Useful work is the starting point.

Application choice decides the outcome

In robotics, the choice of first application matters more than almost any technical decision made afterwards. The same stack succeeds on a bounded, repetitive, physically consistent task and fails on an open-ended one, and no amount of engineering rescues a badly chosen problem.

So the question is not "what could a robot eventually do". It is "which task is tractable now, valuable enough to matter, and forgiving enough to fail safely while it is being learned".

What makes an application tractable

• A bounded task. Clear start and finish, limited exceptions, unambiguous success.
• Physical consistency. Objects, surfaces and spaces similar enough day to day that the problem does not change underneath you.
• Structured environment. Known layout, controlled access, predictable lighting. Unstructured space multiplies difficulty.
• A measurable baseline. An existing figure for time, throughput, error rate or cost to compare against.
• Safe failure. A failure produces a delay and a person stepping in, not an injury or a damaged asset.
• Volume. Enough repetition that reliability can be measured rather than anecdotally assessed.
• No unsupervised judgement about people. Anything requiring a machine to make a call about a person's wellbeing is out of scope.

Measured against those criteria, the three areas below sit in a deliberate order of difficulty, not a product line.

Material handling, the realistic starting point

Moving known objects between known places in a structured space is the closest thing robotics has to a tractable task at scale. The environment can be controlled, the objects are usually consistent, volume is high, baselines already exist because the work is already measured, and failure typically means something sits where it should not for a few minutes.

It is also commercially honest: these are jobs that are already scheduled, already costed and already a source of difficulty, which means improvement is legible to whoever signs for it.

The built world, harder and valuable

Construction sites, facilities and infrastructure are not structured in the way a warehouse is. The useful near-term work there is therefore mostly observational rather than manipulative: capturing what is present, comparing it against what should be present, and turning that into a record someone can act on.

That is valuable because the record is usually the bottleneck. It is also tractable, because being wrong produces a data discrepancy to investigate, rather than a physical consequence.

Everyday assistance, the hardest category

Domestic and personal assistance is where public expectation is highest and where the engineering is least ready. Homes are unstructured, objects are endlessly varied, people move unpredictably, and the tolerance for failure is very low precisely because the setting is intimate.

It is also the category where overclaiming does the most damage. Unsupervised care of children, medical care and safety-critical household assistance are outside this venture's scope, and the useful question is which narrow, supervised, genuinely helpful task could be done well — not which impressive scenario could be filmed.

What this page is not

No application has been built, deployed or tested. These are not products, case studies or available services, and no customer or site exists. Software cannot give a machine a physical capability its hardware lacks, so nothing here should be read as a claim that a particular robot can do a particular job.

The purpose of naming three areas is to be specific about where the venture believes useful work starts, and to invite disagreement from people who operate in them.