Get tomorrow's brief in your inbox
Today: Anthropic just made Claude Code more expensive for third-party tools, forcing OpenClaw users to pay-as-you-go starting today. Oracle shipped AI database tools with no-code workflows for small businesses. AI-generated music fakes are hitting streaming platforms, and distributors are struggling to stop copyright trolls.
Anthropic is splitting its Claude Code pricing into two tiers, and if you're using third-party tools like OpenClaw, you're about to pay more. Starting April 4 at noon Pacific, Claude Code subscribers can no longer use subscription limits for third-party harnesses. Instead, they'll pay extra through separate pay-as-you-go billing. The policy starts with OpenClaw but rolls out to all third-party tools soon.
Anthropic's head of Claude Code says the company's subscriptions "weren't built for the usage patterns of these third-party tools" and they're trying to manage growth sustainably. OpenClaw creator Peter Steinberger called the move suspicious, pointing out that Anthropic copied popular features into their closed harness, then locked out open source alternatives. He tried negotiating but only delayed the price increase by one week before joining OpenAI.
For small businesses already using Claude Code with third-party integrations, this means reviewing your monthly costs. If you're heavily integrated with OpenClaw or similar tools, budget for the pay-as-you-go charges or consider switching to Anthropic's native harness. Full refunds are available for current subscribers who didn't realize third-party tools weren't officially supported.
Oracle AI Database with no-code workflows - Oracle rolled out three new AI database features aimed at small businesses: an Autonomous AI Vector Database with a free tier for building vector apps, a Private Agent Factory for creating AI workflows without coding, and Unified Memory Core for managing context across data types without multiple databases. The no-code approach lets business analysts build AI workflows while keeping data secure and in-house. Initial pricing includes free and low-cost developer tiers with easy scaling as needs grow. The catch: integrating with existing systems can be complex, and you'll need dedicated resources to manage AI applications long-term. If you're looking to add AI without hiring developers, the Private Agent Factory is worth testing on the free tier.
Cisco Zero Trust for AI agents - Cisco announced security measures specifically for businesses deploying AI agents. Key features include Zero Trust Access (each agent gets verified identity tied to a human manager), Dynamic Agent Red Teaming (test AI apps for vulnerabilities before deployment), and automated security frameworks that eliminate manual checks. If you're experimenting with AI agents but worried about security, Cisco's tools add accountability and regulatory compliance. The downside: implementation costs and training requirements for small teams. Best for businesses already using Cisco infrastructure.
Simon Willison's LLM abstraction layer research - Willison is redesigning his LLM Python library to handle new features like server-side tool execution across hundreds of models. He had Claude Code read through client libraries for Anthropic, OpenAI, Gemini, and Mistral to craft curl commands for streaming and non-streaming modes. The scripts and outputs are now in a public repo. If you're building cross-LLM applications, this repo offers real-world examples of handling different vendor APIs.
scan-for-secrets 0.2 - Updated CLI tool for finding exposed secrets in codebases. New version streams results as they're found (better for large directories), supports multiple directories with -d flag, adds -f for scanning individual files, and includes new Python API functions. The -v verbose mode shows each directory being scanned. If you're auditing repos for API keys or credentials before deploying, this tool just got faster and more flexible.
Free business software roundup - Small Business Trends compiled seven free tools worth considering: EngageBay (CRM for 250 contacts with visual sales pipeline), MailerLite (email marketing for 1,000 subscribers, 12,000 emails/month), Trello (unlimited users, 10 Kanban boards), Wave (unlimited invoicing with mobile access), and Zapier (workflow automation connecting apps). If you're bootstrapping, start with EngageBay for CRM and MailerLite for email. Zapier's free tier is limited but useful for connecting two apps.
Apple at 50: playing catch-up in AI - Apple struck a multiyear deal with Google to use Gemini AI as part of a rebooted Siri. Google already pays around $20 billion/year to be the default iPhone search engine. In the AI arrangement, Apple pays Google for the underlying intelligence. The concern: whether Google uses iPhone user data to improve its core business, violating Apple's privacy principles. Apple has kept capital expenditures low compared to Amazon, Microsoft, Alphabet, and Meta, which are collectively committing hundreds of billions to AI infrastructure. For small businesses, this means Siri improvements will come slower than ChatGPT or Gemini updates. If you're betting on Apple Intelligence for business workflows, plan for delays.
AI fakes hit streaming platforms - Folk musician Murphy Campbell discovered AI-generated covers of her YouTube performances uploaded to Spotify, Apple Music, and YouTube Music under her name. She eventually got most removed, but at least one fake song remains on Spotify under a different profile with the same artist name. Spotify is testing a manual approval system for songs before they appear on artist profiles. Separately, copyright trolls used unlisted YouTube videos to claim ownership of her public domain recordings like "In the Pines" and "Darling Corey." Distributor Vydia released the claims and banned the uploader, saying only 0.02% of their 6 million Content ID claims are invalid. For musicians and content creators, this shows how easily AI tools and Content ID can be weaponized. Manually approve everything before it goes live if your platform offers that option.
Human-made content labels proliferate - Creators want an "AI-free" label to distinguish their work, but at least 12 different labeling services exist with varying eligibility criteria and verification methods. Some operate purely on trust (Made by Human), others use AI detection tools (No-AI-Icon), and some manually audit creative processes by reviewing sketches and drafts. The problem: no universal standard, and defining "human-made" gets murky when AI is embedded in creative tools. Instagram head Adam Mosseri suggested fingerprinting real media might be more practical than labeling fakes. For businesses creating content, don't rely on AI detection services (they're unreliable). If you want to prove human authorship, save drafts, sketches, and version history.
Audit your Claude Code integrations - If you're using Claude Code with third-party tools, review your current usage and budget for pay-as-you-go charges. Anthropic is offering full refunds for subscribers affected by the policy change. Check your integration list today before the new pricing kicks in.
Test Oracle's no-code AI factory - If you've been holding off on AI because you lack developers, Oracle's Private Agent Factory offers a free tier for building AI workflows without code. Sign up for the developer tier and prototype one workflow (customer inquiry routing, invoice processing, inventory alerts) to see if it fits your needs before scaling up.
AI pricing is fragmenting. Anthropic's move to charge separately for third-party tools signals a broader trend where subscription models won't cover all use cases. Expect more vendors to split pricing between native integrations and external harnesses. Meanwhile, the AI fake content problem is spreading from images to music and video, with platform verification systems struggling to keep up. Businesses should prioritize provable authorship (version control, audit trails) over relying on detection tools.