Google Is Cutting the Cost of Capable AI Agents

Three weeks was enough.

Google released Gemini 3.7 Flash just three weeks after 3.6 Flash, with better coding and agent performance at an introductory price of $0.75 per million input tokens.

The benchmark gains are substantial: DeepSWE jumped from 49.0% to 65.3%, while AutomationBench rose from 17.0% to 30.4%.

I think the price matters as much as those scores because agents can burn through huge amounts of tokens while repeatedly calling tools and handling multi-step jobs. When a developer watches those calls ticking across a screen, every extra attempt has a cost.

Google is also putting 3.7 Flash into Gemini Spark, its 24/7 agent for Pro and Ultra subscribers, which suggests these improvements are meant for everyday work rather than benchmark charts alone.

The release pace is striking.

How quickly will developers switch models when meaningful improvements now arrive within weeks?

Congress Is Already Using AI Faster Than It Can Govern It

AI is already touching legislation.

The Washington Post reports that congressional offices are using chatbots for work ranging from speeches to constituent mail, while formal oversight has struggled to keep pace.

One telling example surfaced in the annual defense bill, where Rep. Anna Paulina Luna’s office submitted an amendment alongside a summary of what it would do.

That matters because congressional staff produce material that can directly shape laws, public statements and how elected officials respond to voters. If a chatbot gets something wrong, the mistake can move from a glowing laptop screen into official government work before anyone catches it.

I think Congress needs clearer disclosure rules for AI-assisted work, especially when the output touches legislation or communication with constituents. Staff will keep using these tools because they save time, so rules that assume otherwise will quickly become irrelevant.

The harder question is simple: when AI contributes to congressional work, who is responsible for checking it?

Oregon Is Rethinking the Price of AI Data Centers

Oregon has about 125 data centers.

AI is pushing facilities toward much heavier computing demands, and state officials are now examining their electricity use, water consumption, land needs and tax treatment.

The number that stands out is $400 million: that’s roughly how much data centers receive each year in Oregon property tax breaks.

I think that figure changes the conversation because communities have a reasonable case for asking exactly what they receive for incentives of that size. You can see the physical tradeoff around facilities like the AWS site in Boardman, where large buildings and electrical infrastructure occupy real land while drawing significant resources.

Oregon has already approved higher electricity rates for some large users, and a state committee is considering new reporting and transparency requirements.

The pressure is coming from rural voters too.

As AI infrastructure expands, how much will states demand in return for hosting it?

Prohuman team

Free Guides

Explore our free guides and products to get into AI and master it.

All of them are free to access and would stay free for you.

Feeling generous?

You know someone who loves breakthroughs as much as you do.

Share The Prohuman it’s how smart people stay one update ahead.