What we do
Oasis-X is an engineering and research company. We work on systems that run without anyone watching them, which right now means software agents, industrial plants, and energy and climate infrastructure. What those have in common is that they are far easier to deploy than to account for, so the parts we build are the accountable parts: the controls that let a person step in, the records that show what actually happened, and the measurements that survive somebody disputing them.
The people who hire us are usually the ones holding the liability. The work also has to answer to the people living downstream of these systems, who rarely had a say in whether they were deployed.
We call the standard we hold that work to symbiotic intelligence. It is less grand than it sounds. The machine does what it is genuinely better at, the person does the judgment, and authority over the decision stays with the person instead of drifting quietly into the software.
Who benefits
-
People subject to automated decisions
If an automated system acts on your case, you should be able to learn that it did, see the reasoning, and get it undone. That needs an approval step and a record that outlives the session, which is the part we build.
-
Small operators, growers, and independent engineering teams
We publish our control, coordination, and security software as open source, so a two-person operation can run the same tooling a large firm runs, read how it works, and not have to take our word for any of it.
-
Communities exposed to industrial and climate harm
Disputes about environmental harm usually turn on who holds the measurements, and that is seldom the affected party. We produce independent measurement and documentation that a property owner or a community can use in an insurance claim or in court.
-
Industrial operators and their neighbors
Redesigning a process to use less energy, water, and material saves the operator money and cuts what the surrounding area has to absorb. Those two interests overlap more often than either side expects, and most of our process work lives in that overlap.
Oasis-X is the trade name of Open Agriculture Systems & Integrated Supply Inc., a Delaware public benefit corporation. The four groups above are the beneficiaries named in our charter, which makes them an obligation we report against rather than a values statement we can quietly revise.
How the work divides
Three kinds of work go on here and they are funded, judged, and released differently, so it is worth keeping them apart.
01
Ventures
Products with customers, prices, and a support obligation. Seven of them, ranging from live and in production to early development, and each one says which it is rather than leaving you to guess.
- Oasis-Cloud
- Oasis-Firmware
- The Weather Data Service
- Theta Meteorology
- Oasis-Scientific
- Oasis-Kinematics
- NuSci
02
Research
Four programs, each built around a question we cannot answer yet. There is no customer and no delivery date. What a program owes you is a clear argument and whatever evidence we have managed to produce for it so far, including the parts that have not worked.
- The Shape of Good Behavior
- The Structure of Clear Thinking
- Conditions of Strong Takeoff
- Econology Investigations
03
Open source
The coordination, security, and evaluation tooling we run our own work on. We publish it rather than sell it, partly because it is more useful that way and partly because tooling that claims to make AI auditable should be auditable itself.
- dot_swarm — agent coordination
- oasis-claw — agent security
- alignment_research — evaluation
Principles
Global AI Challenges
AI systems have to do more than perform tasks. They have to protect privacy, respect agency, and stay governable across their lifecycle. Researchers at the University of Central Florida named six such challenges. Our answer to each one follows.
| Challenge | Oasis-X response |
|---|---|
| Human Well-Being | Workflows that preserve context across sessions and surface conflicts in human preferences instead of averaging them away. |
| Responsible AI | Multi-objective value modeling, geometric safety barriers, and supply-chain audit on every installed skill. |
| Privacy | Plaintext credentials never enter the LLM context. Local-first architecture. Signed logs for prompt-injection incidents. |
| Human-Centered Design | A public risk ladder. Approval gates in the gateway. Explicit lifecycle controls: an agent can sleep, compact, and hand off. |
| Governance & Oversight | Git-native traces, signed logs, and lifecycle artifacts. We run our own development on the protocol we ship. |
| Human-AI Interaction | Inspectable, interruptible, and accountable. Operator approval runs through an explicit gate, and transcripts are append-only. |
AI Risk Ladder
- 01 Assistive AI drafts, summarizes, suggests, or explains. Human decides.
- 02 Delegated AI performs bounded tasks with reviewable outputs and logs.
- 03 Operational AI can affect systems, customers, money, or data. Requires approval gates, monitoring, and rollback.
- 04 High-Impact AI affects health, legal, finance, employment, or essential services. Requires formal governance.
- 05 Prohibited Deceptive, coercive, privacy-invasive, or designed to remove meaningful human control.
Work with us
Tell us the system you have to keep under control: an AI deployment, a plant, a site, or a claim. We will tell you what we can measure and what we can build.