For what it’s worth, that was pretty obvious from the get go it wasn’t a realistic long term deal. I’ve been building all the libraries I hoped existed over the past 1-2y to have something neat to work with whenever the free compute era ends. I feel that’s the approach that makes sense. Take the free tokens, build everything you would want to exist if you don’t have access to the service anymore. If it goes away you’re back to enjoying writing code by hand but with all the building blocks you dreamt of. If it never goes away, nothing wasted, you still have cool libs

So, antigravity will definitely quickly eat up your pro quota. You can run out of it in an hour (at least on the $20/mo plan) and then you'll be waiting five days for it to refresh.

However, I've found that the flash quota is much more generous. I have been building a trio drive FOC system for the STM32G474 and basically prompting my way through the process. I have yet to be able to run completely out of flash quota in a given five hour time window. It is definitely completing the work a lot faster than I could do myself -- mainly due to its patience with trying different things to get to the bottom of problems. It's not perfect but it's pretty good. You do often have to pop back in and clean up debris left from debugging or attempts that went nowhere, or prompt the AI to do so, but that's a lot easier than figuring things out in the first place as long as you keep up with it.

I say this as someone who was really skeptical of AI coding until fairly recently. A friend gave me a tutorial last weekend, basically pointing out that you need to instruct the AI to test everything. Getting hardware-in-loop unit tests up and running was a big turning point for productivity on this project. I also self-wired a bunch of the peripherals on my dev board so that the unit tests could pretend to be connected to real external devices.

I think it helps a lot that I've been programming for the last twenty years, so I can sometimes jump in when it looks like the AI is spinning its wheels. But anyway, that's my experience. I'm just using flash and plan mode for everything and not running out of the $20/mo quota, probably getting things done 3x as fast as I could if I were writing everything myself.

IMO we are currently in the ENIAC era of LLMs. Perhaps there will be a brief moment where things get worse, but long term the cost of these things will go way down.

Ultimately we'll find more efficient techniques and hardware and AI companies will end up owning Nuclear Power Stations and continue providing models capable of 10x of what they are now.

Valuation have already reached point where these companies can run their nuclear power station, fund developement of new hardware and techniques and boost capabilities of their models by 10x

I still remember those $3 uber rides.

Fellow annoyed Google AI Pro subscriber here!

Can confirm, I initially enjoyed the 5-hour limits on Gemini CLI and Antigravity so much that I paid for a full year, thinking it was a great decision

In the following months, they significantly cut the 5-hour limits (not sure if it even exists anymore), introduced the unrealistically bad weekly limit that I can fully consume in 1-2 hour, introduced the monthly AI credits system, and added ads to upgrade to Ultra everywhere

At the very least the Gemini mobile app / web app is still kinda useful for project planning and day-to-day use I guess. They also bumped the storage from 2TB to 5TB, but I don't even use that

Lights on = Ads in your output. EOY latest; they can't keep kicking the massive costs down the road.

> We may very well look back on the last couple years as the golden era of subsidized GenAI compute.

Looks like enshittification on steroids, honestly.