I've spent quite a bit of time with Codex recently and come to the conclusion that you can't simply say "Let's add custom video controls around ReactPlayer." You need to follow up with a set of strict requirements to set expectations, guard rails, and what the final product should do (and not do). Even then it may have a few issues, but continuing to prompt with clearly stated problems that don't meet the requirements (or you forgot to include) usually clears it up.
Code that would have taken me a week to write is done in about 10 minutes. It's likely on average better than what I could personally write as a novice-mid level programmer.
>You need to follow up with a set of strict requirements to set expectations, guard rails, and what the final product should do (and not do).
that usually very hard to do part, and is was possible to spent few days on something like that in real word before LLMs. But with LLMs is worse because is not enough to have those requirements, some of those won't work for random reasons and is no any 'rules' that can grantee results. It always like 'try that' and 'probably this will work'.
Just recently I was struggled with same prompt produced different result between API calls before I realized that just usage of few more '\"' and few spaces in prompt leaded model to completely different route of logic which produced opposite answers.
This is very true. But each iteration of learning quirks and installing guardrails carries value forward to later sessions. These rough edges get smoother with use, is my point.
It sounds like it takes you at least 10 minutes to just write the prompt with all the details you mentioned. Especially if you need to continue and prompt again (and again?).
Not the OP but, easily. My tasks are usually taking at least that, but up to hours of brainstorming and planning, sometimes I’ll do this over days in between other tasks just so I can think about all and pros and cons. Of course this has always been the way, but now I have an ongoing Claude session which I can come back to at any point, which is holding the context along with my brain. It’s much easier to keep the thread of what I’m working on across multiple tasks.
> I’m writing some (for me) seriously advanced software that would have taken me months to write, in weeks, using Claude and ChatGPT.
Do you understand the code?
What was the speed up from months to weeks? You just didn't know what to type? Or you didn't know the problem domain? Or you found it hard to 'start' and the AI writing boiler plate gave you motivation?
In my experience with AI tools, it only really helps with ideation, most things it produces need heavy tweaking - to the point that there is no time savings. It's probably a net negative because I am spending all of my time thinking how to explain things to a dumb computer, rather than thinking about how to solve the problem.
The main advantage is I can run it in parallel and iterate often.
The speed up is also avoiding looking up reference manuals endlessly just to produce some Qt Widgets.
I’m a fairly recent convert, I only started “vibe coding” a couple of months ago, after hearing how good Opus was. I had been a skeptic until then.
I am a decentralist by nature and prefer open standards and self hosting. I’ve had my own *nix servers since I was twelve (nearing forty) so it really pains me to admit how good it is to use these corporate products.
I am not a programmer by trade.
I use it to write software for my domain of expertise. The value of what I am creating is enormous.
Both ChatGPT and Claude produce good code, in my opinion.
If you have a good network CI/CD pipeline and can trace the time of deployment to when the errors began, it should be easy to reduce your total TTD/TTR. Even when I was parsing logs years ago and matching them up against AAA authorization commands issued, it was always a question of "when did this start happening?" and then "who made a change around that time period?"
I've said it elsewhere, but there must be some point where both parties could have continued operating while reddit made money. Even if it wasn't much, they probably could have covered their costs if they had worked out a pricing plan with the developers.
Based on what I've seen, they went above the standard for API pricing in their industry. Comparatively the pricing offered to the 3rd party app developers was akin to gouging, and was not meant to be realistic.
if 97% of the userbase doesn't use third party apps and "And the opportunity cost of not having those users on our platform, on our advertising platform, is really significant" are true statements then those 3% were extreme power users
The post by the Apollo dev made it sound like they had gone back and forth for a while before the final pricing was laid out. I wonder why reddit didn't come up with a price that worked for all parties. Isn't some money better than none + ill will from the community?
In the recent AMA, they claimed that the average cost per user would be around $1/month. Apollo dev believes it will be much more than.
I feel Reddit could have easily come up with a model where the API usage gets tied to premium account and that decouples apps from API charges.
Reddit has now created a new problem for itself which is that a huge user base that previously didn't bother to look at third-party apps has suddenly become aware of third-party apps. So now, they will start bypassing Reddit's official apps that generate revenue for Reddit in favor of third-party apps that seemingly have a better UX and are ad-free.
Well, maybe, but the API is free only until the end of the month, so… seems like a very temporary problem for Reddit (assuming that the mods generally fold, as I assume they likely will).
I highly doubt a different dollar amount would have changed the outcome. Reddit wanted $2.50 per user from the Apollo dev. Would $2 have been more palatable? Or $1.50? Would the extra few cents per month really make the difference between a user subscribing or not?
The conversation is really only about free vs not free. Everything else is a smokescreen.
Eh, at some threshold, of course a lower price would have changed the outcome. It sounds like the Apollo dev thought API costs came in at 10-20x what he’d been expecting. I think there’s obviously a huge difference between paying Reddit, say, 50 cents of a $3 net payment (assuming a sub for the app is $5/month) and having a $5/user/month cover charge to Reddit as the price of admission, and then having to build a viable app business over that. Seems dubious.
The average user doing ~300 api calls per day would do about 9000 calls per month and at $0.24/1k calls would be $2/user
Apollo also does polling of the message box for each user for push notifications ( https://github.com/christianselig/apollo-backend/blob/main/i... ) which currently has a rate of 1/minute/user. This is another 1.4k calls per day and changes the price that would be paid.
Not sure why you're making up stuff still and cherry picking constants to try and pretend you're right.
Those rates are BEST CASE. I linked the code directly to you that actually does it in your previous comment.
They queue 100[0] users every 5[1] seconds to pull their status, they then update the next check timestamp in the db to be at now + the constant you quote[2] which they use for rate limiting, so at most once per minute.
So unless they have under 1,000 users, then it won't ever be "every single minute."
OK, so whatever numbers you want to use, whether it’s $2 in direct Reddit api access charges or $5, same order of magnitude, probably neither moves the needle that much relative to the other.
But I think there’s a huge difference between that charge being 20 cents and $2 (or 50 cents and $5).
From the API calls that you'd need to pass on to the user, this is a difference of $2/month for no push notifications (9000 calls) vs $10/month (9k calls + 30 * 1.4k calls is about 50k calls). Add on top of that a 30% Apple subscription cost (15% after a year), and we're to $3/month and $13/month and a bit of padding and a slim profit and you're at $5/month and $15/month.
If Reddit is saying "it's $2 for the 300 calls per day" that is claimed for the mobile app usage they're correct looking at what they charge for API calls based on the developer saying that the average user does 345 calls/day from the mobile app.
From the app develop perspective who also has a polling back end and an apple subscription cut and maintaining the same profit as before on top of it all, they're likely looking at closer to $20/month for that same user (noting that previously they were charing $4.99 and it was almost all profit for that user).
I'm curious to see what will happen with the bagholders of office real estate. Will they try to reinvent themselves as mixed use residential/commercial? Or is it possible we will see a wave of defaults as vacancy rates continue to climb?
I've noticed how many fluff-filled articles I see lately about 'needing to get back to the office' in many publications that seem to have a large number of advertisers selling office space. I don't have any numbers, but it seems pretty crazy how many of these have been popping up in different 'business journal' types of sites.
It's not crazy at all. Our entire society is built around exploitatively extracting maximum value from each other and that often requires creative marketing to invent want where none existed before.
I don't think there is a fundamental reason for that. Office buildings are designed to be reconfigured. I guess you'd need to rip out + replace the plumbing + HVAC (or only put in a few ultra-luxury units per floor).
Office buildings tend to have all their plumbing centrally located. Converting to residential means spreading it all around. It's not like just moving partitions around.
Sprinkler systems tend to be evenly distributed around floors so the conduits etc. to run additional piping are already in place. Not every building has such systems but most constructed after 1980 do.
The sewer pipes and vents have to be run everywhere, and they have specific requirements. Sewer pipes, for example, half to have a slope. The vent pipes have to go up. There may be floor loading problems. There's going to need to be a lot more electrical wiring.
There's going to be a massive PR campaign to get employees back into the office. Story after story about how going back to the office increased productivity etc etc. They're going to give businesses the hard sell to lease office space again and get their employees back into the office.
Not everyone will go along with this, but if they could get 80% of the people back it'll probably be enough to prevent commercial real estate from collapsing.
Do we know how much of commercial real estate is owned by foreign entities? ie. Chinese elites laundering money out of China into the West? I wonder if they have any protections to save their investments or is their money made from value generated in China just literally disappearing?
Mastodon is open source. I may be wrong but it looks like this platform is not. There's probably money involved somehow, which is why it is getting the Jack Dorsey bump.
Code that would have taken me a week to write is done in about 10 minutes. It's likely on average better than what I could personally write as a novice-mid level programmer.