# Rubric Labs > Rubric Labs is an AI engineering agency. We build AI agents, products, and models for companies, and ship them to production with your team. ## Pages - [Home](https://rubriclabs.com): Rubric Labs is an AI engineering agency. We build AI agents, products, and models for companies, and ship them to production with your team. - [Blog](https://rubriclabs.com/blog): Field guides and essays from the engineers at Rubric Labs on building AI agents and products: how the software works and where it’s going. - [Brand](https://rubriclabs.com/brand.md): Rubric Labs brand assets: logo files, wordmarks, color palette, and typography, with guidelines for using them in posts, decks, and integrations. - [Careers](https://rubriclabs.com/careers): Open engineering roles at Rubric Labs, a small AI engineering agency with its own research lab. We hire rarely and look for founder-level engineers. - [Book a call](https://rubriclabs.com/contact): Tell Rubric Labs about the AI product you want built. We read every request and reply within a day with times for a 30-minute call. - [Lab](https://rubriclabs.com/lab): Experiments from the Rubric Labs research lab on new AI models, methods, and tools, with the method and data behind every result. - [The Grid newsletter](https://rubriclabs.com/newsletter): Subscribe to The Grid, the Rubric Labs newsletter: what we’re learning about building AI products, sent when we publish something new. - [Privacy](https://rubriclabs.com/privacy): Rubric Labs privacy policy: what data we collect on this site, how we use it, the analytics and third-party services involved, and how to reach us about it. - [Team](https://rubriclabs.com/team): Meet the engineers behind Rubric Labs, an AI engineering agency with its own research lab, and see what each of us works on. - [Work](https://rubriclabs.com/work): AI agents and products Rubric Labs built and shipped for Albertsons, E2B, Gumloop, Graphite, Cal.com, and Sligo AI, and what each one changed. ## Lab studies - [Can a small model learn UI design from visual feedback?](https://rubriclabs.com/lab/uirl): After visual reinforcement learning, Qwen3.5-9B won 67.7% of blind model judgments against its untrained base. It won 47.9% against the base model’s best-of-four approach. - [Does encrypted compaction use fewer tokens than text summaries?](https://rubriclabs.com/lab/luna-compaction): Encrypted compaction used 22% fewer GPT-6 Luna tokens and cost 10% less per run than text summaries. - [Can Jev mask personal data better than an LLM?](https://rubriclabs.com/lab/pii-jev): Jev scored 78.1% mean masking F2, versus 54.6% for GPT-6 Luna and 70.3% for GLiNER. That is 1.43× Luna’s score. Jev led on three of the four sources. - [Can Jev build structured objects without generating text?](https://rubriclabs.com/lab/structured): Jev produced structured objects 4× faster than GPT-6 Luna: 418 ms versus 1687 ms median time. It produced 100% schema-valid objects, with 95.8% exact matches. Luna reached 100% exact matches and cost less. ## Case studies - [Partnering with E2B](https://rubriclabs.com/work/e2b): We embedded with E2B's team and shipped settings, billing, sandbox event visibility, and webhook monitoring to production. - [Partnering with Sligo AI](https://rubriclabs.com/work/sligo): We worked across backend reliability, evaluation systems, multi-agent orchestration, and enterprise security guardrails. - [Partnering with Albertsons](https://rubriclabs.com/work/albertsons): We worked with Albertsons Companies on its AI shopping assistant, which launched in December 2025 on the websites of all its grocery brands. - [Partnering with Gumloop](https://rubriclabs.com/work/gumloop): After a week with Gumloop in San Francisco, we shipped creator publishing and SEO-optimized template pages in one month. - [Partnering with Graphite](https://rubriclabs.com/work/graphite): We turned GitHub stats into a 12-scene script, then rendered each developer's video with Remotion and AWS Lambda. ## Blog posts - [What can you build with Jev?](https://rubriclabs.com/blog/what-can-you-build-with-jev.md): Jev is a frontier model that answers questions with probabilities in milliseconds, helping you make decisions. See five live demos of what you can build. - [Static Analysis for the Agentic Era](https://rubriclabs.com/blog/static-analysis-for-the-agentic-era.md): Testing whether deterministic repository representations help coding agents understand, navigate, and modify unfamiliar codebases. - [Everything we know about good agent design](https://rubriclabs.com/blog/everything-we-know-about-good-agent-design.md): A practical guide to the design choices that help agents plan, use tools, recover from errors, and verify their work. - [What is an Agent Harness?](https://rubriclabs.com/blog/what-is-an-agent-harness.md): A harness is the software that translates a model into a system that can affect its environment. Agent = Model + Harness. - [Challenging the Chatbot](https://rubriclabs.com/blog/challenging-the-chatbot.md): Intelligent applications need new interaction patterns beyond the chatbot: intelligence on every keystroke, in the background, and inside the content. - [Autonomous Iteration](https://rubriclabs.com/blog/autonomous-iteration.md): Testing coding agents' ability to improve a feature automatically: build it, screenshot it, critique the result, and iterate until it meets the bar. - [Every Server Deserves a Coding Agent](https://rubriclabs.com/blog/openclaw-backend.md): The promise and peril of collocating code and coding agent: what happened when we gave an SMS-based OpenClaw agent live access to its own production server. - [Contract Engineering](https://rubriclabs.com/blog/contract-engineering.md): How we define done for AI agents using verifiable contracts. The contracts are the application; the code is an implementation detail. - [Unblocking Agents](https://rubriclabs.com/blog/unblocking-agents.md): Tooling and infrastructure for autonomous software engineering: the access, infrastructure, and identity walls that stop agents, and how we remove them. - [Primitives over Pipelines](https://rubriclabs.com/blog/primitives-over-pipelines.md): Primitive-Oriented Agent Design: replace rigid AI workflows with thin prompts and modular primitives so agents can assemble faster paths for user intent. - [How does Claude Code actually work?](https://rubriclabs.com/blog/how-does-claude-code-actually-work.md): A first-principles breakdown of how Claude Code works, including the agent loop, tool use, context management, and design tradeoffs behind modern coding agents. - [Introducing Rubric Labs](https://rubriclabs.com/blog/introducing-rubric-labs.md): After a year and a half of building in stealth mode, we're excited to officially introduce Rubric Labs to the world. - [Fine-tuning a GPT model for spam detection](https://rubriclabs.com/blog/fine-tuning-for-spam-detection.md): How we fine-tuned a small GPT model to flag spam in our inbound contact form, using dozens of historical spam tags stored in Postgres as training data. - [My Summer at Rubric](https://rubriclabs.com/blog/my-summer-at-rubric.md): A recap of a summer of projects at the Lab, from the Rubric UI component library to Genson, an experiment in generating full React UIs with LLMs. - [Multi-staging → Local → Prod in record time](https://rubriclabs.com/blog/multi-staging.md): How to use multi-staging workflows to build and test multiple full-stack changes in parallel, from local development to production, in record time. - [Planning for near-free AI](https://rubriclabs.com/blog/planning-for-free-ai.md): How trends in open-source large-language models are converging to usher in cheap, abundant intelligence. The state of LLM capability, cost, and timeline. - [Create Rubric App](https://rubriclabs.com/blog/create-rubric-app.md): Create Rubric App is an open-source, full-stack AI agent template in the spirit of Create React App. Learn how it works and how to remix it for your purposes. ## Team members - [Dexter Storey](https://rubriclabs.com/team/dexter): Dexter Storey is the co-founder of Rubric Labs working on infrastructure for AI products, including deployment, permissions, and agent verification. - [Sarim Malik](https://rubriclabs.com/team/sarim): Sarim Malik is the CEO and co-founder of Rubric Labs. He works across product design and engineering, with a focus on coding agents and interaction models. - [Ted Spare](https://rubriclabs.com/team/ted): Ted Spare is the co-founder of Rubric Labs working on agent systems and applied AI products, including model capability and autonomous iteration. ## Open roles - [Product Engineer](https://rubriclabs.com/careers/product-engineer.md): Rubric Labs is hiring a Product Engineer in Canada to design, build, and ship AI products for the fastest-growing companies. ## Interactive demos - [The filterable dashboard](https://rubriclabs.com/blog/iterative-improvement/demos/dashboard): An interactive demo of iterative improvement: the agent built this page, screenshotted it, critiqued the result, and improved it over iterations. - [The landing page](https://rubriclabs.com/blog/iterative-improvement/demos/landing): An interactive demo of iterative improvement: the agent built this page, screenshotted it, critiqued the result, and improved it over iterations. - [The onboarding flow](https://rubriclabs.com/blog/iterative-improvement/demos/onboarding): An interactive demo of iterative improvement: the agent built this page, screenshotted it, critiqued the result, and improved it over iterations.