> They seem mostly useless in a sufficiently large codebase especially when they are messy and interactions aren't always obvious.
What type of documents do you have explaining the codebase and its messy interactions, and have you provided that to the LLM?
Also, have you tried giving someone brand new to the team the exact same task and information you gave to the LLM, and how effective were they compared to the LLM?
> I don't know how much better Claude is than ChatGPT, but I can't get ChatGPT to do much useful with an existing large codebase.
As others have pointed out, from your comment, it doesn't sound like you've used a tool dedicated for AI coding.
(But even if you had, it would still fail if you expect LLMs to do stuff without sufficient context).
This is a major concern for junior programmers. For many senior ones, after 20 (or even 10) years of tenacious work, they realize that such work will always be there, and they long ago stopped growing on that front (i.e. they had already peaked). For those folks, LLMs are a life saver.
At a company I worked for, lots of senior engineers become managers because they no longer want to obsess over whether their algorithm has an off by one error. I think fewer will go the management route.
(There was always the senior tech lead path, but there are far more roles for management than tech lead).
Aren't you still better off than the rest of us who found what they love + invested decades in it before it lost its value. Isn't it better to lose your love when you still have time to find a new one?
Depends on if their new love provides as much money as their old one, which is probably not likely. I'd rather have had those decades to stash and invest.
A lot of pre-faang engineers dont have the stash you're thinking about. What you meant was "right when I found a lucrative job that I love". What was going on in tech these last 15 years, unfortunately, probably was once in a lifetime.
Imagine a senior dev who just approves PRs, approves production releases, and prioritizes bug reports and feature requests. LLM watches for errors ceaslessly, reports an issue. Senior dev reviews the issue and assigns a severity to it. Another LLM has a backlog of features and errors to go solve, it makes a fix and submits a PR after running tests and verifying things work on its end.
They make a quick mention of Turkey. Turkey probably perfected this technique - they've been abusing Interpol this way for years. Many people (e.g. Americans) who have openly criticized Turkey (blog post, Twitter, etc) have found themselves on the Interpol list and in trouble when traveling abroad.
Turkey is quite tolerant - there are quite a lot of women in bikinis and alcohol in the tourist resorts in the south. The majority of the locals don't do that though.
All I can do is throw my anecdotes into the pool: I mostly have met two types of Iranians: Those that fled in the 80's post-revolution, and those that come to the US for university (90's, 00's, and 10's).
All of them have been anti-regime.
I have met a few that came for other reasons (not education and not the 80's stock). Yes, those are either pro-regime or neutral.
My guess is that what rayiner says is correct: The majority of the Iranian diaspora in the US is self selecting and not representative of the full population.
Yeah I'm surprised that they announced it but not the vendor name. I'm sure Google with their infinite resources already know which vendor it is. So who are they hiding it from?
> Keeping the reader glued to the screen is not the primary goal of writing.
This is common advice in English classes and it predates the World Wide Web (and likely the Internet).
Hook them in the first few sentences or lose them.
And yes, of course, it does depend on who the intended audience is. You wouldn't do it in The New Yorker.
> You don't write for the reader. You write for yourself first. Readers sometimes, just happen to appreciate it about as much you do.
Depends very much on the medium. It's definitely not true that most professional writing is written for the author's sake. It is for an audience. Read books on writing and you'll often find the advice to cut out things if they won't interest the reader - no matter how valuable it is to you.
I myself struggle with this. Some years ago, I took a trip to my childhood home in another country after being separated for decades. Almost none of my friends from the time have been there in decades either. I made notes during the trip, and when I got back I started writing what I saw, and shared it with my friends who grew up with me. How various neighborhoods have changed. Anecdotes from my childhood tied to those places. And a lot more.
I got 30% done, and then decided to hold off sharing till I'd written the whole thing. I now have a first draft. It's the size of a proper book. It contains a lot of stuff that is of value to me, but likely not to most of the (small) audience. I know if I share it with them, chances are high no one will read it.
On the one hand, the stuff I wrote is highly valuable to me - it's become an unintentional memoir. But on the other hand, I do want to share quite a bit with my friends, and I know they'll value it if they actually read it.
I'll either have to cut a lot out, or write two versions (impractical).
The point being that even when you have a very limited audience, it is important to care about them and sacrifice your needs to an extent.
Never cut out stuff that you felt important to include. Just forget the reader. The content you write should reflect you, not the reader. It's your expression. Don't make it a sales pitch, or a reflection of average reader's taste.
I get scared when an author is talking to me, the reader. I stop reading when they pretend to be aware of my context. Things like "So you are reading this book because you want to learn about AI" sounds very cheap.
Also I hate when the actors on TV suddenly start talking to the viewer about what they did and why did etc. Disgusting.
Audience want to observe the performers, not converse with them. Your best performance comes out when you are not much aware of the audience. Like a child playing, ignoring people around.
> The content you write should reflect you, not the reader.
It's not a binary proposition. One can write for both, as long as they're willing to compromise. I'd rather something be 80% good and have an audience than 100% good with no audience.
Why write if no one will read? I have no idea if even I will read it - I already "read" it while writing it. Maybe in 20 years I'll revisit? I don't know.
> I get scared when an author is talking to me, the reader.
> Also I hate when the actors on TV suddenly start talking to the viewer about what they did and why did etc. Disgusting.
That's your quirk. Don't assume others have the same preferences as you.
> Things like "So you are reading this book because you want to learn about AI" sounds very cheap.
I don't see anyone advocating writing like this.
In general, most writers disagree with you. I take it you've not heard of "Kill your darlings"?
Things like "kill your darlings" becomes a doctrine purely because of business goals, not artistic goals. It asks you to erase what you like, so that you can sell it. There is nothing visionary or prophecy-like wisdom in that. It's pure selling, which is considered divine by the capitalist western cultures.
The thing with young kids is they tend not to be good at timing restroom breaks with the availability of charging. By the time they tell you, you need to stop at the nearest gas station - they can't wait for you to drive 20 miles to the next charging station.
> No maintenance on, spark plugs, timing belts, gearbox. No oil changes. A quieter ride, especially nice on a road trip.
I'm thinking of getting an EV, so I'll see how much I like this. I can say that this is pretty much not a hassle for me with my ICE car - over the last 20+ years. But then I tend to buy reliable cars and didn't fall for the manufactured "3 months or 3000 miles" rule.
I keep track of all my costs. I average about $500 a year in maintenance (includes tires, oil changes, brakes, etc). I just checked with the insurance company - the increase in my annual premiums for the EV car I'm looking at is $400 more than if I got an equivalent ICE car. And one still needs to change tires, etc on an EV. So the repair/maintenance savings aren't there.
> The insurance part will settle out over time as they get more data I would imagine.
I'm not so sure. The issue is two-fold: First, If you get into an accident and you're at fault, the average damage is a lot more than with an ICE, due to the much heavier weight. Second, compared to an ICE, just about any repair is a lot more expensive. If some of the battery gets damaged, that's crazy expensive. There's also not a good ecosystem for parts - they are more expensive and less modular than with an ICE (or so I'm told).
It apparently is a lot more common for EVs to be declared a total loss compared to an ICE just because of the expense to repair.
> 500$ a year is very little for any car
This is over 3 different cars. And all of them very old (I bought two of them when they were 8 years old, and another when it was 15 years old - still driving that last one).
About $80/year for oil changes. That's it. Then every once in a while there is an expensive repair (brakes, tires, some engine problem, etc). Doesn't happen every year - so the average comes out to $500.
I also don't go to the official dealers. Everything is more expensive with them.
And yeah, the cars are old, so few electronic parts to repair. I imagine if I get another 8 year old ICE, the annual cost to repair will be more just due to the extra safety systems that can go wrong.
> but I opened a Nissan leaf for 8 years and spent less that 2K, of which 1K was for the AC
Leafs are the best case scenario. They're small, not heavy, and thus don't have much tire wear.
I haven't implemented moderation, but in principle it's "easy". On a given post, store as an attribute the IDs of messages you don't want to appear (or even its descendants). The JS will have access to the info and can just filter it/them out.
What type of documents do you have explaining the codebase and its messy interactions, and have you provided that to the LLM?
Also, have you tried giving someone brand new to the team the exact same task and information you gave to the LLM, and how effective were they compared to the LLM?
> I don't know how much better Claude is than ChatGPT, but I can't get ChatGPT to do much useful with an existing large codebase.
As others have pointed out, from your comment, it doesn't sound like you've used a tool dedicated for AI coding.
(But even if you had, it would still fail if you expect LLMs to do stuff without sufficient context).
reply