Meta released Muse Glimmer on Monday, a 30-billion-parameter open-weight model designed to run agentic workflows on local hardware. It’s available The post Meta’s Muse Glimmer fits on a laptop appeared first on The New Stack.
AI generation is temporarily unavailable; this entry was preserved with deterministic fallback metadata.
Meta released Muse Glimmer on Monday, a 30-billion-parameter open-weight model designed to run agentic workflows on local hardware. It’s available The post Meta’s Muse Glimmer fit…
OpenAI made its family of GPT-5.6 models available to its app and API users globally at the start of July. The post “It blows my mind”-“It has a tendency to overengineer things a little”: Developers react to road-testing OpenAI GPT‑5.6 Sol appeared first on The New Stack.
AI generation is temporarily unavailable; this entry was preserved with deterministic fallback metadata.
OpenAI made its family of GPT-5.6 models available to its app and API users globally at the start of July. The post “It blows my mind”-“It has a tendency to overengineer things a…
In the early days of Claude Code, it felt like you either had to approve everything the coding agent did The post Auto Mode will soon be the default in Claude Code — because humans can’t be trusted appeared first on The New Stack.
AI generation is temporarily unavailable; this entry was preserved with deterministic fallback metadata.
In the early days of Claude Code, it felt like you either had to approve everything the coding agent did The post Auto Mode will soon be the default in Claude Code — because human…
AWS on Thursday launched Dogwood, an open-source policy language and reference interpreter that lets developers govern sequences of AI agent The post Your AI agent’s next tool call may be valid but wrong. AWS’s Dogwood promises to fix that. appeared first on The New Stack.
AI generation is temporarily unavailable; this entry was preserved with deterministic fallback metadata.
AWS on Thursday launched Dogwood, an open-source policy language and reference interpreter that lets developers govern sequences of AI agent The post Your AI agent’s next tool cal…
Doist CTO Gonçalo Silva doesn’t pretend to know where AI is headed. The pace of change has made it harder The post Why Todoist says less AI can deliver more appeared first on The New Stack.
AI generation is temporarily unavailable; this entry was preserved with deterministic fallback metadata.
Doist CTO Gonçalo Silva doesn’t pretend to know where AI is headed. The pace of change has made it harder The post Why Todoist says less AI can deliver more appeared first on The…
Most open source maintainers will know the feeling of opening GitHub to a growing pile of issues that can’t feasibly The post Astro’s GitHub issue backlog is heading to zero for the first time in 5 years. Now Cloudflare is open-sourcing the tool that did it. appeared first on The New Stack.
AI generation is temporarily unavailable; this entry was preserved with deterministic fallback metadata.
Most open source maintainers will know the feeling of opening GitHub to a growing pile of issues that can’t feasibly The post Astro’s GitHub issue backlog is heading to zero for t…
Apple now caps how many security reports some researchers can have open at once. And once they hit that cap, The post Apple and Bynario agree GPT-5.5 found a real macOS bug. They disagree on the report cap. appeared first on The New Stack.
AI generation is temporarily unavailable; this entry was preserved with deterministic fallback metadata.
Apple now caps how many security reports some researchers can have open at once. And once they hit that cap, The post Apple and Bynario agree GPT-5.5 found a real macOS bug. They…
Cloud platform company Nscale has agreed to acquire Anyscale, the AI workload scaling specialist, pairing GPU neocloud infrastructure with an independent multi-cloud orchestration layer. Nscale insists Anyscale will stay neutral and performance-led, while analysts warn that 'runs anywhere' and 'runs best somewhere' are different — and neutrality may become just a label. The deal is expected to close in H2 2026 at roughly $1.65 billion.
Nscale acquires Anyscale, combining GPU neocloud infrastructure with a cloud-neutral AI scaling platform.
Nscale says Anyscale will keep its brand and support BYOC across AWS, GCP, and Azure, vowing to win on performance rather than lock-in.
Over 1,100 AI researchers and executives signed an open letter urging governments to slow frontier AI development if safety measures lag. Anthropic CEO Dario Amodei is the only top AI lab CEO to sign; OpenAI and Google DeepMind leaders also signed. The letter cites risks of rapid capability acceleration and self-improving models. Congress is considering related legislation like the AI Kill Switch Act.
More than 1,100 AI experts signed an open letter calling for deliberate slowdown of frontier AI development if safety can't keep pace.
Anthropic CEO Dario Amodei is the only top AI lab CEO to sign; OpenAI and Google DeepMind leaders also signed.
Cursor launches Cursor Start at ₹649/month with UPI, Anthropic introduces rupee pricing but costs more than US, OpenAI's ChatGPT Go at ₹399/month with UPI. Cursor and OpenAI focused on value with tailored plans and payment methods, while Anthropic only converted currency.
Cursor launches Cursor Start at ₹649/month with UPI, excluding frontier models
Anthropic's rupee pricing is more expensive than US equivalent, no UPI support yet
Perplexity launched SPACE, a sandbox platform for its Computer AI agent, focusing on state management rather than isolation. It uses Firecracker, Kubernetes, and Btrfs for efficient snapshots, pause/resume, and forking. SPACE offers 3x performance improvement over incumbents, with rolling snapshots every minute and rewinding up to a week. Security features include RBAC, just-in-time access, and per-tool controls. Future plans include a sandbox API, third-party sandbox compatibility, and local/hybrid deployments. Perplexity also introduced a cost-efficient orchestrator model.
Perplexity's SPACE sandbox platform prioritizes state management (pause/resume/fork) over isolation.
Uses Firecracker microVMs, Kubernetes, and Btrfs copy-on-write filesystem for performance.
Modus exits stealth with $10M in funding to build a 'context warehouse' that continuously maps business operations and provides AI agents with only the relevant context, saving costs and improving accuracy. The startup targets engineering and AI teams, using a Context Miner and Context Composer to dynamically generate skills.
Modus has raised $10M from Insight Partners to launch a context warehouse that sits alongside existing data warehouses.
It continuously crawls assets from sources like GitHub, Jira, and Snowflake to understand business context.
In a podcast appearance, OpenAI CEO Sam Altman discussed AI security, model distillation, and the future of intelligence. He emphasized that scale and usage are more important than high margins, and while security incidents like the Hugging Face event were concerning, model distillation is not a top worry. Altman believes cheap models will exist, and OpenAI must be both the best and cheapest. He also noted the need for better safeguards and credential management, and sees open source models continuing to have a place.
Altman says model distillation is not in his top ten worries; OpenAI focuses on being the greatest and cheapest.
The Hugging Face security incident forced OpenAI to rethink model security; agentic credentials need just-in-time injection.
Diagrid's Catalyst 2.0 adds durable execution and attestation for AI agents built with LangGraph, Microsoft Agent Framework, and others, allowing recovery from interruptions and providing a signed, tamper-evident execution history for compliance in regulated industries.
Catalyst 2.0 provides durable execution for AI agents across multiple frameworks, enabling recovery from the last completed step after interruptions.
It runs beneath existing frameworks, intercepting the agent loop and registering operations as workflow steps using Dapr.
After eight years of building a no-code automation platform, Tines launched 3B Tuesday, a new platform that uses AI to author enterprise workflows but still uses conventional code to execute them.
Tines launches 3B Tuesday, using AI to generate workflow code from natural language, but executing with deterministic code.
CEO Eoin Hinchy states low-code/no-code has a sell-by date as AI can now write code, reducing need for visual builders.
Anthropic CEO Dario Amodei clarifies the company does not want to ban open-weight AI models, but proposes mandatory safety testing for sufficiently capable models before release. The proposal has drawn 50 signatories including OpenAI and Google, with Anthropic and Amazon notably absent.
Anthropic CEO says company does not advocate banning open-weight models, but calls for mandatory safety tests
Proposal includes tighter export controls, cracking down on model distillation, and pre-release evaluations for capable models
Observability platform company Dynatrace announced advancements to its Dynatrace Intelligence service, moving from probabilistic approaches to deterministic real-time context and control with autonomous SRE agents, while maintaining human oversight.
Dynatrace adds autonomous SRE agents that automatically determine if an issue is part of an existing incident investigation.
Agent Builder enables no-code creation of custom AI agents.
Pilot Protocol emerges from stealth with an agent app store and network that allows agents to discover each other, communicate, and pay for tools via wallets. With 250,000 agents already generating 2 billion requests daily, developers embrace this as the future of autonomous usage patterns.
Pilot Protocol provides an agent app store and network for discovery and communication among agents.
Agents receive wallets to pay for tools, enabling a usage-based distribution model for developers.
Cloudflare has open-sourced pvcli, a command-line debugger for privacy protocols like Oblivious HTTP (OHTTP) and MASQUE, designed to troubleshoot split-trust infrastructure used by Apple iCloud Private Relay and Microsoft Edge Secure Network VPN. The tool, built with AI agents in mind, uses curl-like syntax to help smaller companies test and debug privacy-proxied traffic without needing deep protocol expertise.
Cloudflare open-sourced pvcli, a debugger for OHTTP and MASQUE privacy protocols
Addresses debugging challenges in split-trust systems like Apple iCloud Private Relay and Microsoft Edge Secure Network VPN
The Open Secure AI Alliance, formed by 33 partners including Nvidia, Palantir, and Hugging Face, aims to develop techniques and tools to safeguard open-weight AI models by rapidly identifying and patching vulnerabilities. The alliance highlights the regulatory gap for open models and emphasizes infrastructure-level security.
33 partners form Open Secure AI Alliance to protect open-weight AI models. Notable members include Nvidia, Adobe, Cisco, IBM, and Microsoft, but OpenAI and Anthropic are absent.
Experts argue current AI safety regulations focus on closed models, leaving open-weight models in a regulatory blind spot.
The Model Context Protocol (MCP) receives its largest update since launch, removing session state and initialization handshake to simplify remote server operations. The release candidate is frozen, final spec due July 28. Deprecations include core features like Sampling, with migration directions provided.
MCP's update eliminates session affinity by making requests stateless, reducing operational complexity.
Capabilities and protocol version are now carried per-call via _meta, enabling caching and routing.
A shift from correcting AI-generated code to improving the systems that produce it is necessary for software engineers. Patrick Debois of Tessl advocates for context-driven development and harness engineering to scale AI agents effectively.
Engineers should stop correcting AI code and instead improve the system and context.
Context-driven development involves continuous loops of generation, evaluation, distribution, and observation.
The debate over open-weight AI models intensified as 25 major organizations signed a statement warning against premature restrictions, while Anthropic and OpenAI declined. Developers increasingly turn to Chinese open-weight models due to cost, triggering tensions between national security and open technology.
Microsoft, Nvidia, Meta, and 22 other organizations signed a statement defending distillation as legitimate model development.
Anthropic and OpenAI declined to sign, favoring restrictions on Chinese open-weight models.
Anthropic's Opus 5 is cheaper and less restrictive than Fable 5, excelling in agentic coding tasks at a fraction of the cost. However, its autonomy introduces security and cost management challenges for teams.
Opus 5 is priced at $5/$25 per million tokens, outperforming competitors at lower cost.
It can work on programming tasks for extended periods without human input.
Anthropic launched Opus 5 on Friday, the latest version of what used to be the company’s flagship model (before the launch of Fable 5). Opus 5 comes close to Fable 5 performance at half the price, with no data retention policy required. It leads on several benchmarks and introduces new safety features and fast mode.
Opus 5 approaches Fable 5 performance at half the cost, without the 30-day data retention policy.
Tops benchmarks in knowledge work, agentic search, and business workflows, among others.
All four major clouds now provide native agent sandboxes for isolated execution of untrusted code, but each uses a different isolation technology: Firecracker microVMs (AWS), gVisor kernel interception and lightweight Cloud Run isolation (Google), Hyper-V boundaries (Azure), and containers plus dedicated VMs (Cloudflare). The article argues that sandboxes are just containment boundaries; the real challenge is governance of credentials and behavior.
All four major cloud providers now offer agent sandboxes as a native primitive, with significantly different isolation stacks and lifecycle models.
AWS Lambda MicroVMs use Firecracker for dedicated VMs with up to 8 hours runtime and suspend-resume support.
OpenAI and Anthropic released competing voice updates. OpenAI's ChatGPT Voice aims to control desktops and AI agents, while Anthropic's Claude Voice focuses on conversational reasoning for complex problems.
OpenAI introduces GPT-Live for desktop control via voice, allowing multitasking and app integration.
Anthropic enhances Claude Voice Mode for iterative problem-solving through conversation.
NVIDIA's senior director of generative AI software, Joey Conway, discusses how local open models are increasingly working alongside frontier models, with routers deciding which to use, enabling organizations to achieve better outcomes at lower cost and latency.
NVIDIA advocates combining local and frontier models via intelligent routing to match task complexity.
Hardware like DGX Spark allows running up to 200B-parameter models locally, offering full data control.
Cursor launches a model router that directs coding tasks to the best-fit AI model, claiming 30-50% cost savings. Ramp and Meta also release similar tools. Cursor itself is developing proprietary models and taking control of its AI stack.
Cursor Router uses a triage system to assign requests to optimal models, cutting costs without quality loss.
Cursor has unveiled its own models, Grok 4.5 and Composer 2.5, to reduce reliance on external providers.
Software delivery lifecycle company Harness launched its AI Agent Development Lifecycle (DLC) service to apply the same governance, testing, and security used for application code to AI agents. The challenge is agents' non-deterministic nature; Harness focuses on making the pipeline predictable rather than the agent itself. It introduces five new capabilities: AI Evals, Agent deployments, AI configs, AI asset catalog, and AgentTrace, along with open-sourcing foundational components. The goal is to enable safe, governed agentic deployments.
Harness launches AI Agent DLC to apply code delivery pipeline governance to agent development.
Agents are non-deterministic; Harness advocates for predictable pipelines around them.