Hey {{first name | there}}. This week, Linus Torvalds put his foot down. Told the contributors grumbling about AI they could fork the kernel or walk away, and called AI a useful tool, a real turn from the man who dismissed it as 90% hype last year

The rest of the week showed why the tool still needs watching.

In today's roundup:

You will also find upcoming events, a podcast worth listening to, and curated job opportunities across the ecosystem.

Let’s dive in 👇🏼

Linus Torvalds tells AI haters to fork off
Torvalds drew a line as top maintainer: Linux isn't an anti-AI project, and objectors can fork it or leave. He calls AI a useful tool now, a sharp softening from his "90% hype" stance last year.

‘HalluSquatting’ Compromises AI Coding Agents to Install Malware
Researchers registered the fake package names coding agents tend to invent, laced them with malware, and waited. Agents hallucinated false repo names 85% of the time, and skill names 100%

AI Agents with Cloud Credentials Are Outrunning Billing Guardrails Built for Human-Speed Mistakes
Agents and leaked credentials can burn thousands on cloud AI before anyone notices, since billing lags a day behind. One three-person shop's usual bill jumped from $15 to $14,000 overnight.

Microsoft is reportedly training salespeople to talk down OpenAI and Anthropic
Bloomberg reports Microsoft's sales team is now told to pitch against Claude and OpenAI, the very companies whose models power its own apps, as it swaps them out for cheaper in-house ones.

Checkmarx introduces self-healing application security
Checkmarx now runs a find-and-fix loop inside AI coding tools, detecting a flaw, generating a fix, and verifying it before code is committed, aiming to cut manual remediation by up to 70%.

SpaceX open-sources Grok Build while beaming users' repos to the cloud
After researchers found Grok Build packaging entire repos and sending them to the cloud, SpaceX open-sourced the tool and promised to delete all retained data. The upload code is still there, apparently reversed. 

GitHub's redesigned PR inbox tackles the review bottleneck AI created
GitHub's new pull request inbox sorts reviews by urgency and even attributes agent-opened PRs to the human behind them. Heavy AI teams merge 98% more PRs, but review time is up 91%.

The real AI race may no longer be at the frontier
While everyone watched Anthropic and OpenAI, open models took over production. Chinese open-weight models passed US ones at 41% of Hugging Face downloads, and hold the top six spots on OpenRouter.

Developing and Deploying a Platform that the Business Understands and Developers Actually Want
Two engineers rebuilt adoption of their internal platform by selling it like a product: talk to stakeholders, measure value with DORA, and make the pain personal. What finally landed was the story, not the tech.

SmartBear adds integrations with Claude, Atlassian, GitHub, and Kiro
SmartBear is embedding its testing and governance tools directly inside the AI-native IDEs and assistants developers already work in. Hence, quality checks happen in the flow rather than in a separate tool.

Mark your calendars!