POCs and demos are easy to build by anyone these days. The last 10% is what separates student projects from real products.
any engineer who has spent time in the trenches understands that fixing corner cases in code produced by inexperienced engineers consumes a lot of time.
in fact, poor overall design and lack of diligence tanks entire projects.
I got laid off thrice during my 25+ year career during which i have worked at 9 different companies. twice from startups in each of the previous busts and once from a large company. job hunts are not easy and i dont interview well.
however, there are tech transitions and it appears we are in the midst of one - after which a new set of tech careers seem to emerge. internet/web/mobile/cloud/saas and whatnot. i fully expect the same to happen here.
after the dot-com bust, i took a job that involved a 70 mile commute (each way) which was about 100 minutes on the road on average each way. At the time, i thought that was the only option for me. turns out it was one the worst career choices i made.
not sure i agree, i went to school in the 90's as well, and out of 9 jobs I have held since, 4 came through cold-applications and 5 through referrals. referrals imo have a better chance of success particularly if you are a senior engineer, and you could have a bigger eco-system of referrers. I would also like to add that with more experience, you need to build bigger referrer networks which help in this whole process.
i try to help anyone i know (with referrals) as i have personally seen the anxiety, stress, and emotional roller-coaster that accompany the job hunt.
nice project ! and attempt to modernize the spreadsheet tech stack. be warned though, that the feature-set of the modern spreadsheet is gigantic, with tentacles in hundreds of nooks and corners. Any attempt to play catch-up is a losing proposition. Hence you could position it so as not to directly compete with excel - in areas that are new or emerging.
I've said this in some other comment. We can't and will never be as feature full as Excel. We want to find different use cases. A super light spreadsheet with MIT license that you can use anyway you want.
If I were to go closed source, for instance, I think I loose 90% of my selling points.
One of my dreams if I were to create my own web-based spreadsheet engine would be to make worksheets much more general purpose and not stuck in tab UI like Excel. You would be able to embed them separately in the page so multiple worksheets could be viewed at the same time. Example: you might have multiple small worksheets with financial summaries where the columns are slightly different. You could also do things like include text above/below worksheets, to present the data in a document-like format.
well said. agree 100%. papers like these - and i did skim through it, are thinking "within the box" as follows: we have a system, and it has a problem, how do we fix the problem "within" the context of the system.
As you have put it well, there is no notion of truthfulness encoded in the system as it is built. hence there is no way to fix the problem.
An analogy here is around the development of human languages as a means of communication and as a means of encoding concepts. The only languages that humans have developed that encode truthfulness in a verifiable manner are mathematical in nature. what is needed may be along the lines of encoding concepts with a theorem prover built-in - so what comes out is always valid - but then that will sound like a robot lol, and only a limited subset of human experience can be encoded in this manner.
I will have to disagree with the author on specifics (but agree on the broad premise)
it is not what the foundational knowledge "is" - but "how" the foundational knowledge is "unpacked" by each person from the first principles.
this was articulated by Descartes in "Rules for the direction of the mind" which remains sadly unread and forgotten.
the difference between the 2 approaches is that the author's approach does not clearly delineate what is meant by "learning". In fact, it could even encourage rote-learning of the foundational material. Descartes outlines an alternative.
An interesting anecdote about Descartes is that he "unlearnt" everything he knew early in life, and attempted to "rebuild" his knowledge. Few of us have the luxury to do that.
thanks for this comment ! it clarifies the function of the llm well.
ie, use it as a template-generating search-engine helper for most common things.
for uncommon things, you have to prompt-guide it to get what you want.
I would have liked to see a giant ppt of an agentic framework or architecture.
Call it Enterprise Agentic Framework or something like that.
The architecture diagram would fill an entire ppt slide and bedazzle its customers.
every hype cycle runs through a predictable course.
we are at a phase where the early adopters have seen the writing on the wall.. ie that llms are useful for a limited set of usecases. but there are lots of late adopters who are still awestruck and not disillusioned yet.
POCs and demos are easy to build by anyone these days. The last 10% is what separates student projects from real products.
any engineer who has spent time in the trenches understands that fixing corner cases in code produced by inexperienced engineers consumes a lot of time.
in fact, poor overall design and lack of diligence tanks entire projects.