FlowLens
activeUnderstand a software repository before changing it. Flow Lens extracts architecture, responsibilities, dependencies, and boundaries into an explorable system view.
INDEPENDENT SOFTWARE + AI STUDIO
Useful systems, thoughtfully made.
Pegasus Loop is an independent software and AI studio that builds practical tools, experiments with emerging technology, and explores better ways for humans and AI to work together.
Products
Snapshot last updated
Pegasus Eval
developmentEvaluate AI systems with reproducible benchmarks instead of vibes. Pegasus Eval runs models against controlled question sets and deterministic grading rules, then makes accuracy, latency, failures, and model-to-model differences inspectable.
WHAT WE DO
BUILD
Practical software
We build software from focused experiments to public applications, with an emphasis on usefulness, clear architecture, and systems that can evolve over time.
EVALUATE
Evidence over hype
We test AI and software against observable behavior—not just impressive demos. Evaluation, provenance, failure analysis, and measurable quality are part of the build process.
EXPLORE
Learn by building
We experiment with AI, local computing, information systems, developer tools, and human-computer interaction. Useful discoveries can become products; unsuccessful experiments still teach us something.
VISION
AI should increase what people can build—not eliminate the judgment required to build well.
Pegasus Loop is built around a simple idea: better automation should expand human capability.
AI can write code, analyze information, operate tools, generate alternatives, and dramatically reduce the cost of experimentation. That changes what a small team—or even one person—can build.
But execution is not the same as judgment. People still define the problem, choose what matters, evaluate the result, understand consequences, and decide what should exist.
We build systems around that division of labor: automate what machines do well, preserve evidence where decisions matter, and keep human judgment where judgment is actually required.
SERVICES
Selected technical consulting and product work.
Pegasus Loop works with organizations on problems where AI, software, product thinking, and evaluation overlap. Engagements can range from focused technical analysis and prototypes to working systems.
AI + LLM Systems
Applied LLM systems, retrieval and knowledge workflows, conversational AI, agent behavior, prompt and workflow design, and local or hosted model integration.
Evaluation + Quality
Evaluation pipelines, ground-truth design, model quality measurement, failure analysis, provenance, automated validation, and human-in-the-loop review systems.
Product + Prototyping
Rapid technical prototypes, AI-enabled product concepts, workflow automation, architecture exploration, and turning ambiguous ideas into testable software.
WRITING / RESEARCH
Ideas from the work.
Notes, essays, experiments, and longer-form work about AI, software engineering, evaluation, human-computer interaction, and what we learn while building Pegasus Loop projects.
Pattern IQ
What We Observed
Seven models. Fifty deterministic pattern-reasoning questions. A closer look at scores, shared failures, disagreement, and why a failed grade does not necessarily mean failed reasoning.
Read research note →ABOUT
Small by design.
Pegasus Loop is an independent software company and experimental studio. We build our own products, publish selected experiments and research, and take on consulting work where our technical experience is useful.
We deliberately keep the organization lean. Modern AI and automation make it possible to experiment faster and operate software with far less organizational overhead—but the work is still directed, tested, and evaluated by humans.
CONTACT
Start a conversation.
For project inquiries, consulting, or general questions, get in touch by email.
- lydia@pegasusloop.ai