Hi if the Gemini API team is reading this can you please be more transparent about 'The specified schema produces a constraint that has too many states for serving. ...' when using Structured Outputs.
I assume it has something to do with the underlying constraint grammar/token masks becoming too long/taking too long to compute. But as end users we have no way of figuring out what the actual limits are.
Other than that, good work! I love how fast the Gemini models are. The current API is significantly less of a shitshow compared to last year with property ordering etc.
levocardia 24 minutes ago [-]
Add me to the list of "saw nano banana pro, attempted to get an API key for like 5min, failed and gave up." Maybe I am a dummy (quite possible) but I have seen many smart people similarly flummoxed!
You can walk into a McDonalds without being able to read, write, or speak English, and the order touchscreen UI is so good (er, "good") that you can successfully order a hamburger in about 60 seconds. Why can't Google (of all companies) figure this out?
andy99 8 minutes ago [-]
I tried at some point to sign up for whatever IBMs AI cloud was called. None of the documentation was up to date, when you clicked on things you ended up in circular loops that took you back where you started. Somehow there were several kinds of api keys you could make, most seemingly decoys and only one correct one. The whole experience was like one of those Mario castle levels where if you don’t follow the exact right pattern you just loop back to where you started.
It makes sense for IBM, seems like google is just reaching that stage?
JohnMakin 23 minutes ago [-]
because the bubble in which googlers exist is inherently user-hostile, even to their own detriment. been like this for a while
dannyobrien 5 hours ago [-]
The odd thing about all of this (well, I guess it's not odd, just ironic), is that when Google AdWords started, one of the notable things about it was that anyone could start serving or buying ads. You just needed a credit-card. I think that bought Google a lot of credibility (along with the ads being text-only) as they entered an already disreputable space: ordinary users and small businesses felt they were getting the same treatment as more faceless, distant big businesses.
I have a friend that says Google's decline came when they bought DoubleClick in 2008 and suffered a reverse-takeover: their customers shifted from being Internet users and became other, matchingly-sized corporations.
cortesoft 4 hours ago [-]
I have had way too many arguments over the years with product and sales people at my job on the importance of instant self-signup. I want to be able to just pay and go, without having to talk to people or wait for things.
I know part of it is that sales wants to be able to price discriminate and wants to be able to use their sales skills on a customer, but I am never going to sign up for anything that makes me talk to someone before I can buy.
Workaccount2 2 hours ago [-]
The number one rule of business that should just be passively reiterated to everyone working in any type of transactional field:
1. Never make it hard for people to give you money.
the_snooze 1 hours ago [-]
Parking apps don’t seem to care much for that. They know you’ll jump through their shoddy UIs and data collection because they have a local monopoly. Often with physical payment kiosks removed and replaced with “download our shitty app!” notices.
prasadjoglekar 22 minutes ago [-]
There's also the unfortunate stick of a much larger parking ticket that is even more trouble to contest.
Workaccount2 58 minutes ago [-]
Plenty of people on here looking to disrupt a market with tech...c'mon guys, get on it
Edit: On second thought, there is a perverse incentive at work (and probably one of the "lowest friction" ways to get money), which is issuing government enforced fines.
brightball 1 hours ago [-]
It depends on the environment.
If a platform is designed in a way that users can sign up and go, it can work well.
If an application is complicated or it’s a tool that the whole business runs on, often times the company will discover their customers have more success with training and a point of contact/account manager to help with onboarding.
AznHisoka 2 hours ago [-]
You are also a developer though, and developers are notorious for wanting self serve.
Someone who works in finance or conpliances might want a demo, or views those things as signals the product is for serious use cases.
dpkirchner 2 hours ago [-]
Sure, and they should have that option. But in my experience business-folks ask techies to evaluate services all the time, and ideally we can just start out in the low-/no-touch tier to feel things out. If that tier isn't available, us techs might just try a different service.
timr 6 minutes ago [-]
The kind of products hidden behind sales calls are generally the sort where the opinion of IC-level tech staff is next to irrelevant. With these kinds of products, the purchase decision is being made at a group level, the contract sizes are large, and budgetary approvals are required. It’s a snowball the size of a house, and it started rolling down the mountain months (or years) before it got to your desk. Literally nobody cares if you buy a single license or not, and if you (personally) refuse to try it because it doesn’t have self-service, you’ll be ignored for being the bad stereotype of an “engineer”, or worse.
About the only time you’ll be asked to evaluate such a product as an IC is when someone wants an opinion about API support or something equivalent. And if you refuse to do it, the decision-makers will just find the next guy down the hall who won’t be so cranky.
sh34r 3 hours ago [-]
> I know part of it is that sales wants to be able to price discriminate and wants to be able to use their sales skills on a customer
You say that as if it isn’t the entire reason why these interactions should be avoided at all costs. Dynamic pricing should be a crime.
pooper 3 hours ago [-]
> You say that as if it isn’t the entire reason why these interactions should be avoided at all costs. Dynamic pricing should be a crime.
Does segmentation also count as dynamic pricing?
--
The IT guy at Podunk Lutheran College has no money: Gratis.
The IT guy at a medium-sized real estate agency has some money: $500.
The IT guy at a Fortune 100 company has tons of money: $50,000.
Not parent poster, but I think a more practical approach is to ban secret pricing.
If everybody can see the prices that are being quoted/paid by other people, that exerts a strong moderating force against abuse.
It won't help you if there's a monopoly, but I consider that a separate problem needing separate solutions.
transcriptase 1 hours ago [-]
The entire lab supply industry is disgusting in this respect. The funding (and recent grants) that a given professor or research lab has is generally publicly available information that vendors will buy in easily digestible formats from brokers and companies that scrape the websites of major granting agencies.
All of their products, however realistically commoditized, will require a drawn out engagement with a rep who knows how much money you’ve received recently and even has an outline what research you plan to do over the next few years since even the detailed applications often get published alongside funding allocations.
The exact same piece of equipment, consumables required to use it, and service agreements might be anywhere from X to 10X depending on what they (as a result of asymmetrically available knowledge) know you need and how much you could theoretically spend.
lokar 52 minutes ago [-]
A group of research universities should start a non-profit co-op to produce this for them.
Getting just the university of California should be enough critical mass.
nicbou 2 hours ago [-]
Certain purchases (like health insurance in my country) should be a conversation, because the options are fiendishly complex and the attributes people typically use for comparison are wrong. The consequences are lifelong.
xboxnolifes 3 hours ago [-]
Pricing tiers are a form of dynamic pricing. Service free tiers basically couldn't exist without dynamic pricing, as they are subsidized by the paying tiers.
Sevii 4 hours ago [-]
That has definitely changed. Google AdWords today is one of the most unfriendly services to onboard I've ever encountered. Signing up is trivial, setting up your first ad is easy, then you instantly get banned. Appeals do nothing. You essentially have to hire a professional just to use it.
LiamPowell 4 hours ago [-]
Yet it's still absolutely inundated with scams and occasionally links that directly download malware[1] that they don't action reports on. I don't think the process needs to be easier if they already can't keep up with moderation.
It might seem vindictive, but these are the ads that google shows people who block all of Googles tracking or are new/blank profiles. Hear me out...
When Google has a bad/empty profile of you, advertisers don't bid on you, so it goes to the bottom feeders. Average (typically tech illiterate) people wandering through the internet mostly get ads for Tide, Chevy, and [big brand], because they pay Google much more for those well profiled users. These scam advertisers really don't pay much, but are willing to be shown to mostly anyone. They are a bit like the advertiser of last resort.
All of that is to say, if you are getting malware/scam ads from Google, it's probably because (ironically) you know what you are doing.
dekhn 3 hours ago [-]
The thing to understand about google services is that they see so much spam and abuse that it's easier for them to just assume you are a spammer rather than a legitimate customer, unless you go through other channels to establish yourself.
binsquare 4 hours ago [-]
Also adding onto this, it is impossible to get human support!
One of my co-workers left with an active account and active card but no passwords noted. The company gave up and just had to cancel + create a new account for the next adwords specialist.
fersarr 4 hours ago [-]
My attempts always had validation issues that stopped the ads from running but I never figured it out and stopped trying
obmelvin 4 hours ago [-]
I don't understand the multiple posts / comments I've seen about this.
That docs page has a link in the first primary section on the page. Sure, it could be a huge CTA, but this is a docs page, so it's kinda nice that it's not gone through a marketing make over.
* besides sponsored result for AI Studio
(Maybe I misunderstood and all the complaints are about billing. I don't remember having issues when I added my card to GCP in the past, but maybe I did)
leopoldj 3 hours ago [-]
As the article states, generating the key itself is easy. But getting credit and billing are the issues.
knollimar 14 minutes ago [-]
I have it running and calling but it's not showing the usage and I set it up the day gemini 3 came out
yawnxyz 3 hours ago [-]
I've to this day never been able to pay for Gemini through the API, even though I've tried maybe 6-7 times
If you bring it up to Logan he'll just brush it off — I honestly don't know if they test these UX flows with their own personal accounts, or if something is buggy with my account.
This is my experience as well in my personal account, however at work given we were already paying for Google Cloud it was easy enough to connect a GCP account.
But somehow personally even though I'm a paying Google One subscriber and have a GCP billing account with a credit card, I get confusing errors when trying to use the Gemini API
verdverm 3 hours ago [-]
Most of them are correlating gemini-cli experience (trash) with the broader access to Gemini via studio or cloud (not at all a problem)
amluto 3 hours ago [-]
> Gemini via studio
Excuse me? If you mean AI Studio, are you talking about the product where you can’t even switch which logged in account you’re using without agreeing to its terms under whatever random account it selected, where the ability to turn off training on your data does not obviously exist, and where it’s extremely unclear how an organization is supposed to pay for it?
rezonant 2 hours ago [-]
Yes, much like admin.google.com (the GSuite admin interface), which goes ahead and tries to two-factor your personal GMail account every single time you load it instead of asking you which of the actual GSuite accounts you're signed into you'd like to use...
amluto 46 minutes ago [-]
I love how the two factor screen has no obvious way to tell it that you want a different account.
Hint: you can often avoid some of this mess by adding the authuser=user@domain to the URL.
Every aspect is at least partially broken several times a day, and even when there isn't a temporary outage of something somewhere, there are nonsensical "blocks" for things that ought to just work.
I've been using the AI Studio with my personal Workspace account. I can generate an API key. That worked for a while, but now Gemini CLI won't accept it. Why? No clue. It just says that I'm "not allowed" to use Gemini Pro 3 with the CLI tool. No reason given, no recourse, just a hand in your face flatly rejecting access to something I am paying for and can use elsewhere.
Simultaneously, I'm trying to convince my company to pay for a corporate account of some sort so that I can use API keys with custom tools and run up a bill of potentially thousands of dollars that we can charge back to the customer.
My manager tried to follow the instructions and... followed the wrong ones. They all look the same. They all talk about "Gemini" and "Enterprise". He ended up signing up for Google's equivalent of Copilot for business use, not something that provides API keys to developers. Bzzt... start over from the beginning!
I did eventually find the instructions by (ironically) asking Gemini Pro, which provided the convenient 27 step process for signing up to three different services in a chain before you can do anything. Oh, and if any of them trigger any kind of heuristic, again, you get a hand in face telling you firmly and not-so-politely to take a hike.
PS: Azure's whatever-it-is-called-today is just as bad if not worse. We have a corporate account and can't access GPT 5 because... I dunno. We just can't. Not worthy enough for access to Sam Altman's baby, apparently.
BoorishBears 2 hours ago [-]
As the other comments pointed out, that's not covering billing...
But also the (theoretical) production platform for Gemini is Vertex AI, not AI Studio.
And until pretty recently using that took figuring out service accounts, and none of Google's docs would demonstrate production usage.
Instead they'd use the gcloud CLI to authenticate, and you'd have to figure out how each SDK consumed a credentials file.
-
Now there's "express mode" for Vertex which uses an API Key, so things are better, but the complaints were well earned.
I did this same thing and this was my first result too. I am just not seeing how the author ended up where they did, unless knowing how to use Google search is not a core skill.
mediaman 2 hours ago [-]
Read the full post. Partway down you will see they agree with you that getting an API key is not hard.
Paying is hard. And it is confusing how to set it up: you have to create a Vertex billing account and go through a cumbersome process to then connect your AIStudio to it and bring over a "project" which then disconnects all the time and which you have to re-select to use Nano Banana Pro or Gemini 3. It's a very bad process.
It's easy to miss this because they are very generous with the free tier, but Gemini 3 is not free.
malfist 3 hours ago [-]
I did notice in their post instead of searching for answers, they asked Gemini how to do things, and when that didn't work, they asked Claude.
I often see coworkers offload their work of critical thinking to an AI to give them answers instead doing the grunt work nessecary to find their answers on their own.
dugidugout 3 hours ago [-]
This rhetoric worries me. If you insist on degrading others at least fix it to something like:
> [They seemingly] can't think on their own without an AI [moderating]
They _literally_ can think on their own, and they _literally_ did think up a handful of prompts.
A more constructive way to make what I assume to be your point would be highlighting why this shift is meaningful and leaving the appeal to ego for yourself.
ipaddr 53 minutes ago [-]
There is some truth in that statement.
Low energy afternoons you might be able to come up with a prompt but not the actual solution.
There are people offloading all thoughts into prompts instead of doing the research themselves and some have reached a point where they lost the ability to do something because of over AI use.
malfist 3 hours ago [-]
I agree with your assessment, I am in the wrong here. It's easy to be extra judgmental to anonymous figures on a blog you'll never meet. Thank you for reminding me to give people the benefit of doubt and not jump to worst case assumptions.
I've edited my post to be more charitable
plaidfuji 4 hours ago [-]
> The “Set up billing” link kicked me out of Google AI Studio and into Google Cloud Console, and my heart sank. Every time I’ve logged into Google Cloud Console or AWS, I’ve wasted hours upon hours reading outdated documentation, gazing in despair at graphs that make no sense, going around in circles from dashboard to dashboard, and feeling a strong desire to attain freedom from this mortal coil.
100% agree
avereveard 3 hours ago [-]
Is the outdated aws documentation in the room with us now?
neom 4 hours ago [-]
I complained about this on HN recently and Logan responded and asked me to email him with feedback on how I'd like the experience to work (I didn't, sorry Logan, been busy :)) - Logan, to his credit, is very active everywhere reading and soliciting feedback. I think they're going to be giving it a pretty big bump on ux/ui of AI studio next month. It's easy to see he's a super smart guy trying to build something complex within a massive machine - given how focused on the product he appears to be, I have high hopes.
I don't know, he announced on Bluesky that they are dropping a big vibe coding update to aistudio next year
1. cart out in front of the horse a bit on this one, lame hype building at best
2. Not at all what I want the team focusing on, they don't seem to have a clear mission
Generally Google PMs and leaders have not been impressive or in touch for many years, since about the time all the good ones cashed out and started their own companies
mvkel 37 minutes ago [-]
This has often been the case (Google dev rel soliciting feedback) and they very rarely take meaningful action. Like firing a bug report with Apple.
They'll get to it when it becomes strategically important to.
Why making it easier to pay them isn't always strategically important, I'm not sure.
niwtsol 4 hours ago [-]
The article lightly mentions it, but how AWS and Google Cloud Console are so absolute nonsensical in UX and ease of use is beyond comprehension.
polalavik 4 hours ago [-]
holy hell google cloud is so confusing i just ended up using (a much more expensive) digital ocean droplet instead for a little project. I guess they only really care about enterprise customers who can burn tons of money figuring it out, but it made me never want to use it again.
Same with google ads - super fuckin shit UI/UX, super confusing to understand what is going on.
companies like digital ocean, supabase, etc can make money (from people like me) because they just circumvent the bullshit or wrap the dogshit experience (aws) into a much better experience. bless supabase.
herpdyderp 3 hours ago [-]
DigitalOcean is such a dream to use. I also really appreciate all their guides for almost everything web server related.
sofixa 4 hours ago [-]
Baring them using specific marketing terms (so you have EC2 for what are basically virtual machines), for which both the docs and the portal itself provide helpful information, what do you mean? I find GCP's console and whole set up to be slightly better, but both it and AWS are fine.
Now Azure, or anything made by VMware, you just know they hate you.
koinedad 3 hours ago [-]
I have always found Google products incredibly confusing and difficult to use. I have had a very similar experience to this a number of times.
squirrellous 23 minutes ago [-]
As a former employee, the engineering culture at Google gives me old-school hacker vibes, so users are very much expected to “figure it out” and that’s somewhat accepted (and I say this with fond memories). It’s no surprise the company struggles with good UX.
Havoc 5 hours ago [-]
Yeah can't figure out WTH is going on in google's AI ecosystem either.
They absolutely deserve credit for their free tier API keys though. That's unheard of in big cloud - an actual you can't shoot yourself in the foot with a life ruining bill thing. Can't recall what part of their product maze I got it from but it seems to do what it says on tin
impure 3 hours ago [-]
The free tier is good but they've been cracking down on rate limits. Just recently they significantly dropped the max requests per day.
BoorishBears 4 hours ago [-]
You can sum it up as: Gemini from AI Studio and Gemini from Vertex AI Studio have independent rate limits.
-
And I guess to add some context, it's because Google seemingly realized that Google Cloud moves so glacially slow, and has so much baggage, that they could no longer compete with scrappier startups like OpenAI and Anthropic on developer mindshare.
So there's a separate product org that owns AI Studio, which tries to be more nimble, and probably 50x'd Gemini adoption by using API Keys instead of Service Accounts and JSON certs that take mapping out the 9th circle of hell to deploy in some environments. (although iirc Vertex now has those)
They definitely do ship faster than Google Cloud, but their offerings actually end up feeling like a product team with fewer resources than OpenAI or Anthropic (like shipping purple tailwind-slop UIs as real features), which is just nuts.
modeless 4 hours ago [-]
Seems like the real problem is something about his account or credit card tripped some fraud detectors and he got stuck in a part of the system designed to prevent credit card fraud rather than facilitate legitimate use. I can certainly imagine that Google gets a lot of chargebacks from people who had their credit card numbers stolen to mine bitcoin or whatever on Google Cloud.
nl 4 hours ago [-]
You literally cannot buy Antigravity with a non-personal Google account.
I read someone on here who is using Gemini via OpenRouter because it was the only way they could pay for it.
verdverm 4 hours ago [-]
I'm using it fine through both aistudio and vertex ai, direct API calls
It's not at all hard generally, the core of this issue is centered around gemini-cli which is a hot pile of trash. The inability to get keys or account credentials (like why even use an API key, Google is top notch in auto-auth/WIF)
Insanity to me how gemini-cli is so bad at the basics with so many great Google packages in open source that handle all this transparently. All I need to do is have my gcloud authd with the right account/project. I sarcastically assume his is because they vibe coded gemini-cli and it implemented everything from scratch, missing out on reusing those great packages
nl 1 hours ago [-]
> I'm using it
Do you mean Antigravity or Gemini?
If you mean Antigravity then.. how? Their docs say you can't do this.
If you mean Gemini then I personally haven't had issues but haven't tried to productionize a Gemini app. The OPs account seems to reflect other comments here.
inquirerGeneral 4 hours ago [-]
[dead]
2 hours ago [-]
happyopossum 3 hours ago [-]
The underlying issue here is that 3.0 is still in preview. Once it’s a GA model, you can just use your $20 consumer Ai pro sub and skip all the GCP stuff…
aerhardt 5 hours ago [-]
I had to warm up a Gemini API project worth a few thousand hours during weeks so that I could get to the tier that allowed me to carry out the workload.
How can you have any tokens if you haven’t finished your tokens?!
throwup238 4 hours ago [-]
> How can you have any tokens if you haven’t finished your tokens?!
Another rate limit in the wall.
rtaylorgarlock 4 hours ago [-]
Interesting perspective. I've mainly felt like i have 'American privilege' regarding the ease with which i open accounts of basically any sort on a whim, usually with little friction.
mrj 4 hours ago [-]
Oh no, this is just a Google thing. I've done the same verification bs for four different companies now, multiple times for each of them. I just keep an image of my license on my computer so I can upload it on demand. Google's payment verification is byzantine.
It'll trigger when you sign up.
It'll trigger if you create an Android developer account.
It'll trigger if you get a new phone.
It'll trigger if your card expires.
It'll trigger the month before your card expires. Why? Fuck you, that's why.
jpollock 3 hours ago [-]
There is a lot of fraud with UPI, specifically social engineering to obtain UPI OTP codes.
Since the card and the account haven't been previously associated, that's probably a risk model saying a human needs to verify the account before activation.
Indian cards also (I believe) have a mandatory 24 notice period prior to money being pulled - giving fraudsters a 24 hour starting gun to spend like crazy. That makes merchants that provide variable cost service on credit products twitchy.
Adding another layer on top of Google's own APIs adds latency, lowers reliability, and (AFAIK) doesn't allow batch mode - but if that's tolerable, it avoids the mess that is Google Service Account JSON and Cloud Billing.
binarymax 1 hours ago [-]
The same billing experience applies to the Google programmable search api. Easy to get a key, but a Byzantine maze to pay for more than the free version.
axi0m 4 hours ago [-]
Gosh, this story resonates so much with me... I had the exact same experience few days ago, desperately trying to get a small agent prototype working for a quick demo. I spent an good hour dealing with that pile of nonsense. Online payments and accounts management have been mastered for 20 years now, why do we still have to endure such things? It just kills me. The same goes with Azure (and all MS online-related services), if not even worse.
tigranbs 5 hours ago [-]
Yeah, then try adding more quotas to scale your usage; you will feel the pain!
But, to be fair, it is way easier than the AWS Bedrock or Microsoft Azure!
semi-extrinsic 4 hours ago [-]
I recently had the distinct displeasure of being the first to set up a service in Azure on a new tenant at Microsoft.
Of course I first had to faff about adding the company credit card, which took five tries and two days. Then I found I had to create the appropriate resource group, before I could set up a service. Fair enough, it might make sense later to have costs divided up like that. After I got the resource group, I then thought to start simple and spin up a single VM.
This gave me an error message saying that my request exceeded the quota. Which quota? The built-in copilot in Azure chewed on the raw error in its JSONness, and helpfully told me I could find the Azure quota page by searching for it in the Azure portal.
Once I entered the quota page, I was greeted with a message saying that I was now in the new quota experience in public preview mode. After many clicks I found the appropriate line for the desired VM SKU in the desired region, where it said I had used 0 of the quota of 30. So why didn't it work? I tried to request an increased quota, just in case. That process spent five minutes on "please wait", then failed with a generic error message.
At that point I started googling around, and eventually in some forum thread I found the missing piece: my resource group did not yet have a subscription. After more faffing about, I got a subscription associated with my resource group. What is a subscription, you ask, and what is the relation between a tenant, a subscription and a resource group? I haven't the foggiest, but I've clicked enough buttons to make the errors go away. Por ahora.
cj 4 hours ago [-]
Isn't OpenAI equally annoying?
I remember multiple waiting periods, and multiple requirements to cross spend thresholds to increase in tiers. I remember at one point spamming the OpenAI API with garbage just to consume credits in order to get to the next tier to increase rate limits.
More recently (couple months ago) I tried using a 3rd party client for ChatGPT which needed a OpenAI API key. I gave up after 20 mins.
ipaddr 32 minutes ago [-]
The OpenAI api key generation was simple and using it no problem. No different from stripe.
The limits are annoying.
arielcostas 4 hours ago [-]
I'd say Bedrock is the easiest since you just log into your AWS account, get an AWS credential in the same way you would for any other service (if you're on EC2 it's even easier) and call the endpoints from the SDK. Azure though...
intothemild 2 hours ago [-]
I like using Gemini for general stuff,
I have that Gemini AI plan thing, and it's great. But I absolutely will not plug my credit card into Google cloud services, no way.. I know I can put guardrails up, but I just am terrified that I'll get a gigantic bill that I cannot afford.
Nope sorry no way. I want a simple $X per month sub.
Claude gives me that. Which is why Claude wins.
ra 2 hours ago [-]
That's fine but you're missing out, and you're most likely paying waay more than you need to.
ipaddr 46 minutes ago [-]
Missing out on nothing
eezing 39 minutes ago [-]
It looks AI won’t replace software engineers after all.
postsantum 4 hours ago [-]
Congrats, you have sampled the life of android developer. I've been avoiding touching Gemini exactly for the reason "Your account is in good standing. For now". When it's not, enjoy your ban for life
horaceradish 3 hours ago [-]
[dead]
senthilnayagam 2 hours ago [-]
I am from India, Have had similar experience. After billing they mentioned models which ai was looking were not launched in india ( Veo3 and nano banana )
I use replicate, fal for all api and for LLM openrouter
heystefan 3 hours ago [-]
I had the same experience. Plus you never know what's the best way to use eg. Nano Banana -- it works better in AI Studio versus their regular Gemini chat.
notepad0x90 5 hours ago [-]
I think they're just too focused on enterprise billing. Someone at google doesn't get that individuals trying it out is how they go their work and recommend this stuff.
Googlers tend to exist in an isolated bubble. In the corporate world, Azure is the default and they have Azure OpenAI. Why would someone bother with Gemini? Unless the devs at companies have a good experience with it of course.
Googlers are awesome/mean well, if only enough of them lurked here :)
marcuskaz 4 hours ago [-]
Developer experience matters. This is what Vercel figured out and why their admin screens are sooooooo much better than anything AWS or Google creates.
9rx 4 hours ago [-]
"Developer experience matters" and "Vercel" being the example is something I never thought I would see together.
I actually do agree that Vercel's admin screens are quite good compared to the other usual suspects. But I don't consider that to be on the development side of things. It's done decently well because it is geared towards the business folks who are paying the bills.
Developers writing code on top of the development solutions produced by Vercel have been completely forsaken.
marcuskaz 3 hours ago [-]
:thinking-face:
- How builds and deploys are configured
- The simple aspect of connecting a GitHub repo and you get auto deploys
- Auto creating branch environments that make testing as easy as a new link
- Just configuring users and permissions and not seeing IAM anywhere is a huge win
My billing admins don't do any of this stuff.
9rx 2 hours ago [-]
None of those are development tasks. IT tasks, I'd buy, but anyone deeply entrenched in IT are more likely going to want more powerful tools (even if harder to use). Vercel is geared towards the small groups where there are some developers on staff, but the budget makers are playing double-duty in IT roles.
4 hours ago [-]
h02 5 hours ago [-]
Just wait until you find out that Tier 1 only gives you up to 250 requests a day, and if you want more than that you'll have had to have spent over $250 in Google Cloud spend, and your first payment has to be more than 30 days ago. I was going to build my side project using Gemini 3 Pro, but gave up after that.
ryuuseijin 3 hours ago [-]
Just a note that you can use opencode with their API gateway (they call it "zen") to get access to all the most popular models using a single account, including gemini. (Although this wouldn't have helped the author, since they wanted to try the Gemini CLI specifically).
consumer451 4 hours ago [-]
Wasn't this type of Google thing clearly called out in the Karpathy Software 3.0 talk?
It's interesting to me this UX problem is not readily solved.
What is the sticking point in a big org? I don't have a point of reference.
chillfox 59 minutes ago [-]
It will be politics, it's always politics. Large orgs works a lot like the feudal system.
robertheadley 3 hours ago [-]
I had issues too, I wanted to use my free Google API credits with Roo Code, but I could never get it to work.
I eventually got Gemini Cli and now Antigravity to work.
huevosabio 16 minutes ago [-]
This nonsense alone justifies the existence of OpenRouter.
nikanj 5 hours ago [-]
This reminds me of the ”I just want to serve 5 terabytes” thing
Google does not want your money, they don’t know how to count so low
I use OpenAI and Anthropic APIs every day for work. I have never used google Gemini precisely because there seems to be a whole different set of friction involved in getting an account. First I don’t want to tie anything to my google account, especially any form of payment (no idea if I actually need to do this). Second I don’t want AI studio or whatever, I just want a similar api to the others I can hit.
I admit I’m completely ignorant about what’s really involved, I have never tried and am just going on vague things I’ve heard but stories like this definitely reinforce my perception. I even have a mistral account, grok, etc, but google feels like a whole other level of complication.
0cf8612b2e1e 3 hours ago [-]
I feel you on not wanting to tie anything additional to your Google account. Will I somehow do something “naughty” (say spam an emoji during a livestream) that gets me permanently banned for life from all services?
Google really needs to evaluate separating service bans. I cannot be the only one who would rather go to a competitor than risk angering the black box and destroying my digital life.
2 hours ago [-]
athrowaway3z 5 hours ago [-]
I didn't even need to read the article to upvote, and doing so just confirmed my assumption that somebody finally wrote down their frustration with Gemini.
My fucking god, how has Google not flagged the failure of onboarding devs like Claude / Codex?
3 days ago I was literally thinking, I want to throw 20$ to try out Gemini alongside my Claude and Codex subs.
It took me a few minutes to realize its just not worth my time to figure out how.
mox1 4 hours ago [-]
I wanted to try Googles Image generation models, to compliment my Claude API sub.
Holy Crap, I got about 45 minutes into setting up billing and just gave up and un-did everything.
Hint: If you want to put a spending Limit on your google cloud account, its not trivial.
I will say that Stability AI is similar to Claude, they will just let you buy credits and hit an API.
sirfz 4 hours ago [-]
I find GCP frustrating (coming from AWS) but luckily asking Gemini how to do things makes it much easier.
gxs 4 hours ago [-]
This is a longstanding issue we’ve had, not just with Gemini
Even with something as simple as google workspace - permissioning service accounts and authentication are a pain in the ass
The docs suck and of course there’s no one to help
arielcostas 4 hours ago [-]
Wait until you see Azure. Apparently you need to create either an "Azure OpenAI" or a "Microsoft Foundry", where AFAIK (got an email last week) Foundry now includes everything AI including "Azure OpenAI", the former "Cognitive Services" (for speech, computer vision and other stuff) and inference on non-OpenAI models. But wait, because once you create that, you are told to go to another portal (ai.azure.com) where you get an "old" foundry experience and anew one that can't be enabled for every project. Oh, wait, did I mention there apparently used to be a "Foundry" and a "Foundry Project"? Oh, and all those apparently work with a single API key, unless (I guess) you set up authentication with the Azure SDK, which makes you go back to Azure Portal (or maybe Entra ID?).
All of that while trying to explain to your non-technical boss how he can browse the voices available at "the Azure thingy" to pick his favourites to then pick and use in the project due relatively soon. Since, of course, you told him the original Cognitive Speech Services (or Speech Services, or Cognitive Services-Speech, or whatever they decided to call it on that specific page) semi-public URL where he could browse the gallery was "speech.microsoft.com" which is now semi-dead with awful loading times that seem some server issue and has been happenning for a few months now. Or tell them to go to the "new foundry" where he might not be able to find the resource or might not have stuff in the regions you were using up until then, or whatever crap this 3.56 trillion-dollar company decides to throw at you to prevent you from using their services.
And all of this is the exploration phase, where you just use the GUIs and copy things around until they work. Then you need to figure out what you did (and more importantly, where) to be able to write some Terraform/OpenTofu or Bicep or similars to try and keep the environment replicable to avoid the excruciating pain of repeating every single step you followed to get it on a working state.
At the very least, Google was nice enough to launch Vertex AI inside GCP for enterprises that have figured that out, and then Google AI Studio as an almost completely separate thing that only is bound to Google Cloud for billing purposes, similar to how Firebase is integrated too.
g-unit33 2 hours ago [-]
Wait until the 429 error message
5 hours ago [-]
notatoad 3 hours ago [-]
i saw somebody saying this last week, so i tried it just out of curiosity, and it isn't that hard? i went to google ai studio, i clicked on the button to get an api key, and i got an api key. what's so hard about this?
it seems like people are just clowning on google for this for pageviews, and people believe it because based on google's reputation it seems like it is probably true. but it really isn't.
BeetleB 2 hours ago [-]
I know it's against HN rules, but ...
You really should read the submission before leaving a comment. Or even copy/paste into an LLM to summarize.
Rendered at 02:42:56 GMT+0000 (Coordinated Universal Time) with Vercel.
I assume it has something to do with the underlying constraint grammar/token masks becoming too long/taking too long to compute. But as end users we have no way of figuring out what the actual limits are.
OpenAI has more generous limits on the schemas and clearer docs. https://platform.openai.com/docs/guides/structured-outputs#s....
You guys closed this issue for no reason: https://github.com/googleapis/python-genai/issues/660
Other than that, good work! I love how fast the Gemini models are. The current API is significantly less of a shitshow compared to last year with property ordering etc.
You can walk into a McDonalds without being able to read, write, or speak English, and the order touchscreen UI is so good (er, "good") that you can successfully order a hamburger in about 60 seconds. Why can't Google (of all companies) figure this out?
It makes sense for IBM, seems like google is just reaching that stage?
I have a friend that says Google's decline came when they bought DoubleClick in 2008 and suffered a reverse-takeover: their customers shifted from being Internet users and became other, matchingly-sized corporations.
I know part of it is that sales wants to be able to price discriminate and wants to be able to use their sales skills on a customer, but I am never going to sign up for anything that makes me talk to someone before I can buy.
1. Never make it hard for people to give you money.
Edit: On second thought, there is a perverse incentive at work (and probably one of the "lowest friction" ways to get money), which is issuing government enforced fines.
If a platform is designed in a way that users can sign up and go, it can work well.
If an application is complicated or it’s a tool that the whole business runs on, often times the company will discover their customers have more success with training and a point of contact/account manager to help with onboarding.
Someone who works in finance or conpliances might want a demo, or views those things as signals the product is for serious use cases.
About the only time you’ll be asked to evaluate such a product as an IC is when someone wants an opinion about API support or something equivalent. And if you refuse to do it, the decision-makers will just find the next guy down the hall who won’t be so cranky.
You say that as if it isn’t the entire reason why these interactions should be avoided at all costs. Dynamic pricing should be a crime.
Does segmentation also count as dynamic pricing?
--
https://blog.codinghorror.com/oh-you-wanted-awesome-edition/If everybody can see the prices that are being quoted/paid by other people, that exerts a strong moderating force against abuse.
It won't help you if there's a monopoly, but I consider that a separate problem needing separate solutions.
All of their products, however realistically commoditized, will require a drawn out engagement with a rep who knows how much money you’ve received recently and even has an outline what research you plan to do over the next few years since even the detailed applications often get published alongside funding allocations.
The exact same piece of equipment, consumables required to use it, and service agreements might be anywhere from X to 10X depending on what they (as a result of asymmetrically available knowledge) know you need and how much you could theoretically spend.
Getting just the university of California should be enough critical mass.
[1]: https://adstransparency.google.com/advertiser/AR129387695568...
When Google has a bad/empty profile of you, advertisers don't bid on you, so it goes to the bottom feeders. Average (typically tech illiterate) people wandering through the internet mostly get ads for Tide, Chevy, and [big brand], because they pay Google much more for those well profiled users. These scam advertisers really don't pay much, but are willing to be shown to mostly anyone. They are a bit like the advertiser of last resort.
All of that is to say, if you are getting malware/scam ads from Google, it's probably because (ironically) you know what you are doing.
One of my co-workers left with an active account and active card but no passwords noted. The company gave up and just had to cancel + create a new account for the next adwords specialist.
I google `gemini API key` and the first result* is this docs page: https://ai.google.dev/gemini-api/docs/api-key
That docs page has a link in the first primary section on the page. Sure, it could be a huge CTA, but this is a docs page, so it's kinda nice that it's not gone through a marketing make over.
* besides sponsored result for AI Studio
(Maybe I misunderstood and all the complaints are about billing. I don't remember having issues when I added my card to GCP in the past, but maybe I did)
If you bring it up to Logan he'll just brush it off — I honestly don't know if they test these UX flows with their own personal accounts, or if something is buggy with my account.
But somehow personally even though I'm a paying Google One subscriber and have a GCP billing account with a credit card, I get confusing errors when trying to use the Gemini API
Excuse me? If you mean AI Studio, are you talking about the product where you can’t even switch which logged in account you’re using without agreeing to its terms under whatever random account it selected, where the ability to turn off training on your data does not obviously exist, and where it’s extremely unclear how an organization is supposed to pay for it?
Hint: you can often avoid some of this mess by adding the authuser=user@domain to the URL.
I've been using the AI Studio with my personal Workspace account. I can generate an API key. That worked for a while, but now Gemini CLI won't accept it. Why? No clue. It just says that I'm "not allowed" to use Gemini Pro 3 with the CLI tool. No reason given, no recourse, just a hand in your face flatly rejecting access to something I am paying for and can use elsewhere.
Simultaneously, I'm trying to convince my company to pay for a corporate account of some sort so that I can use API keys with custom tools and run up a bill of potentially thousands of dollars that we can charge back to the customer.
My manager tried to follow the instructions and... followed the wrong ones. They all look the same. They all talk about "Gemini" and "Enterprise". He ended up signing up for Google's equivalent of Copilot for business use, not something that provides API keys to developers. Bzzt... start over from the beginning!
I did eventually find the instructions by (ironically) asking Gemini Pro, which provided the convenient 27 step process for signing up to three different services in a chain before you can do anything. Oh, and if any of them trigger any kind of heuristic, again, you get a hand in face telling you firmly and not-so-politely to take a hike.
PS: Azure's whatever-it-is-called-today is just as bad if not worse. We have a corporate account and can't access GPT 5 because... I dunno. We just can't. Not worthy enough for access to Sam Altman's baby, apparently.
But also the (theoretical) production platform for Gemini is Vertex AI, not AI Studio.
And until pretty recently using that took figuring out service accounts, and none of Google's docs would demonstrate production usage.
Instead they'd use the gcloud CLI to authenticate, and you'd have to figure out how each SDK consumed a credentials file.
-
Now there's "express mode" for Vertex which uses an API Key, so things are better, but the complaints were well earned.
At one point there were even features (like using a model you finetuned) that didn't work without gcloud depending on if you used Vertex or AI Studio: https://discuss.ai.google.dev/t/how-can-i-use-fine-tuned-mod...
Paying is hard. And it is confusing how to set it up: you have to create a Vertex billing account and go through a cumbersome process to then connect your AIStudio to it and bring over a "project" which then disconnects all the time and which you have to re-select to use Nano Banana Pro or Gemini 3. It's a very bad process.
It's easy to miss this because they are very generous with the free tier, but Gemini 3 is not free.
I often see coworkers offload their work of critical thinking to an AI to give them answers instead doing the grunt work nessecary to find their answers on their own.
> [They seemingly] can't think on their own without an AI [moderating]
They _literally_ can think on their own, and they _literally_ did think up a handful of prompts.
A more constructive way to make what I assume to be your point would be highlighting why this shift is meaningful and leaving the appeal to ego for yourself.
Low energy afternoons you might be able to come up with a prompt but not the actual solution.
There are people offloading all thoughts into prompts instead of doing the research themselves and some have reached a point where they lost the ability to do something because of over AI use.
I've edited my post to be more charitable
100% agree
https://x.com/OfficialLoganK
1. cart out in front of the horse a bit on this one, lame hype building at best
2. Not at all what I want the team focusing on, they don't seem to have a clear mission
Generally Google PMs and leaders have not been impressive or in touch for many years, since about the time all the good ones cashed out and started their own companies
They'll get to it when it becomes strategically important to.
Why making it easier to pay them isn't always strategically important, I'm not sure.
Same with google ads - super fuckin shit UI/UX, super confusing to understand what is going on.
companies like digital ocean, supabase, etc can make money (from people like me) because they just circumvent the bullshit or wrap the dogshit experience (aws) into a much better experience. bless supabase.
Now Azure, or anything made by VMware, you just know they hate you.
They absolutely deserve credit for their free tier API keys though. That's unheard of in big cloud - an actual you can't shoot yourself in the foot with a life ruining bill thing. Can't recall what part of their product maze I got it from but it seems to do what it says on tin
-
And I guess to add some context, it's because Google seemingly realized that Google Cloud moves so glacially slow, and has so much baggage, that they could no longer compete with scrappier startups like OpenAI and Anthropic on developer mindshare.
So there's a separate product org that owns AI Studio, which tries to be more nimble, and probably 50x'd Gemini adoption by using API Keys instead of Service Accounts and JSON certs that take mapping out the 9th circle of hell to deploy in some environments. (although iirc Vertex now has those)
They definitely do ship faster than Google Cloud, but their offerings actually end up feeling like a product team with fewer resources than OpenAI or Anthropic (like shipping purple tailwind-slop UIs as real features), which is just nuts.
I read someone on here who is using Gemini via OpenRouter because it was the only way they could pay for it.
It's not at all hard generally, the core of this issue is centered around gemini-cli which is a hot pile of trash. The inability to get keys or account credentials (like why even use an API key, Google is top notch in auto-auth/WIF)
Insanity to me how gemini-cli is so bad at the basics with so many great Google packages in open source that handle all this transparently. All I need to do is have my gcloud authd with the right account/project. I sarcastically assume his is because they vibe coded gemini-cli and it implemented everything from scratch, missing out on reusing those great packages
Do you mean Antigravity or Gemini?
If you mean Antigravity then.. how? Their docs say you can't do this.
If you mean Gemini then I personally haven't had issues but haven't tried to productionize a Gemini app. The OPs account seems to reflect other comments here.
How can you have any tokens if you haven’t finished your tokens?!
Another rate limit in the wall.
It'll trigger when you sign up.
It'll trigger if you create an Android developer account.
It'll trigger if you get a new phone.
It'll trigger if your card expires.
It'll trigger the month before your card expires. Why? Fuck you, that's why.
Since the card and the account haven't been previously associated, that's probably a risk model saying a human needs to verify the account before activation.
Indian cards also (I believe) have a mandatory 24 notice period prior to money being pulled - giving fraudsters a 24 hour starting gun to spend like crazy. That makes merchants that provide variable cost service on credit products twitchy.
https://support.stripe.com/questions/background-on-indian-go...
Adding another layer on top of Google's own APIs adds latency, lowers reliability, and (AFAIK) doesn't allow batch mode - but if that's tolerable, it avoids the mess that is Google Service Account JSON and Cloud Billing.
Of course I first had to faff about adding the company credit card, which took five tries and two days. Then I found I had to create the appropriate resource group, before I could set up a service. Fair enough, it might make sense later to have costs divided up like that. After I got the resource group, I then thought to start simple and spin up a single VM.
This gave me an error message saying that my request exceeded the quota. Which quota? The built-in copilot in Azure chewed on the raw error in its JSONness, and helpfully told me I could find the Azure quota page by searching for it in the Azure portal.
Once I entered the quota page, I was greeted with a message saying that I was now in the new quota experience in public preview mode. After many clicks I found the appropriate line for the desired VM SKU in the desired region, where it said I had used 0 of the quota of 30. So why didn't it work? I tried to request an increased quota, just in case. That process spent five minutes on "please wait", then failed with a generic error message.
At that point I started googling around, and eventually in some forum thread I found the missing piece: my resource group did not yet have a subscription. After more faffing about, I got a subscription associated with my resource group. What is a subscription, you ask, and what is the relation between a tenant, a subscription and a resource group? I haven't the foggiest, but I've clicked enough buttons to make the errors go away. Por ahora.
I remember multiple waiting periods, and multiple requirements to cross spend thresholds to increase in tiers. I remember at one point spamming the OpenAI API with garbage just to consume credits in order to get to the next tier to increase rate limits.
More recently (couple months ago) I tried using a 3rd party client for ChatGPT which needed a OpenAI API key. I gave up after 20 mins.
The limits are annoying.
I have that Gemini AI plan thing, and it's great. But I absolutely will not plug my credit card into Google cloud services, no way.. I know I can put guardrails up, but I just am terrified that I'll get a gigantic bill that I cannot afford.
Nope sorry no way. I want a simple $X per month sub.
Claude gives me that. Which is why Claude wins.
I use replicate, fal for all api and for LLM openrouter
Googlers tend to exist in an isolated bubble. In the corporate world, Azure is the default and they have Azure OpenAI. Why would someone bother with Gemini? Unless the devs at companies have a good experience with it of course.
Googlers are awesome/mean well, if only enough of them lurked here :)
I actually do agree that Vercel's admin screens are quite good compared to the other usual suspects. But I don't consider that to be on the development side of things. It's done decently well because it is geared towards the business folks who are paying the bills.
Developers writing code on top of the development solutions produced by Vercel have been completely forsaken.
- How builds and deploys are configured
- The simple aspect of connecting a GitHub repo and you get auto deploys
- Auto creating branch environments that make testing as easy as a new link
- Just configuring users and permissions and not seeing IAM anywhere is a huge win
My billing admins don't do any of this stuff.
It's interesting to me this UX problem is not readily solved.
What is the sticking point in a big org? I don't have a point of reference.
Google does not want your money, they don’t know how to count so low
https://youtu.be/3t6L-FlfeaI (2010)
To be fair, a lot of this changed after that video became a meme.. but I'd bet that the broccoli man template is still trending on memegen
I admit I’m completely ignorant about what’s really involved, I have never tried and am just going on vague things I’ve heard but stories like this definitely reinforce my perception. I even have a mistral account, grok, etc, but google feels like a whole other level of complication.
Google really needs to evaluate separating service bans. I cannot be the only one who would rather go to a competitor than risk angering the black box and destroying my digital life.
My fucking god, how has Google not flagged the failure of onboarding devs like Claude / Codex?
3 days ago I was literally thinking, I want to throw 20$ to try out Gemini alongside my Claude and Codex subs.
It took me a few minutes to realize its just not worth my time to figure out how.
Holy Crap, I got about 45 minutes into setting up billing and just gave up and un-did everything.
Hint: If you want to put a spending Limit on your google cloud account, its not trivial.
I will say that Stability AI is similar to Claude, they will just let you buy credits and hit an API.
Even with something as simple as google workspace - permissioning service accounts and authentication are a pain in the ass
The docs suck and of course there’s no one to help
All of that while trying to explain to your non-technical boss how he can browse the voices available at "the Azure thingy" to pick his favourites to then pick and use in the project due relatively soon. Since, of course, you told him the original Cognitive Speech Services (or Speech Services, or Cognitive Services-Speech, or whatever they decided to call it on that specific page) semi-public URL where he could browse the gallery was "speech.microsoft.com" which is now semi-dead with awful loading times that seem some server issue and has been happenning for a few months now. Or tell them to go to the "new foundry" where he might not be able to find the resource or might not have stuff in the regions you were using up until then, or whatever crap this 3.56 trillion-dollar company decides to throw at you to prevent you from using their services.
And all of this is the exploration phase, where you just use the GUIs and copy things around until they work. Then you need to figure out what you did (and more importantly, where) to be able to write some Terraform/OpenTofu or Bicep or similars to try and keep the environment replicable to avoid the excruciating pain of repeating every single step you followed to get it on a working state.
At the very least, Google was nice enough to launch Vertex AI inside GCP for enterprises that have figured that out, and then Google AI Studio as an almost completely separate thing that only is bound to Google Cloud for billing purposes, similar to how Firebase is integrated too.
it seems like people are just clowning on google for this for pageviews, and people believe it because based on google's reputation it seems like it is probably true. but it really isn't.
You really should read the submission before leaving a comment. Or even copy/paste into an LLM to summarize.