how do RAG implementations work with generic prompts vs specific prompts? meaning, there are prompts that could easily be answered by the base model itself and doesn't require RAG. but some prompts might involve questions about something proprietary where RAG is actually useful.
so is the default to just run the RAG search index on every prompt and if it returns nothing then you get the plain answer from the base model otherwise you get the augmented answer?
Oh, yeah. I hear people in their thirties, complaining about being treated as "old."
When the CEO is 26, then it's easy to have a young workforce. In The Days of Yore, the C-suite was generally folks in their fifties, and the youngsters were forced to work with their chronological seniors. The older folks had the money and power, but they needed the creativity and energy of the younger workers.
If the workforce is all older folks, you get shipping product that no one wants. If the workforce is all younger folks, you get ... FTX.
Look at some of these "full team" photos, for many of these new companies. You won't see a grey hair anywhere, and, if you do, a bit of research usually shows them to be a Principal.
I fully admit that it's a world that doesn't want or need people like me. It really pissed me off, at first, but, in the aggregate, it has resulted in the first truly happy work that I've done in decades, so it's all good.
That said, some of my former employees are near my age, and were able to find work, but it took each of them, several years, and they didn't have the baggage of being a former manager.
That's the thing: I'm not working there, because they'd never hire me.
In my (limited) experience (about five years ago), the recruiters were always engaging, helpful and friendly. However, as soon as one technical person (usually young) got involved, the temperature dropped about thirty degrees.
I may not be God's gift to programming, but I'm not that bad. I do, for example, have over thirty years' experience shipping extremely high-Quality products, in very challenging environments. In many fields, that usually commands a tiny bit of respect.
In extremely regulated environments you would typically have someone with legal background to boil down the requirements into their absolute essence and give them to the dev team.
I work in one of those environments and we do have those people (lawyers). The next question is how should we prioritize solving for potential legal risk vs reliability risk vs new feature work. You’d expect it’s black and white, I promise you it’s not. The PM has to figure that out.
Good product managers I've worked with are better than me at navigating the domain and compliance worlds as tourists as they are tourists in the engineering side as well. David Ricardo was right, comparative advantage is a good thing.
If I consult a solicitor about data protection regulation and they tell me what I need to build to satisfy the requirements of soft opt-in under PECR, that does not make them a product manager!
No, but it's something a good product manager does for you, and with the economy of comparative and potentially absolute advantages. This let's you focus on the craft of delivery without encumberance of spending research into compliance.
I don't consider the engineering having a deep understanding of the domain and compliance to be a "encumberance". The more engineers get into the domain themselves, the more they can make deep decisions about how to build the most effective solutions.
> This let's you focus on the craft of delivery
I don't give a rat's ass about the craft of delivery! What a reductionist thing to think of engineers as "delivery machines". Me and my team mates are not a factory that simply takes input from a task master and produces code to their bidding.
I care only about one thing: solving the problem! That encompasses learning, understanding and, finally: delivery. To do that, I need to speak to stakeholders MYSELF, I need to see how they are using the product MYSELF, and any regulation I am gonna implement, I need to understand in depth. I don't need somebody to summarize findings for me, I can make my own damn conclusions.
Sure, I will ask the domain experts of their opinion, but me and my teams will decide what to build, when, and how, and do we will do it together, based on all our shared learnings, not based on what a PM told us.
The developer teams, of course. They are solving the problems, so why should they not prioritize them? They just need to have other parts of the org heard, but if you are solving it, you get to choose how to do it, in what order and with what priority.
If that's not to somebody's liking, they are welcome to prioritize and solve the problems themselves.
What area of insights does the development team need to be able to prioritize for better utilization of their product?
---
> If that's not to somebody's liking, they are welcome to prioritize and solve the problems themselves.
Ah, I see. There is a concept known as a separation of concerns that prevents this on most teams.
There is a good saying for complex service delivery that most developers operate with in application development and support. If you want to go fast, go alone. If you want to go far, go together. That "together" brings more skillsets that most small engineering teams have on staff, and is typically not beneficial for them to focus on gaining.
> They are solving the problems, so why should they not prioritize them?
> If that's not to somebody's liking, they are welcome to prioritize and solve the problems themselves.
There's two types of problems that developers solve. Problems that they, themselves created (oh no, I wrote a bug), and problems that are given to them.
The second type (traditionally, the "business problems") are being prioritized and solved by the people who care about them. Namely, the CEO and the exco. Those guys set overall strategy, culture etc. and then the delegate the implementation and some of the detail to people with more expertise.
Developers do not run the business. They are delegated to to solve specific problems in the context of running the business. Not all of the business' problems will be solved by developers, but all of the business' problems should be solved in a coherent and harmonious way that accounts for the context of all the other problems in the domain (market, regulatory, finance & tax etc.)
For a specific product, the place where this context is managed is usually by someone wearing a hat with "Product Manager" written on it.
> if you are solving it, you get to choose how to do it, in what order and with what priority
In times of plenty, this works, and I think it can have high velocity. Especially with very small teams that all have high levels of direct investment in the result.
There are also times of famine, however, and the work that keeps the lights on is often not work that even invested software developers want to prioritize. When making money is on the line, there's always somebody who's there to make you eat your vegetables. Who, in your conception of the problem, is that?
> and the work that keeps the lights on is often not work that even invested software developers want to prioritize.
What kind of work is that? And why are the developers willingly not wanting to prioritize it, if the alternative of not doing it is becoming jobless? Do we need some kind of PM figure to say things like: "If we don't make this feature for the client, we are doomed, so start coding. Chop chop!". And do we need to keep that guy on the payroll for simply relaying a message?
Right - like, you'd think it's obvious that of course you'd just do the right thing. But developers are expensive and market research, talking to customers, planning for the future, etc. becomes rivalrous to building things. And that stuff is generally much less interesting, so without compulsions to solve those problems, they frequently then don't get solved.
"But if you don't do these things, the company fails!" Yeah, and look at all those small, engineering-heavy companies that do exactly that! And further--eventually, as you scale you reinvent division of labor because as much as we like to kid ourselves, software developers are not, universally, better at doing everyone else's job. (If you find yourself in a position where you are better at doing everyone else's job, you should probably find a new company to work at.)
If the PM does that, the developer has no idea what part of the requirements is important, and won't be able to deal with it when the requirements unavoidably break each other or reality. The developer has to be there too.
noob question but how does one actually use prettier? i find that the formatting just make the code less readable. is the goal to train yourself to be able to read prettier-formatted code as well as you could read your own hand-formatted code?
when i'm in the thick of it working on a piece of code, writing some logic and have everything laid out the way that makes sense to me and then i hit cmd + s and BOOM all the lines get shifted and wrapped and i'm ripped out of my flow i wondering what the heck it is i'm now staring at. sometimes i disable prettier on save and just run it on a pre-commit hook because at that point who cares that the code is unreadable to me.
I think the assumption that Ilya was the ringleader still holds and all the evidence still points in that direction. Adam represents the board NOW (after the signed letter and change of heart), but it doesn't mean that Ilya wasn't the key instigator at the start of this and all the way through Monday morning.
If the key instigator flipped, why would a supposedly disinterested third party continue to hold the line to the point where talks break down? Doesn’t hold up.
Maybe the other three weren't thrilled with Sam, and hadn't been for some time, but were just three of six so they didn't have the votes and weren't planning to press the issue until Ilya brought it up. With Sam and Greg gone, they are now three of four and even if they don't feel strongly that Sam was the wrong choice they may find it unappealing to vote to bring him back, and in fact it would take at least two of them changing their minds to vote with Ilya.
That’s possible, but I don’t think it’s an accident that coordinate with the firing of Sam they reduced the board to being outside-controlled, with only one actor who can be said to be seasoned in corporate politics.
The other two board members have nothing in their history to suggest they have the stomach or experience exercising agency to be a significant part of this resistance. But they might be pretty inclined to make a useful ideological stand.
I’ll reiterate a point I and others have made: the person most hurt by recent OpenAI events is D’Angelo. All he has to do is make the valid point that OpenAI’s charter is to make AI available to all and twist it a bit to his extreme personal advantage. The difference is that most people seasoned in corporate warfare can see he is not acting out of altruism but to his personal advantage, and apparently the other board members cannot.
Inversely to the other board members not having a history of individual stands, D’Angelo does, and he also lacks indications in his work history that he is concerned with altruistic actions in general.
1. He’s the actual ringleader behind the coup. He got everyone on board, provided reassurances and personally orchestrated and executed the firing. Most likely possibly and the one that’s most consistent with all the reporting and evidence so far (including this article).
2. Others on the board (e.g. Adam) masterminded the coup and saw Ilya as a fellow traveler useful idiot that could be deceived into voting against Sam and destroy the company he and his 700 colleagues spent so hard to build. He then also puppeteer Ilya to do the actual firing over Google Meet.
If #1 is real, he’s just the biggest weasel in tech history by repenting so swiftly and decisively… I don’t think neither the article, nor the broader facts really point to him being the first to cast the stone.
Wait I’m completely confused. Why is Ilya signing this? Is he voting for his own resignation? He’s part of the board. In fact, he was the ringleader of this coup.
It sounds like he’s just trying to save face bro. The truth will come out eventually. But he definitely wasn’t against it and I’m sure the no-names on the board wouldn’t have moved if they didn’t get certain reassurances from Ilya.
so is the default to just run the RAG search index on every prompt and if it returns nothing then you get the plain answer from the base model otherwise you get the augmented answer?