Presented by Teleport

There are 144 non-human identities for every human identity. And that number is growing at 44% year over year.

Yet most organizations are still securing AI Agents with old patterns. 

↳Static credentials. 
↳Long-lived tokens. 
↳Shared service accounts.

These were already a liability when humans held them. And we hand them over to an autonomous system provisioning your infrastructure, querying databases, and calling APIs at scale.

Now, you may ask, why are organizations making the same mistakes? Well, the answer is simple. Traditional access models were never designed for autonomous, always-on systems. 

So what can you do? Firstly, treat AI Agents, MCP servers, LLM tools, etc. as first-class identities – same as engineers. 

And that's only possible when you use an Agentic Identity Framework. 

Building on over 10 years of securing the world's most critical software, Teleport has designed an identity model that can contain and control your AI infrastructure. 

Check it out 👉🏾 https://fandf.co/4dpusPW

Thank you to Teleport for sponsoring this newsletter.

Hey {{first name | there}},

Another week wrapped🥳!

Today’s issue spans AI breaking things, scaling fast, and seeping into every layer of the stack, from CI and cloud costs to security, tooling, and infrastructure.

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

Let’s dive in👇🏼

GitHub Actions outage told devs 'your account is suspended'
GitHub Actions goes down for hours, blocking CI/CD pipelines and even showing some users a false “account suspended” error, exposing how fragile build systems can be when the control plane fails. Read more.

The internet is being rebuilt for machines
AWS is redesigning OpenSearch Serverless for AI agents, letting search and vector workloads scale instantly up and down to zero as agent traffic spikes and disappears, aiming to match machine-driven workload patterns. Read more.

GitHub Slashes Agent Workflow Token Spend up to 62%
GitHub says it cut AI agent workflow costs by up to 62% by trimming unused MCP tools, replacing some agent calls with CLI commands, and using agents to continuously audit and optimise token usage inside CI pipelines. Read more.

Bye-bye, Gemini CLI; Google's gone and swapped you for a closed-source AI
Google is replacing Gemini CLI with Antigravity CLI, pushing most users toward a less open replacement that launches without full feature parity. Developers are already criticizing the move over usage limits and the shift away from openness. Read more.

IBM and Red Hat Announce Project Lightwell
IBM and Red Hat are launching Project Lightwell, a $5 billion effort combining AI systems and 20,000 engineers to help enterprises find, patch, and manage vulnerabilities across open-source software supply chains. Read more.

Pullfrog AI: Open-Source CodeRabbit Alternative Powered by GitHub Actions
Pullfrog is a new open source AI GitHub bot that runs entirely inside GitHub Actions, giving developers a model-agnostic alternative to tools like CodeRabbit for PR reviews, issue triage, and CI automation. Read more.

Frog Report Surfaces Need for Rapid DevSecOps Change in AI Era
JFrog says attackers are increasingly targeting AI developer tools, uncovering hundreds of malicious AI models, extensions, and agent skills while teams struggle to secure AI-generated code and modern software supply chains. Read more.

Microsoft wants safer C# without turning it into Rust
Microsoft is redesigning C#’s unsafe model to improve memory safety, making unsafe code more explicit and easier to track across callers while still keeping automatic memory management. The changes are planned for C# 16. Read more.

Mark your calendars!

Cloud / DevOps

AI / ML

Software Engineering / Web

Open Source / Security

General Tech

Presented by Teleport

There are 144 non-human identities for every human identity. And that number is growing at 44% year over year.

Yet most organizations are still securing AI Agents with old patterns. 

↳Static credentials. 
↳Long-lived tokens. 
↳Shared service accounts.

These were already a liability when humans held them. And we hand them over to an autonomous system provisioning your infrastructure, querying databases, and calling APIs at scale.

Now, you may ask, why are organizations making the same mistakes? Well, the answer is simple. Traditional access models were never designed for autonomous, always-on systems. 

So what can you do? Firstly, treat AI Agents, MCP servers, LLM tools, etc. as first-class identities – same as engineers. 

And that's only possible when you use an Agentic Identity Framework. 

Building on over 10 years of securing the world's most critical software, Teleport has designed an identity model that can contain and control your AI infrastructure. 

Check it out 👉🏾 https://fandf.co/4dpusPW

Thank you to Teleport for sponsoring this newsletter.