Measure real outcomes
Define success up front and verify the product on real tasks. Actual use and outcomes inform our decisions.
Who we are
We are an AI product company based in Shenzhen. We design agent workflows around business goals and evaluate them by the outcomes they deliver. We also explore how AI can change the way people divide work, collaborate and build healthier working relationships.
01 / ABOUT USAbout
Start with the outcome
Pioneering builds products around what a business wants to achieve. We design a path that can be executed and verified, then turn it into a product.
In the SaaS era, companies embedded their management practices in tools and relied on people to learn the software and move work forward. Today agents can take on some of that execution. We can work backward from a business outcome, decide what AI should do and where people need to step in, then test the design against real results.
Tools such as Claude Code and Codex have lowered the cost of reproducing common SaaS interfaces and workflows. What matters to us is understanding real tasks and making our products faster and more reliable over time.
02 / PRODUCTSProducts
Open tools, actively maintained
We open-source the tools we build. You can inspect, run and adapt the source, or use the versions we maintain.
We aim to make everyday access free where possible, with usage-based charges for higher resource consumption. Users who need AI can also access reliable models at competitive costs through us.
Open source still takes long-term work. We respond to practical needs with short release cycles, fix issues and improve reliability.
03 / CAPABILITIESCapabilities
Turn goals into executable workflows
We work backward from outcomes to define tasks, dependencies, permissions and acceptance criteria. Agents execute within clear boundaries and keep evidence and execution records available for human review.
When facts cannot be verified, output may be unreliable, or a decision involves a business trade-off, the workflow pauses. A person checks the facts, handles exceptions and makes the decision before the agent continues.
We evaluate open and closed models on real tasks, comparing quality, reliability, speed and cost. Users can choose a model based on their workload, openness and data-control preferences, and budget.
04 / VALUESValues
How we build
Keep judgment with people
Make model limits and uncertainty visible so people can inspect the evidence, correct mistakes and decide what happens next.
Stay open and accountable
Publish our tool source and give users a choice of models and deployment options. Follow through with maintenance, evaluation and iteration.
SPACE FOR A BETTER WAY OF WORKING.Illustrative image
Technology moves forward.
So should working life.
05 / PEOPLE, FIRST.Work with us
We want people to do work they are proud of, speak openly when they disagree and have room for life outside work.
35+Experience deserves attentionPriority for ages 35 and over, with competitive tech-industry pay.
We give priority to candidates aged 35 and over. Experience with complex projects, hard trade-offs and leading teams takes time to build. We want to understand how that experience can help the team and offer pay benchmarked against the technology industry.
0PUAWork together with respectReview management practices and reject manipulation.
We review how managers and teams work together. Goals should be clear, feedback should be grounded in facts and disagreements can be discussed openly. Managers are responsible for resolving conflicts and fixing unreasonable arrangements. Humiliation, manipulation and intimidation have no place at work.
4on, 3 offMake time for life outside workOpen to a four-day working week.
We are open to a four-day working week. Each location can adopt it based on customer and team needs, with a decision by its business lead and notification to the company. We agree on delivery, coverage and handovers first so people can rest and teams know who to contact.