The problem is that we were on the Team plan which is a flat rate plan.

You raise a valid point but it has no substance. Management confronted us and said "you are doing it wrong!" And I stopped them "no, we aren't. even if we are, we can try to economize on tokens. We can get a 100% improvement, 200%? 500%?

It's still order of magnitude LESS than what we were doing and it's costing us, in dollars, twice what we were paying before.

You see, the problem with the Enterprise plan is that, unlike Team or Pro, it does not include any (subsidized) usage. Every token counts towards the budget. And corporate either did not read the fine print, or they had more than 150 developers (the maximum allowed by Team plan) and they were forced to switch.

They were expecting it would last for the entire month, like the Team plan was.

But it turns out the Pro and Team plans are subsidized and include far more tokens, that reset every week, than the "actual price" that they charge large companies.

It would be a valid point if one or two devs got it wrong. But it happend for all of us (and we work in different projects).

I'm Argentinian. And I'm speaking from experience.

and they are not yet charging for the "tokens" what they need to charge to become profitable

We recently got access to Claude Enterprise and found how expensive it is. We were given $45 a month of budget. Everyone in the team blew through that in 2 days. And considering this is still being "subsidized" I honestly don't see what's the future for "AI Coding".

As to HR, a critical function of HR is resolving conflicts.

No it's not? HR's only role is to protect the company from their employees.

HR is not your friend. Raise any concerns with HR and you will be flagged as problematic. They are not there to resolve conflicts. They're there to AVOID conflicts. And the best way to avoid conflict is not having "conflictive" people in the first place.

HR does not work in your interest. HR works in the interest of the company.

HR is there to document your mistakes so the company can fire you cleanly. That's HR's main function.

You didn't answer what they were asking.

They are behind CGNAT and Cloudflare provides a cloud VPN service that you can use to dial out, and then open ports in the cloudflare console, into your computer.

You are talking about two different things cloudflare does.

well, at my current job they use NoSQL, in this case it's DynamoDB and it's been frustrating at times. So I asked the question: why are we dealing with these problems day in, day out, if the problems we're trying to solve have been solved half a century ago with SQL?

The answer is cost. The way we access data may be convenient to do with SQL, but it's also expensive. We have big (not webscale but large) volumes of data coming in every day. Having this on SQL would cost us tens of thousands a month. Keeping it in DynamoDB costs us a few hundred. And it's stupidly fast - if we wanted to get that kind of performance from SQL we'd have to pay for a supercharged overprovisioned server.

And honestly it's been fun. It's turned "boring business software development" back into more of an engineering problem.

I see the problem as a more "get off my lawn" types here. They have fully adopted "vibe coding" as "anything made with AI assistance" as much as older people call anyone younger than them "millennials".

There's a big difference between an experienced programmer providing the AI with clear, concise prompts and guidance; than having someone with zero knowledge trying to build an entire app from scratch.

One is "augmented capabilities", the other is vibe coding. But the haters here just refuse ANY sort of AI involvement.

I refuse to believe the claim that "this would require billions of dollars and at least five years to get a factory operational."

There is clearly enormous amounts of money circulating in the industry right now. If a company like Nvidia genuinely wanted to manufacture its own memory, it absolutely could. Even with initially poor yields, the economics could still work. A 50% yield rate is far less concerning when RAM prices have increased by 200%, especially for a company purchasing memory in massive volumes alongside its hardware partners.

From my perspective, this looks less like an unavoidable technical limitation and more like market consolidation and price coordination. Companies have become comfortable charging substantial premiums for RAM, and the current situation provides a convenient justification for it.

I was never offered a free upgrade path and I only have 2 accounts: mine, and the admin one they force you to pay for. I was on the legacy plan and they forced me to pay.

A news aggregator should write blurbs like a news site, and not assume every reader knows what every proper noun is. Plex, an animal husbandry fanservice anime sharing company, ...

they also remove drag-hover-drop . it's so infuriating to have to organize your windows in a specific way to drag a file over to another window, OR use ctrl-c/ctrl-v

it was as easy as drag the icon to the next window "through the taskbar" which made the other window come front, and drop the icon.

i guess they removed that option since they started forcing taskbar grouping by default. a feature i remove from every windows and KDE machine I set up. I don't see any benefit in "grouping" or "compacting into an icon". if i wanted that behavior i'd just get a mac.

we have two options: live with a bug, or hope the fix doesn't introduce another bug.

i mean that's not a bad thing either. I sometimes DO NOT want to learn "new to me" things. I've been contributing to an ancient, but still used software called Xastir. It's VERY OLD spaghetti code, low level X11 with Motif. I DO NOT want to learn Motif. It's not a marketable skill or something I'll ever need. But I let the AI code a few contributions (one of them was replace some parts with Cairo fonts for antialias in high dpi scerens, and the other was fixing a very old screen drawing routine that took 2-3 minutes on a Raspberry Pi 2 and cut it down to 5 seconds). Could I have fixed this bug? Not even in my wildest dreams. Do I care how it was fixed? Oh no. No I don't. I just checked that the output of the LLM was reasonable.