BIG TECHAWAYS

WHAT MATTERS TODAY

​Google splits Gemini 3.8 Flash into two tracks​

Google has brought Gemini 3.8 Flash into the race to run longer agent tasks while introducing a Flash Cyber version that is not broadly available. The two-track launch shows how the same model capability can come with different access policies: the standard Flash targets reasoning and coding, while Flash Cyber is available through the Fairwind Program to trusted defense organizations.

Google is keeping its introductory price at $0.75 per million input tokens and $3.75 per million output tokens through December 31, 2026. For model buyers, that is a reference point, not the whole equation: execution time, number of steps, and output volume can vary across workflows even under the same price sheet.

The practical question is which model fits the job and the level of control a team needs, and only then the cost per completed task. Third-party evaluations are additional context, not benchmarks published by Google, so there is not enough basis to call Gemini 3.8 cheaper, faster, or better than every frontier model.

​Muse Spark 1.3 helps agents stop retracing their steps​

Muse Spark 1.3 is now available in Muse Code and the Meta Model API, bringing the change to users through tools they already use rather than only through a new model page. Meta says the version is stronger at coding and longer-horizon agent tasks, asks for clarification when requirements are unclear, and has stronger defenses against prompt injection.

In an internal comparison with version 1.2, Meta reports about 20% fewer tool calls and 25% fewer tokens. Those are Meta-provided figures, not an independent benchmark.

They point to a product shift: agent reliability is becoming part of the experience. Retaining important instructions and avoiding rework can free up time in a workflow, although the actual savings will depend on the task.

PRESENTED BY CASH APP

Teach them money habits that last

Helping your teen learn to manage money is a big step, and Cash App Card makes it easy for both of you. They get a secure debit card and easy ways to save and send money, while you stay in the loop every step of the way. All with no monthly or hidden fees for you or your teen.

Why millions of families trust Cash App Card:

Instant discounts on everyday spending

Add offers and save on groceries, restaurants, rides, and more.

No monthly or hidden fees Shop in-store and online where Visa is accepted with no monthly or hidden fees.

24/7 fraud monitoring Avoid scams before they can happen with proactive security features.

Oversight and control Review your teen’s spending, control who can send them requests, and lock their card if it’s ever lost or stolen.

​Get started with Cash App​

Cash App is a financial services platform, not a bank. Banking services provided by Cash App’s bank partner(s). Prepaid debit cards issued by Sutton Bank, Member FDIC. Cash App Visa® Debit Flex Cards issued by Sutton Bank, Member FDIC, and The Bancorp Bank, N.A., pursuant to a license from Visa U.S.A. Inc. See terms and conditions for the Sutton prepaid card (https://cash.app/legal/us/en-us/card-agreement), Sutton debit flex card (https://cash.app/legal/us/en-us/debit-flex-card-agreement-sutton), and Bancorp debit flex card (https:// cash.app/legal/us/en-us/debit-flex-card-agreement-bancorp).

BIG THINK

​Data centers are competing for workers and HVAC capacity​

Chicago Fed President Austan Goolsbee told Fortune that the expansion of data centers is competing with other industries for construction workers and HVAC capacity. He distinguishes a region where resources are being pulled toward one use from an economy-wide overheating problem.

The examples are concrete: some businesses in the Seventh District have scaled back plans because construction labor costs are rising and HVAC crews are hard to find. Before AI creates broad inflation, it may already be changing prices and deployment schedules for projects competing in the same region.

Goolsbee also said current labor-market stability is not mainly coming from data centers. That is an assessment of crowding-out risk, not evidence that AI has caused economy-wide inflation.

SHIFT SIGNALS

BEHIND THE HEADLINES

• ​US AI policy disagreements surface at the G20: Axios reports that US officials want to project AI leadership at the G20 while disagreements between the Commerce Department and the White House technology policy team continue at home. Policy is becoming a variable in AI infrastructure and distribution.
• ​Palantir and Anduril win a $192 million contract for eight TITAN systems: The US Army signed contracts worth a combined $192 million with Palantir and Anduril to produce eight Tactical Intelligence Targeting Access Nodes. AI-enabled defense is moving from prototype stories into procurement processes.
• ​Aitan raises $41 million for edge AI weapon systems: Axios reports that Aitan has emerged from stealth with $41 million and offers edge AI weapon systems. Claims about prior operational experience come from the company; the funding shows capital moving toward autonomous systems that operate outside a centralized cloud.
• ​Delivery robots return to college campuses: College campuses are becoming test environments for delivery robots. Their relatively contained geography still provides enough real-world situations to test physical AI reliability before expansion.
• ​AI starts helping allocate firefighting teams: Axios reports that research groups are testing AI to help officials allocate firefighting teams across multiple fires. The models prepare options and compare trade-offs, while the final decision remains with people.
• ​Gemini 3.8 Flash shows how model release cycles are speeding up: Google released Gemini 3.8 Flash and a cybersecurity-oriented version only weeks after the previous Flash generation. Shorter model cycles mean buyers have to measure results by task and workflow instead of comparing list prices alone.