Comment by jjcm

Disclaimer that I haven't tried this since January, so things may have changed in the last 7mo, but this was my experience at that time: https://x.com/pwnies/status/2010523020629274723

At a high level though, as a rule of thumb Google assumes that they're serving companies at Google scale first, and at a human scale second. For other companies it's the opposite. Generally what that means is the first experience you get with a Google product will route you through 8 different dashboards to set up ACLs before you've hired your 2nd employee.

Personally I went to https://console.cloud.google.com since I already had some GCP projects. Then I searched for Gemini API Key. It brought me to https://console.cloud.google.com/agent-platform/studio/setti.... Then, there was a banner saying "Enable APIs to access full platform capabilities." Then, I did that, which took quite a while (minutes). Finally, I was able to see the way to create an API key.

The fact that there's two ways to get keys is also very confusing.

You need to create a Google Cloud project to create an api key and when you try to create one you very often get error messages like:

“Failed to create project, The request is suspicious. Please try again” or “ You do not have permission to create a key in this project”. You can then navigate multiple screens in GCP to make it work but it’s a hassle compared to any other provider (OAI/Ant/OpenRouter or any of the Chinese labs).

We have probably 10+ years old account with Google cloud etc. We recently had a production deployment, I went over to AI studio to get new keys and it kept failing saying "Failed to generate API key, The request is suspicious. Please try again" - It was through my standard browser, same geo-ip. And it just worked after 2 days.

> Can you help me understand how it is hard to get an API key from Google?

Using Google products in general is an effing nightmare as soon as you have to give them money.

The one thing you want in a business is to remove friction when people want to give you money, a concept Google has never been able to understand.

Maybe things have changed but it was a big mess trying to getting an API key from Google as an individual a few years ago. Way too much conflicting documentation.

Eventually I gave up and run a few hundred million tokens (edit a few billion) through openrouter.ai using Gemini Flash 1.5 to Flash 2.5

Every since price increases on Flash 3.0 I've stopped using Gemini, too expensive for basic classification, sentiment detection, ocr etc.

As other posters said Google assumes you are some bigcorp trying to use their products. The Vertex versus AI studio confusion did not help.

I tried to use Gemini for one of my projects a month ago. Immediately after signing up and paying for credits, I got an email saying, “Action required: your billing account {redacted} is past due or has invalid payment information.” I have no idea why it says this. My credit card on file works. My balance updated with a new amount from that card. 11 days later, my account was terminated. I still don’t understand what went wrong or how to fix it.

I do pay for OpenAI, Anthropic, and ElevenLabs keys.

I have a Google account that I had tied to a domain I originally purchased from google (when they had the .dev offerings), and now that it's been purchased by square space my ability to use AI through that account is in a bizarre state. Even my free gmail account has more AI offerings, and I'm not allowed to pay for improved AI offerings on my custom domain & google workspace account.

I know this is probably a pretty small edge case, but it is a bit frustrating. Any other provider lets you sign up with an email and give them a payment processor/card, but because google wants me to only use their unified workspace for signing up, I'm completely locked out now.

it worked for me okay when I needed it for myself in my personal account. But when I tried setup this for a company I spent almost a day solving lot of small puzzles in GCE like how to tell CEO that he have to connect billing account created for other purposes (and he not even remember at time that it exist) to new project and all other things that others talking about.

As someone who has been running Gemini models in production for a year, recently (last 2 months), I have been actively moving away from it.

The primary reason for me has been that Google autonomously decides to downgrade usage tiers and then upgrade them again - and does this incorrectly.

Over the last week itself, in the span of two days, our account for first downgraded and then upgraded. This is despite matching the criteria to remain at the tier we operate at throughout.

Google Support (when you finally get to a human) has accepted that these are potentially bugs, but the first time it happened, we were rate limited so severely for ~4 hours that I find it really difficult to continue trusting Google.

Simple question - can i use Gemini 3.7 flash with a subscription in my own harness and not in agy client ? You're from Google so the question.

By the way - I love Gemini's personality . it's phenomenal to work with

The reason i want my own harness - is the custom tools that i provide vs the low tier tools that come with the custom harnesses.

You'll would really benefit , if we could use Gemini in our own harness and not be forced to use it via agy . i've tried using gemini to circumvent - but not been successful.

If you see this - please reply here

On top of being the hardest website to navigate, Google console a) doesn't have real time billing (!) b) doesn't allow you to set a budget limit.

Sorry but it's not worth waking up with a 100k$ bill, fix your platform first.

GCP/vertex is a maze

Same experience. It took minutes to find the API key from the console.

Now following up with Google support team without luck to find the logs. Prompts send to the model and the responses including the thinking was available in the ai studio. But it’s unclear where to find the same in console.

To make matters worse there is vertex api and rebranded to Gemini something and making it very confusing.

i use gemini api in the third-party platform like openrouter and evolink.ai , easier to get key and manage my bill

That is easy but I’ve also found myself in account setup dashboards that were obviously geared toward enterprise trying to set up access to tinker with something AI related. It might have been TTS but it’s been a little while and I can’t quite remember.

The problem is that this doesn’t work for enterprise. The rate limits of that is super low. Then you need to migrate to Vertex and that is just a pain. Who ever thought of using JSON instead of an api key…

I even know the link existed but forgot what it was specifically and couldn’t remember what AI* property it was offered under and it took me a long time to figure it out.

I haven’t tried in about a year, but I could never do any meaningful work outside 1P Google apps (antigravity) due to such fast throttling.

I just asked Gemini how to do it and that's exactly what it sent me. Was up and running in a few minutes.

[dead]

Yep, I use Gemini for this too and it’s great - very fast and high quality.

I’d be very willing to try it out as an API, but it’s far too complicated to set up payment, and I don’t want to risk taking a wrong step and being locked out of other Google services. So Anthropic and Mistral get my money instead.

Is it? I think waferscale might actually be cheaper per-token, it's just so many more tokens, and of course right now it's not a full buildout so the availability is limited as well. I'd imagine they'll be migrating to whichever inference method is least expensive, and I expect asics to be the ultimate answer.

Moving from either frontier intelligence or frontier latency to a single model that does both at the same time is potentially a game changer in certain industries. I can easily see e.g. hedge funds dropping tons of money on this, because it means they can now do the same thing as their competitors, but much faster. That's basically a license to print money.

5.6 Luna costs far less and benchmarks far better, have you compared for this task?

Probably cheaper to run a Mac Mini with VisionKit (private APIs if you need bounding rects).

> You reach for it every time you do a Google search

[my self-important Kagi shtick awakens, pokes at it's restraints]

I already use GCP and Google for work, and getting an API key was so annoying that even I couldn't be bothered after a while of looking around.

Maybe things there have improved some, but when I was looking it was a huge runaround.

I like this concept of "vocational awe", and I do think it's why, for example, there is so much sexual and financial abuse in the movie, music and game industry: those people are willing to be treated poorly so they can work their craft - do the thing they love. Basically they are trading away good working conditions in exchange for actually doing work you like or find meaningful, the opposite of someone who does something they hate or find meaningless but pays a high salary and where you are treated really well.

I don't know how well this actually translates to AI, though. We all understand that AI art has a pretty low quality, but sometimes low quality is enough. If I want an image for my D&D character that only I and my DM will likely ever see, I am fine with a 7 cent AI-generated image, but I'm not willing to pay 150$ for an artist to do it - not that I don't value their time, I just don't value the image that much. Before AI this was the same, I'd just have used an image from Pinterest and thought "Well, this isn't exactly a good match, but I can't find anything better".

But I assume real illustrators do things like illustrations for children's books? I'd like to believe that those are still done by actual people, not AI.

i do feel like opus here is the most natural. theres something off about 3.7 and grok while its an improvement feels flat and not complete

i do wonder why gpt sol was not compared here but honestly it's not really known to be the best at UI

a fable 5 comparison would've been also interesting and likely the best.

I don't think I'd say rarely. Companies rarely allocate the design resources to produce that, but the companies that do are typically much larger, so the actual number of individual developers that get rich mocks is probably closer to 40-50%.

Depends who you work with, what's the intention, budget, etc. I'd agree this is a really good one.

I'm used to incremental Figma wireframe -> final product and working together with a designer.

Unfortunately, as soon as they can’t find work, everybody interested in the more easily automated dev work will suddenly become very interested in up-skilling into all other kinds of dev work. So then you have an excess supply, which means little job security and littler salaries. Developers were in the cool kid club in SV because it was more painful to fill dev roles than to treat developers with kid gloves. Without high labor demand, there is no leverage for developers. Increasingly, management has the leverage. Oh well.

This is pretty interesting take on design and your trick to make the surface maps is pretty slick!

I did have one question about the tool, is it possible to set how many variations you want per step? I would rather be able to guide it manually at some steps where maybe I know pretty well what I want or just need minor tweaks, and then let it loose on others when really trying to experiment with an idea.

did you build your own diffusion model?

why Grok not through `Grok Build` ?

[dead]