Anthropic draws a line on third-party tools

Anthropic tightens control over developer usage

Anthropic just cut off a popular way developers were stretching Claude Code subscriptions.

Starting April 4 at noon PT, users can’t apply their subscription limits to tools like OpenClaw and now have to pay separately on a usage basis. The change will extend to all third-party “harnesses,” even as OpenClaw’s creator moves to OpenAI and keeps the project open source.

This looks deliberate. Anthropic says it’s about compute and sustainability, but it also pulls developers back into its own environment right as rivals compete for the same users.

The timing matters because Claude Code has been gaining traction with engineers who rely on external tools to customize workflows. Charging separately for that behavior makes experimentation more expensive and less predictable, which tends to slow down power users first.

Picture someone watching usage costs tick up on a second monitor while running builds. That friction pushes teams to rethink which platform they standardize on. Does this clean up costs, or just send developers somewhere else?

AI moves into frontline operations

Sona just raised $45 million to rebuild how shift-based businesses run.

The company now has over $100 million in funding and is pushing into the US with software that combines scheduling, HR, payroll, and analytics in one system.
It already works with brands like Popeyes and Tao Group, using live inputs like bookings, revenue, weather, and past shifts to adjust staffing in real time.

This moved fast. The pitch is simple: frontline software stayed frozen while everything else upgraded, and now AI is compressing that catch-up into a short window.

I think the interesting part is Forge, which lets companies build their own internal tools on top of Sona’s system using their actual operating data. That turns the product from a tool into infrastructure, which is harder to replace once it is embedded across payroll, scheduling, and decision-making.

You can picture a manager checking staffing on a phone during a noisy dinner rush.
If this works, smaller operators get capabilities that used to require dedicated ops teams. That matters. But can one platform really own all of this without breaking under complexity?

AI companies still won’t stand behind outputs

Microsoft’s Copilot terms quietly say the product is “for entertainment purposes only.”

The language, last updated October 24, 2025, warns users not to rely on outputs for important decisions and to use it at their own risk. Microsoft now says this is outdated and will be revised, while similar disclaimers remain standard across OpenAI and xAI products.

This is awkward. Companies are selling AI into core workflows while their legal terms still treat it like a novelty that can’t be trusted.

The gap matters because enterprise adoption depends on reliability, and legal language is usually the most honest signal about how a product actually performs under pressure. If the terms lag behind the product story, buyers will notice, especially in regulated industries where risk teams read every line.

You can imagine someone scrolling through terms late at night before signing a contract. That sentence stands out more than any demo. At what point do these tools take real responsibility for their output?

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.