Who are these imaginary people going to websites and just being like "I'm not spending $10 on this!! I'll just clone it with $500 worth of tokens!"

I don't think they exist, but I think there are loads who would go to this website and then leave, preferring to download free higher-quality apps instead.

Still cheaper than spending $2,300 to initially build it.

Eventually the cost will go drastically down as always.

From what I've heard, personas give a greater chance that the LLM will answer confidently.. and also a greater chance it'll hallucinate something when the data is sparse. Supposedly "grounding" the personas on real documents/web searches is the best approach. Anecdotal though.

I don't think we're going to be "dependent", because I can't really think of anyone that "needs" this stuff (well, unless you're like attempting to build a business off skills you don't have). I guess this really comes down to if you believe the productivity story. I don't. I think there are some gains, but the evidence that isn't just anecdotes from vibe coders seems to be modest.

... and building a business off of skills you don't have based on a strategy already exists! You use capital to pay humans that do have the skills.

Or capital a comparable sum to pay an AI to approximate the skills of humans I guess is the proposed future?

Play stupid games, win stupid prizes.

Well, in the brief window that I got to test Fable 5, my brief review is: somehow an (already specced!) minor feature in my 150k loc codebase ended up costing.. $153! For like, an hour or two worth of work and maybe 8 or 9 requests overall. I'd say it was not remotely worth it.

I asked it to tweak the fonts/colors of a very very simple static page and it blew through $35 (which is a lot for me lol; it's 10 days of my monthly codex plan).

You shouldn’t be using Fable for that, that’s Haiku work.

I think they were hoping for more advanced model’s more advanced “taste”

This is common sentiment amongst many users

I will never understand why people recommend using models with the capabilities of early 2025.

They cannot even move 10 existing lines of code around without breaking it in the process half of the time.

I very much doubt they are up to the task of implementing any sort of plan with a reliability that allows to complete the work faster than writing the code by hand.

If I used a racecar to go 25mph in a residential neighborhood, I’d make a similar conclusion.

It was made available in my subscription so I tested it out. I'm glad I tested it in a subscription, since I'd be pretty irritated if I had spent that amount of money accidentally in API usage. I guess what I've learned is what I already know, which is that the newer models seem to increase costs a lot with no perceptible benefit to my workflow.

I said "cost", not "cost me". I use `ccusage` to track what my unsubsidized token spend would be since I'm sure these subscriptions won't stick around forever and I want to have a realistic idea of what these things actually cost in a professional setting.

To be fair though, even if it's not costing me that much it's evidently costing Anthropic a pretty penny, I'm up to like $800 in spend on my $200 subscription in less than a week.

What makes you think the API based cost is the cost to Anthropic?

Well, they're selling a non-essential commodity in a highly competitive industry with about zero lock in or customer loyalty, using the standard VC playbook of "capture the market then worry about costs". I don't see space for large margins in there, and if there are margins they're probably recent because of the IPO.

Hard agree

I'd have to see the implementation in the editor to really know if I'd want it or not, but my initial impression is that I don't think it's something I would use. I've worked on a couple of products with in-editor collaboration built in (not text editors, but, content creation tools). For the most part collaboration just wasn't something users wanted that much. The problem was that the in-editor collaboration was only really useful for things that were 100% on the web, it never caught on with the more file-centric workflows. It works in something like Google Docs because it's natural that your data is in the cloud there, but for something that lives on your filesystem it feels like it takes extra steps that most people aren't going to be willing to do (especially if they have to use an editor different from their normal one)

I guess it would be perfect for pair programming in a remote work environment.

When I use these models, I honestly can't tell much of a difference between any of the frontier ones. Admittedly I'm not sitting around benchmarking them during the hype cycles, but as a coder I have zero issues switching to whatever's cheapest. Custom harnesses or whatever are not much of a moat (honestly Claude is so buggy right now that I've been using codex and opencode just so I don't have to deal with a flickery mess that screwed up my arrow keys)

I just don't see how being the "premium" provider really works if much cheaper models are basically good enough.

You should probably be more pessimistic about Meta. Look at their last major venture, the Metaverse, which was basically embarassing. Their AI strategy is incoherent.

I just looked at ccusage for a personal project. In 5 days (doing it as a hobby) I've managed to spend $250 in API tokens on a $200 subscription. 5 days, and thats on one computer (I split time using 3 of them). If I had to pay $2000 a month -- no fricking way, not worth it.

Elon Musk just got the rules of the NASDAQ changed so he can more or less force index funds to buy his shell company and take money from people's 401k. Feels very grifty to me.

That is a very misleading summary of what happened.