AWS, Google, and Microsoft walk into a session

Weekly Update | Issue 527

TNS essential reads

AWS, Microsoft, and Google agree the session is the new unit of compute. They disagree on how to isolate it.
Across AWS, Microsoft, Google and Anthropic, agent runtimes are being rebuilt around the session rather than the request — and session-aware execution is fast becoming the baseline for production-grade AI.

After Fable 5 ban, Anthropic and 19 organizations launch open source security body

Hosted by the Linux Foundation, Akrites aims to fix open source security vulnerabilities before attackers can exploit them — founding members include Anthropic, Google, OpenAI, and Microsoft.

“Code should be regenerated, not maintained”: Codeplain makes the case for spec-driven development
The phoenix of Greek myth burns and is reborn from its own ashes -- a metaphor this founder is applying to AI-generated code.

Cursor quietly acquires Continue, an open-source alternative to GitHub Copilot
Cursor acquires Continue, the open-source AI coding assistant with 34K GitHub stars, in a quiet acqui-hire that shuts down the product and hands the codebase to the community.

OpenClaw and Hermes agree on what an agent is. They disagree on what controls it.
Agent harnesses turn AI models into autonomous systems. OpenClaw bets on broad gateways while Hermes bets on memory — and platform vendors want both.

PagerDuty’s CAIO says most AI incident tools are missing a critical layer
Discover why MCP isn't enough for AI incident response and why software teams need an agentic harness with memory to prevent downtime.

Your AI pipeline is broken, and your dashboards don’t know it
Stop treating AI bugs like logic flaws. Discover how to debug and monitor probabilistic LLM pipelines using modern asynchronous tracing.

“I think this is costing us more than we realize”: what permissionless ownership really asks of you
Permissionless ownership means turning "someone should fix this" into "I can help." Learn how to bridge the gap and drive real impact.

Together with Heroku

The developer's guide to connecting CRM data, AI, and app experience at scale

Have you downloaded your copy of our newest eBook yet? Written by Drew Robb and sponsored by Heroku, this brand-new playbook is designed to help you build AI on a data foundation that actually works in production. Learn:

Why the data foundation, not the model, is almost always the deciding factor in AI success.
How to give AI agents access to the unstructured enterprise knowledge that makes up 90% of your organization's information.
How retrieval-augmented generation resolves the accuracy-versus-latency tradeoff that slows most deployments.

Get your free copy

WeAreDevelopers is coming to the US to give unsung developers a bigger voice

The WeAreDevelopers conference expands to San Jose, California, this September. In this episode, WAD co-founder and CEO Sead Ahmetovic and Entire CEO Thomas Dohmke speak with The New Stack about the future of software development ahead of the event. Tune in to learn more!

Catch the episode

Featured webinars

July 22

OpenSearch vs. the black box: one open-source stack for traditional observability and AI agent evaluation

Logs, metrics, and traces work great — until an AI agent starts hallucinating, overconsuming, or making tool calls you didn’t anticipate. Learn how to monitor traditional services and evaluate AI agent behavior in production using one open-source stack built on OpenSearch and OpenTelemetry.

Register today →

July 23

Log Everything, alert smarter: A SecOps framework for AI-driven detection

SecOps teams are caught in a never-ending alert-fatigue cycle. More detection rules generate more alerts, more alerts overwhelm analysts, and overwhelmed analysts miss the threats that matter. Get a practical framework for SecOps teams working to get control of their signal-to-noise problem with specific guidance on where AI makes the biggest operational difference.

Register today →

July 28

Still managing separate access tools for humans, pipelines, and AI agents?

Most access strategies weren’t designed for non-human identities. Traditional PAM tracks admin sessions, and VPNs grant broad network entry. In this session, we will work through what a unified access architecture actually covers today: human developers, contractors, pipelines, and AI agents, under one consistent policy model, without adding IT overhead or slowing down engineering teams.

Register today →

On Demand

From Silos to Governance: Securing IT/OT Data Movement

Find out how organizations managing critical infrastructure are replacing one-off file transfers with governed, resilient data movement at scale. See where IT/OT integration security actually breaks down in practice — and what a modern architecture does differently.

Watch on-demand →

On Demand

The Kubernetes rightsizing trust gap: Why the stakes just got higher

AI workloads have made the Kubernetes rightsizing trust gap expensive. Learn how teams move from human-approved every change to continuous closed-loop optimization — in practice.

Catch the replay →

On Demand

Operationalizing AI in Observability: From Debugging to Automated Remediation

Engineering teams have more data than ever, but humans are still the bottleneck, manually stitching together logs and traces to find answers. Get a practical look at how Datadog’s Bits AI moves teams from manual investigation to autonomous remediation.

Register today →

From our sister site, Towards Data Science

How to create powerful loops in Claude Code

Loops are changing how developers work with coding agents — set a goal, let the agent self-verify, and only come back to you when it's done. Here's what this guide breaks down:

Why self-verifying loops let you run multiple agents in parallel and ship more with less back-and-forth.
The simple command that keeps Claude Code working autonomously until a task is fully complete.
The two-step verification method that dramatically reduces bugs before they hit dev.

Explore the guide

Connect with The New Stack