OurColorings
Launching soonColoring images and books made from a person's own photographs and their own life stories, with a printed book at the end of it. The first product to come out of the laboratory.
The laboratory Artificial Intelligence without limits
AImitless means Artificial Intelligence without limits — the laboratory where the company researches, implements and puts to work the best of what artificial intelligence can do, according to the case.
That last phrase is the whole discipline. A capability is chosen because it fits the problem, not because it is new. Most software companies now use artificial intelligence. Fewer set out to understand it, and the difference shows up later, when something does not work and someone has to know why.
01
What the laboratory is
AImitless is the first of the Ximitless laboratories, and the form is meant to be repeated: one laboratory per frontier domain the company enters. This one takes artificial intelligence. It reads the field, tries what is new on real problems, and builds the products that come out of what it learns.
AImitless is internal. It is not sold, not licensed and not offered as a service. What it produces is the understanding the products are built from.
02
Why a laboratory, and not simply using AI
Almost every software company now says it uses artificial intelligence, and that sentence has stopped carrying information. Four differences make the case for owning the understanding instead.
- Using a technology means borrowing an understanding
- You adopt what is popular, you inherit its failure modes, and you find out what it cannot do when a user finds out first.
- Researching it means owning the understanding
- You know what a model is good at, where it breaks, what it costs and what it would take to replace it — before it ships inside a product someone paid for.
- A frontier moves faster than a roadmap
- Something that was the correct choice at the start of a build may be the wrong one by launch. Only a standing research capability notices this in time to matter.
- Choosing well requires having tried the alternatives
- According to the case is not a judgement you can make from a launch announcement. It is one you make from having put several options on the same real problem.
03
What happens inside
Four things, most of the time at once.
- Reading what is new
- Papers, releases, model cards, benchmarks, and the work that has not been written up yet. Knowing what already exists is most of the job.
- Trying it properly
- A model or a method is not adopted because it is new. It is put on a real problem from a real product, and it stays only if it is better on that problem.
- Learning it well enough to teach it
- The people here learn the technology, not the tutorial. Something is understood when someone can explain why it fails.
- Building
- This is also where the software gets written. The laboratory is not separate from the work — it is where the work is messy, and that is the point.
04
What comes out of it
What the laboratory has fed so far. What was learned making each one is still in the room.
Compliance automation, unnamed
In internal developmentThe recurring tax and legal obligations of a Chilean SpA, automated. Ximitless is its first user.
One laboratory per frontier. The first is AImitless. The list is not meant to end there.
The laboratory is upstream of the products, not beside them. That was a decision: research that never reaches a product is research someone else eventually commercialises.