Guardrails against what? Rehashing public wikipedia information?
Execution matters, and they did a trurly horrible job that crippled their product to the point of being useless and a joke. Huge mistakes were made and im sure they regret it already, heads will roll.
Building resilient, robust, self-healing, best-effort systems. Important in computer architectures, new chip designs, medicine, material sciences... I can see this being useful in different domains.
I used dspy in production, then reverted the bloat as it literally gave me nothing of added value in practice but a lot of friction when i needed precise control over the context. Avoid!
They are hiring "architects", or do we call them analysts. The impression is we're going back to analysts drawing those pld school UML-like diagrams etc. Also, a lot of the devs are on the brink of just quitting, because it's "not programming" anymore. So, not only will you still need devs, or people massaging those specs, you'll also need enough "product" people to keep that engine fed! If your management isn't lazy, I can see the need for growing people count will continue to rise within such companies. That doesn't mean the work will be ...satisfying for devs.
I don't think it's implemented that harsh or enforced so hostile, but they have these rict procedures now on how the code is to be developed. That procedure they follow is all centered around automated code generation. So they simply... don't anymore in practice, it is not part of the job description so to speak. He wasn't happy I can tell, but also acknowledged it was working very well.
I also thought it was pushing it to the limit, but I think this is just some Founder of a successful company deciding engineering was going to transform to this way of working. A huge bet, but the implementation didn't feel amateuristic or ad hoc. Just not very pleasant for most devs to work that way. I'm sure some will look elsewhere. I know I would!
Like I said, devs don't like it. He said productivity went up 3-4x. "It works". There was no question of denying that as far as he was concerned. At the same time he was going to look for another job as it was just painful to work like that.
I see whole teams pushed by c- level going full in with spec driven + tdd development. The devs hate it because they are literally forbidden to touch a single line if code. but the results speak for themselves, it just works and the pressure has shifted to the product people to keep up. The whole tooling to enable this had to be worked out first. All Cursor and extreme use of a tool called Speckit, connected to Notion to pump documentation and Jira.
It says a lot about US academic culture that they think in terms of hiring. There is an important educational commitment requirememt to the role of professor, at least in Europe. Hiring is to the betterment of your own goals and almost orthogonal to the educational mission. A lot of unethicalities fond their root in this schizophrenic mission statement of doing professional competitive scientific research and at the same time education of graduates.
I suspect the number of startups will skyrocket the nexr few years. Fired engineers will start to compete against the establishment that fired them. Competition may get a lot more fierce for a while.
"only for private pharma/bio/tech firms to add a thin layer of additional research (or design) on top"
Citation needed.
Go to market cost billions and takes a decade. Doesn't sound like a thin layer. I'm not disputing fundamental research in academia is an essential fuel to keep innovation engines running. But the contributions of biotech is not "thin".
I hate typing strings of syntax. So boring. Never saw the appeal. I do like tinkering with ideas, concepts, structure... just not the mechanical interaction part. Im not tbe best typist...then again, its the same with playing factorio. I love the concept of building structures, but fighting the UI to communicate my ideas is such a drag...
Because typing in text and syntax is now becoming irrelevant and mostly taken care of by language models. Computational thinking and sematics on the other hand will remain essential in the craft and always have been.
LLMs translate specs into code, if you master conputational thinking like Antirez, you basically reduce LLMs to intelligent translators of the stated computational ideas and specifications into a(ny) formal language + the typing. In that scenario LLMs are a great tool and speedup the coding process. I like how the power is in semantics, whereas syntax becomes more and more a detail (and rightfully so)!
I live there! Crazy...