cplat

I agree with this. LLMs can actuate over capabilities we expose, so the blame isn't only on LLMs for this.

That's not quite how auto-regressive models are trained (the expression of "ideas" bit). There is no notion of "ideas." Words are not defined like we humans do, they're only related.

And on the latent space bit, it's also true for classical models, and the basic idea behind any pattern recognition or dimensionality reduction. That doesn't mean it's necessarily "getting the right idea."

Again, I don't want to "think of it as a probability." I'm saying what you're describing is a probability distribution. Do you have a citation for "probability to express correctly the sentence/idea" bit? Because just having a latent space is no implication of representing an idea.

I don't understand. Deterministic and stochastic have very specific meanings. The statement: "To continue my reply I could say this word, more than the others, or maybe that one, a bit less, ..." sounds very much like a probability distribution.

GeorgiaTech's online Master's in Computer Science has lower costs. About $8k in total. I did it along with a full-time job since I only took one subject per semester.

Which aspects? Foundational textbooks would focus on principles, not necessarily implementations, and don't go "outdated" the same way a snippet does.

To be fair, he never said "every single role at your company requires a secret clearance." He has specifically mentioned that you can get by even without them. You have misquoted.

I think there's merit to both design docs and prototypes.

At the same time, your argument that "you'll be replaced by GPT in no time" is also an opinion that you've not supported with any data; the same thing that you're accusing the OP of.

I mean If I stopped reading opinions, 99% of the HN comments would disappear.

Good for you.

I agree with you. I've wrote very big web applications in Rails, Django, and Java/Spring.

I understand the argument for things being "explicit" in Python, but then I just prefer Java if I'm going to be very verbose about what I want (I'm only talking about backend APIs here). It's just my opinion that whether explicit is good or not depends on the level of abstraction we're interested in, and I don't believe that explicit is always good. (I like the concept of meta-algorithms, for example)

But if there's a one-person project that needs to be scaled quickly, I prefer Rails. (The article mentions how Django makes model fields explicit in the models file, but doesn't talk about schema.rb in Rails which doesn't require you to view each migration to know how the database looks.)

Yes, big projects in any language can get messy, but that's a software engineering problem, not a framework problem.

I recently wrote a FastAPI project that was db-driven, with all the necessary test cases, etc. The amount of lines it took to express the controller, the schemas and models separately, the dependencies for auth and stuff, and especially elaborate test cases was pretty substantial. Yeah, the code was all explicit, but it was not enjoyable.

Yes, which is why the rest of my sentence is important. I said, “I haven’t found a need to use a proprietary language.” I didn’t claim that one will never be needed by anyone. :-)

Yes, I’m aware, which is why it reminded me of PureBasic with which I’m more familiar. :-)

I'm familiar with PureBasic (although didn't use it a lot). I got introduced to it in the 2000s (2000-2007 or something), along with DarkBasic, GameMaker, and the likes.

In today's era, however, I have not yet found a need to use a proprietary language.

This is a very good comment and does reflect my experience. As engineers, we're the only people who can estimate something close enough, and it becomes our job to do that while taking into account the risks.

Our bad assumption is in thinking that only the final output matters, regardless of when and where it is delivered. Like saying that it only matters if the train arrives at the station, regardless of when it does.

The problem is anyone depending on us downstream will get impacted. And yes, estimation is tough, requires foresight, and maybe a lot more things, but that's what being a professional means.

This is it. I'm a hardcore engineer at heart who has a lot of these sales, marketing, and product folks as friends, and can attest to the fact that they also have constraints.

The whole world runs on deadlines and timelines. Even a president is elected for a specific duration. If you're in a B2B setting, the customer demands (sometimes even contractually binding) at least the Quarter when something will be delivered.

Time is the only common denominator by which different activities can be coordinated. Without some heed to time, there will be no coherence.

If your knowledge of Python comes from JavaScript, I would not blame Python for it. It's the failure of the person to not "read the instructions" and assume instead. Maybe conduct interviews in languages that you're familiar with?

By that logic, the 30% Apple tax must be "extremely fair" because all iOS apps are built on both the software and hardware built by Apple, which are arguably far more complex than a piece of PHP software. (/sarcasm)

Being "built" on something does not in itself imply value. What has to be taken into account is also "value that is added." If the underlying platform had high intrinsic value in itself, then no value would have had to be built on it to profit from it, which would mean WP Engine would not be the only entity profiting from it.

And to add to the reductio ad absurdum, are PHP engineers required to "donate part of their salaries" to the project just because "they're profiting off of that language?" (/sarcasm)

Looks awesome. Compliance in itself is a huge space that can benefit from a lot of automation. And the pain point of avoiding back-and-forth and giving quick feedback is real, because people usually don't read instructions, but rely on "feedback" to instruct them. All the best!

As someone who has done a lot of those things (beyond programming), I can assuredly say that "writing about something" is neither a proof of understanding nor a proficiency in a subject. (Case in point, our educational system)

But wouldn't the value of data be reflect in the cost of exploitation? (By cost of exploitation, I don't mean to say the resources needed to exploit, but what a company would stand to lose if exploited). The values of the variables, sure, can be different. I don't see why the equation has to be.

During my Master's, security was one of the subjects I took. It started with an equation that related risk (how much you'd lose if something bad happened), the probability of that risk, and the cost of mitigating that risk. The instruction being, one tries to find a mitigation that costs less than the exploitation of the risk. And note here that "cost" does not refer to just money, but could be computational cost, energy consumed, etc.

Yes. Although I’d like to see a deeper investigation. Of course, quality of completions have improved. But there could be a confounding phenomenon where newer folks might just be accepting a lot of suggestions without scrutiny.

The title is correct. But I agree that saying "50% increase" conveys less information than saying "1% to 1.5%" (because you don't know 50% of what), and seems to have been used only to make the title more dramatic.

Good points. Although, having worked on many high scale architectures before, I always err on the side of thinking that any technical solution of this magnitude would have far too many nuances for a blog post to capture. And I believe Figma’s post also talks mainly about the common denominator that’s easier to communicate to the external world.

For me to understand their choices, I’ll first have to understand their system, which can be a task in itself. Without that, I’d not become aware of the nuances, only general patterns.

Congrats on writing the book! As I'm a student of GeorgiaTech, I have access to Oreilly's Safari. I just found the book over there and am looking forward to reading it.

The 2nd point really resonates with me. I was reading this book called “Ultralearning” and feedback was one of the biggest points mentioned there. People who get prompt feedback can immediately use it to learn from experience. People who don’t have to keep guessing of what they’re doing is right or wrong.

I might need someone who has done the on campus version to pitch in. However, I’m assuming that the assignments are on par with the on campus version, and proctored exams almost too.

The difference is in the number of subjects that you can take per semester, and that also adds to the difficulty. Also, I’m not entirely sure but some courses are bound to be labs heavy. Those courses probably are not available on omscs.

[OP here] I agree with you. Even if GT's courses, the course material is pretty standard and doesn't actually drastically increase your understanding. It's the assignments that give you the most value, which are similar to Udacity Nanodegree's projects.

[OP here] You're right. I actually did my bachelors in Electronics and Telecommunications engineering, which can only be done in college because of the expensive equipment required to learn anything of value.

[OP here] Yes, you're right. In my case, I had the fortune of starting to program at a very young age, about 11. I started doing these courses to get on the whole ML/AI bandwagon (although I had been exposed to neural networks in the early 2000s).

The broad understanding I came to was, it's important to make your own conclusions, and not take anyone else's conclusion for granted, whether it comes from universities, courses, or even blog posts like mine. In reality, however, it's difficult to devote that much time to personally research everything. So you should at least do it with the things that matter the most to you in life.

Cool project. On a side note, Apache Drill (https://drill.apache.org/) is also worth checking out. The homepage doesn't make it very apparent, but I've used it to query CSVs, etc., using SQL.