Glad I’m not looking for a job.
https://www.nytimes.com/2026/06/11/technology/bezos-prometheus-ai-engineer.html
{{ Backed by more than $12 billion in funding, the company intends to create new engineering tools using many of the techniques used to build chatbots and other A.I. technologies. The hope is that these automated systems will improve the design and manufacture of practically any device, from computers to jet engines. }}
intercst
This is more complex than they think. Will they buy the computer codes of private companies and feed it into their AI learning? For example, very component of a nuclear power plant uses computer codes developed by reactor suppliers and their sub-suppliers and by national engineering societies like ASME, ASCE, IEEE, etc. These codes are the core of of their engineering and if they sell their codes they are out of business. Furthermore, NRC and most other regulatory agencies on the federal, state and local level need to have these companies certify the safety of their systems, structures and equipment.
Who will then be licensed to certify the safety of any systems, structures and equipment designed by AI? I would never feel safe in any automobile, aircraft, ship, elevator, generator, transformer, pressure vessel, control system, electrical system, chemical reactor, bridge, high rise building, etc. without proper regulation and certification.
It’s difficult to think of a better use for $12 billion.
Can the new engineer chatbots help fight wars?
Companies are already using AI as part of their engineering efforts. Identifying and organizing specs is extremely menial. Engineering requires multiple layers of review. There are plenty of opportunities to use AI as part of the process, with humans still driving the actual design.
Hyped up headlines. I was wondering when Bezos would compete better with Musk in AI and space development.
This will cost a lot less than $12 billion. It will probably be a desktop AI. That is where everything is going quickly.
Desktop AI won’t do a lot of “learning”. The AI will endlessly factor “odds”. Meaning it reviews odds, it discusses with engineers odds, it reports back with odds, it summarizes with odds, it suggests with odds. The words it uses to do all of this, the odds, the math, all odds, it is right or wrong. It is an extensive “odds” computing. Suggestions to carry out and double checks things are correct.
In coding, we are there. No matter how extensive the computer programming job.
The odds have proven AI very useful to professionals.
The productivity gains for engineers will be tremendous. This will be a very profitable application of AI.
Last night I asked Claude on a scale of 1 to 10 with 10 being the most difficult how hard is my game project. Claude said 9, but because I have gone into the aspects of each difficulty with programming the risks on the same scale were much lower that I get the project done. Almost no solo developers have ever gotten this far.
Computer programmers have used sub-routines for selected activities for generations. I see no reason engineering cannot be reduced to a collection of subroutines.
The creative part is where AI comes in. Running through a collection of sub routines to see which materials and designs meet requirements and are cost effective seems very possible to me. Doesn’t cad/cam software do this already?
Thousands of possible situations might come up but the most used subroutines get perfected first. Completing all possible ones might take decades but that is progress.
So does submitting documents to the Courts, but as we have seen those “layers of review” are often called out for not having reviewed”. There are dozens of instances where fantasy cases have been submitted, attorneys sanctioned, fines imposed, and it doesn’t seem to have eliminated the problem.
Do you really want your parking garage to be designed by an AI bot who conjures up supports that don’t exist, concrete that’s active only in its imagination, HVAC systems that are inadequate?
I do. Because different situations change the basic calculus of which solutions worked before. Nothing is perfect, I admit. Medical trials can’t possibly divine how any particular drug will use with the billions of combinations of other drugs people take, but I’d like to know if my skyscraper was designed by “a collection of subroutines” or by humans, pedigreed, vetted, and responsible for the oversight of the work.
That is still all on the engineer who applies his AI tools. Structural analysis can be run through more thoroughly with AI, but someone applies the program and signs off on the results. An experienced engineer will find AI useful for this.
That is only a drop in the million gallon bucket.
No, and that’s not what I meant. Here’s what is already happening:
Again, while in the early stages, all of this is already happening.
Agreed, they’re already finding it very useful. You bring up an excellent point. Some fear that AI will make experienced engineers obsolete. That’s not accurate, it will make inexperienced engineers obsolete. Those menial tasks that inexperienced engineers typically get saddled with are being handled by AI. The challenge AI presents to engineering - how do you keep the pipeline full of future experienced engineers, when the inexperienced engineers won’t be learning the way they used to.
I am an inexperienced programmer with some college education in coding and other programs.
Claude rates my game development on a scale of 1 to 10 with ten as the most difficult a 9. Further Claude qualified that my inroads to programming the game brought down the development risks considerably. Claude stated almost no solo developers would have done what I am doing. This only means I am early to do it because I am an experienced designer.
Claude’s idea of difficulty is complexity with a focus on the individual tasks.