Critical Software is rebuilding itself around AI (200 ppl being layed off)

Co-founder Gonçalo Quadros says the Coimbra company will no longer be the software developer it has been, but will keep making systems for aviation, aerospace and cars that humans must still verify.

A cockpit instrument panel showing a large module being replaced by ten smaller, uniform components, illustrating the refounding of software development processes.

Critical Software, the Coimbra-based technology company, is being refounded because of AI. "We are refounding the company. There is a before and after artificial intelligence," co-founder and chairman Gonçalo Quadros told the Lusa news agency on Tuesday, according to Público.

Asked about a recent wave of layoffs, Quadros declined to give details but acknowledged that "some people" had been let go because of the changes AI has brought to the company. He did not say how many. The aim, he said, was to keep the organisation agile and fast by cutting whatever makes it less flexible. He added that the company is not currently in a job-cutting process.

Quadros described a sequence inside the company. Routine work went to AI first, notably testing a system and checking how it behaves, which the machine does with what he called "absolutely brutal" effectiveness and efficiency. AI then moved into producing code itself. Two or three years ago that was not the case, he said, but today the machine writes code of "brutal" quality.

Critical Software still develops software today. Quadros believes that, in a not very distant future, most software "will be made by the machine", and says what the company does can no longer be what it has been until now. The company told Público by email that the future of its software development will depend on adapting to the needs of its sectors.

The work that stays is technology for critical systems in aviation, aerospace and automotive. A system that controls an aircraft or ensures a train does not pass a red signal is a piece of technology that includes software but must be designed, verified and validated by humans, Quadros said, and he does not expect that to disappear.

He floated a ten-fold gain: AI might multiply the company's productive capacity by ten, so it could build technology ten times faster. That could make viable systems that were not economically efficient before, because they needed large amounts of software development that a machine can now take on.

Quadros also voiced concern about the race among the big AI developers. If machines reached recursive self-improvement, meaning a system able to optimise itself, and control were lost, he said, "we may be on the edge of the abyss". The probability of something dystopian is higher than it has ever been, he said, without saying whether it is large or small.