Get tomorrow's brief in your inbox
Daily AI intelligence for small business owners, solopreneurs, and IT professionals
Today: Anthropic refused a Pentagon deal over weapons and surveillance concerns, got blacklisted by the Trump administration, and then watched Claude climb to No. 2 in the App Store as the controversy generated a surge in public support. Google pulled its robotics platform Intrinsic into the main company and is making a serious push into physical AI, projecting a $370 billion robot market by 2040.
This week's biggest AI story isn't about a new model or feature. Anthropic tried to negotiate safeguards into a Department of Defense contract, specifically to prevent its AI from being used for mass domestic surveillance or fully autonomous weapons. The Trump administration responded by directing all federal agencies to stop using Anthropic products. Defense Secretary Pete Hegseth went further, designating Anthropic a "supply-chain threat."
The immediate business fallout surprised everyone: Claude jumped from outside the top 100 apps in late January to No. 2 in the US App Store by Saturday, surpassing Google Gemini and sitting just behind ChatGPT. The controversy drove millions of users to try Claude for the first time. OpenAI moved fast, announcing its own Pentagon agreement the same week. CEO Sam Altman claims it includes safeguards on domestic surveillance and autonomous weapons, though the specific terms were not made public.
What does this mean for your business? If you're using Claude through the API or building on it for client work, there is no service disruption. Anthropic's commercial operations are entirely separate from any government contracts that fell through. The practical upside: this controversy accelerated Claude's user growth significantly, which typically translates into faster product investment and more resources for improving the model. The bigger business lesson here is that when an AI company holds a public line on ethics, it can become a trust signal worth more than any marketing campaign. If you've been evaluating Claude vs. ChatGPT for your team, this week's events are a meaningful data point about how each company weighs commercial pressure against stated principles. Source: TechCrunch | Source: Inc.
Claude memory export is now live. Anthropic launched claude.com/import-memory, a feature that lets you export everything Claude has stored about you into a single copyable block. The prompt Anthropic provides asks Claude to output instructions you've given it (tone, format, preferences), personal details, project context, tools and frameworks you use, and any corrections you've made over time. This matters if you're switching devices, onboarding a new Claude instance, or just auditing what the model has accumulated about your work style. Takes under two minutes. Source: Simon Willison
Square Register 2.0 is out. Square announced a faster hardware update to its register, focused on checkout speed for retail and high-volume transaction businesses. If you're running Square POS and experience any slowdown during peak hours, this is worth considering in your next hardware refresh cycle. No major AI angle here, but faster checkout hardware still reduces friction and cart abandonment for in-person retail. Source: Small Business Trends
Simon Willison published a practical guide on a problem every AI-assisted developer will eventually hit: your agent writes working code, but you no longer fully understand how it works. He calls this "cognitive debt," and it's a real risk, especially as more business logic gets delegated to coding agents.
His solution is to ask Claude to build interactive explanations rather than just written walkthroughs. He had Claude Opus 4.6 create an animated HTML page that visually shows how a word cloud placement algorithm works, step by step. The animation made an obscure concept ("Archimedean spiral placement") immediately clear in a way that a code comment never could.
If you're using Claude Code, Cursor, or any AI coding assistant, build this into your workflow. After any agent-generated feature that touches core logic, ask: "Build me a short animated HTML page that walks through how this algorithm works, step by step." The result becomes documentation your whole team can reference. Claude Opus 4.6 is particularly strong at producing these. The upside is not just understanding your own codebase; it's being able to confidently plan new features on top of it, because you're no longer reasoning about a black box. Source: Simon Willison
According to Inc., companies are starting to deploy AI agents in settings where they make decisions with million-dollar consequences, including procurement routing, operational risk assessment, and resource allocation. The article is thin on specifics, but the direction is clear: the model is shifting from "AI assists a human reviewer" to "AI decides, human reviews exceptions."
If you're evaluating where to apply agents in your business, the best candidates are high-volume decisions with clear data inputs and defined success criteria, such as lead scoring, invoice categorization, inventory reordering triggers, or support ticket routing. These are decisions that are currently costing you time to make manually but are rule-based enough that an agent can handle the majority of cases reliably. Start there before moving into anything that requires contextual judgment. Source: Inc.
Google moved Intrinsic, its robotics software platform, out of "Other Bets" and into the main company. Intrinsic is positioning itself as the Android of robotics: a software layer that runs across hardware from multiple manufacturers, including FANUC, Universal Robots, and KUKA, rather than building its own robots from scratch. The Android comparison is deliberate. Google's playbook in mobile worked because it made development accessible to manufacturers and app builders who couldn't afford to build a full stack from scratch.
With Intrinsic now inside Google, the platform will be tightly integrated with Gemini AI models and DeepMind's research. Google also partnered with Foxconn late last year to deploy AI robots for electronics assembly in US factories. McKinsey projects the general-purpose robot market at $370 billion by 2040.
For most small businesses, this is a 5-to-10-year story. But if you operate in manufacturing, logistics, fulfillment, or any sector where robotic automation has been cost-prohibitive because of high integration complexity, Intrinsic is worth watching. If Google succeeds in building an Android-style layer here, the cost and complexity of adding robotic automation to smaller operations drops substantially. That is the actual bet being made. Source: CNBC
US air strikes on Tehran and Iranian missile sites sent oil prices spiking and Wall Street futures lower on Saturday. This is a reminder that operational resilience matters. When energy and supply costs spike unpredictably, businesses that have already automated high-cost manual processes have more flexibility. If you've been putting off AI adoption for cost reasons, macro volatility cuts both ways: it makes automation more attractive while also pressuring budgets. Focus on automations with the shortest payback period first. Source: Inc.
Export your Claude memory profile. Visit claude.com/import-memory to get started. Paste the export prompt Anthropic provides and Claude will return a full profile of everything it has stored: your instructions, preferences, project context, and any behavioral corrections you've made. This is useful for auditing your data, backing it up before switching devices, or just understanding what the model has learned about your work habits over time.
Eliminate one high-volume manual decision this week. Pick a decision your team makes repeatedly with a predictable data input, such as routing support tickets, scoring inbound leads, or flagging invoices for review. Write down the rule you apply mentally when you make it. Then test whether an AI agent or even a simple prompt chain can handle 80% of those cases reliably. If it can, the remaining 20% still goes to a human, but you've just cut the time cost of that process significantly.
The Anthropic-Pentagon story will keep developing. Watch for whether the executive order gets challenged legally or expanded to other AI vendors, and whether OpenAI's Pentagon deal produces any transparency about the actual safeguards. Google's Intrinsic move is the clearest signal yet that major AI players see physical robotics as the next expansion market; expect more announcements from Amazon, Tesla, and smaller robotics startups in response. The relatively thin week on new AI tools suggests the industry is in a consolidation phase between major model releases. This is a good time to deepen expertise with the tools you already have rather than chasing new launches.
Sources: TechCrunch, CNBC, Simon Willison, Inc., OpenAI Blog, Small Business Trends