’s at door 5.
Agent2:15 PM
Field guides
Model
What is an Agent Harness?
7 min read
Everything we know about good agent design
12 min read
✻ Welcome to Claude Code!
> fix the login bug
Read(src/login.ts)
Update(src/login.ts)
Bash(npm test)
⎿ 1 failed: empty password
Read(src/login.test.ts)
Update(src/login.ts)
Bash(npm test)
⎿ 12 passed
Fixed. All 12 tests pass.
>
? for shortcuts44% context used
Request
Tool call
Tool result
Answer
Context window
How does Claude Code actually work?
11 min read
Everything else
- Re-introducing Rubric LabsThree years in, and what comes nextAnnouncement · October 27, 2026AnnouncementOctober 27, 2026
- What can you build with Jev?Practical use cases for fast, cheap decision modelsField guide · October 7, 2026Field guideOctober 7, 2026
- Static Analysis for the Agentic EraEvaluating how repository maps affect coding agent performanceExperiment · September 22, 2026ExperimentSeptember 22, 2026
- Challenging the ChatbotIntelligent applications need new interaction patternsEssay · July 7, 2026EssayJuly 7, 2026
- Autonomous IterationTesting coding agents' ability to improve a feature automaticallyExperiment · June 10, 2026ExperimentJune 10, 2026
- Every Server Deserves a Coding AgentThe promise and peril of collocating code and coding agentExperiment · May 1, 2026ExperimentMay 1, 2026
- Contract EngineeringHow we define done for AI agents using verifiable contractsEssay · April 9, 2026EssayApril 9, 2026
- Unblocking AgentsTooling and infrastructure for autonomous software engineeringEssay · March 31, 2026EssayMarch 31, 2026
- Primitives over PipelinesGive agents modular functions instead of prescriptive workflowsEssay · March 4, 2026EssayMarch 4, 2026
- Introducing Rubric LabsThe bridge between AI research and productsAnnouncement · March 31, 2025AnnouncementMarch 31, 2025
- Fine-tuning a GPT model for spam detectionHuman-level spam tagging, running 24/7, for a couple hours of devExperiment · November 29, 2024ExperimentNovember 29, 2024
- My Summer at RubricBridging LLMs with component libraries and design systemsEssay · September 8, 2024EssaySeptember 8, 2024
- Multi-staging → Local → Prod in record timeGive every developer their own local and staging environmentExperiment · February 16, 2024ExperimentFebruary 16, 2024
- Planning for near-free AIWhat would you build if GPT-4 were 100x cheaper?Essay · January 7, 2024EssayJanuary 7, 2024
- Create Rubric AppThe fastest way to bootstrap a full-stack AI projectAnnouncement · November 1, 2023AnnouncementNovember 1, 2023