The Informed Layer · Newsscan
Daily briefing

Agentic dev tools surge; AI labeling bill introduced

← All issues

Anthropic / Claude ecosystem

Anthropic launches drug discovery programs using Claude Science

Anthropic is using its Claude Science AI workbench to accelerate early-stage drug discovery for neglected diseases that Big Pharma often finds unprofitable. The company aims to expand Claude's capabilities from coding and summarization to preclinical drug research, providing AI tools for the broader industry.

Anthropic reduces Claude Code system prompt by 80%

Anthropic cut Claude Code's system prompt size by 80%, allowing newer Fable 5 models to operate with fewer instructions and examples. The change suggests that the latest models require less explicit steering, improving efficiency and potentially reducing token usage.

Frontier model providers

DeepSeek launches image recognition mode on app and web

DeepSeek has officially launched its image recognition function, now available on both the mobile app and web platform. The new feature enables users to upload images for analysis, expanding the model's multimodal capabilities.

AI developer tooling & infrastructure

Chinese startup Z.ai launches ZCode, undercutting US rivals with agent-first IDE

Z.ai has released ZCode, a free desktop IDE environment built around its GLM-5.2 model that emphasizes agent-driven development. It aims to undercut US coding assistants on price while handling long-horizon software tasks.

DuneSlide: Cursor IDE flaws allow remote code execution via prompt injection

Two zero-click vulnerabilities in Cursor IDE (CVSS 9.8) enable attackers to escape the sandbox and execute arbitrary code through crafted prompts. Users are urged to upgrade to Cursor 3.0 immediately.

LlamaIndex introduced Legal-KB, a new application designed for agentic retrieval workflows in legal contexts. It demonstrates the platform's expansion into domain-specific enterprise solutions.

mcpsnoop: open-source proxy brings transparency to MCP debugging

mcpsnoop is a new open-source tool that acts as a transparent proxy in the MCP data path, observing production traffic and JSON-RPC frames in real time. It reveals what the standard MCP Inspector misses, enabling deeper debugging of agent interactions.

pxpipe hides text in PNGs to slash Claude Code token costs by 70%

An open-source tool called pxpipe reduces AI token costs by rendering text as images (PNG format) before sending to Claude Code or Fable 5. While it trades some accuracy and speed, the approach significantly cuts per-token costs.

Cloud & platform providers

No significant new developments.

AI policy, regulation & governance

UN scientific panel releases preliminary report on AI capabilities, risks, and opportunities

The Independent International Scientific Panel on Artificial Intelligence (IISPAI) has published its first global assessment, covering AI’s opportunities, risks, and impacts. The report aims to inform international governance discussions with evidence-based analysis.

US AI Labeling Act of 2026 introduced in Senate

Senator Brian Schatz introduced the AI Labeling Act (S. 4915), which would require AI-generated content to be labeled across platforms. The bill proposes a dedicated standards group to develop best practices for labeling.

India considers dedicated AI legislation

India is exploring the development of a standalone AI law, signaling a shift toward formal regulatory governance of artificial intelligence. This move aligns with global trends but will need to balance innovation with safety.

Tasmania plans legislation to combat revenge porn, deepfakes, and digital tracking

The Tasmanian government is introducing draft laws to criminalize the non-consensual sharing of intimate images, including deepfakes, and covert digital tracking. The legislation aims to modernize protections against technology-facilitated abuse.

Industry & market moves

Aily Labs partners with AWS to deploy AI decision intelligence agents

Aily Labs announced a strategic partnership with AWS to make its AI Decision Intelligence agents available on AWS Marketplace. The integration enables enterprise customers to deploy autonomous decision agents within their own cloud environment quickly and securely.

Microsoft names Logicalis Australia Frontier Partner for governed enterprise AI deployments

Microsoft has designated Logicalis Australia as a Frontier Partner, a rare status that signals a governance-forward approach to enterprise AI rollouts in the Australian market. This partnership aims to accelerate secure, compliant AI deployments for Australian organizations.

Datadog acquires Adaptive ML to build specialized AI agents

Datadog has acquired Adaptive ML to strengthen its AI capabilities, enabling the development of specialized agents and models that leverage real-time signals. The deal expands Datadog's AIOps and observability offerings.

Zoom acquires Common Room to boost AI-powered sales tools

Zoom has acquired Common Room, an AI-based go-to-market intelligence platform, to enhance its sales and pre-sales capabilities. The acquisition signals Zoom's expansion beyond video conferencing into AI-driven workflow automation.

CPP Investments and EQT commit $2.4B to EdgeConneX for AI infrastructure

Canadian Pension Plan Investments and EQT have committed $2.4 billion to EdgeConneX, a data center operator, to expand AI-focused infrastructure. The investment aims to meet the growing demand for AI compute capacity.

AI product & feature launches

Microsoft Copilot to merge into single 'Fusion' app, signals paid-adoption crisis

Microsoft plans to consolidate Copilot into a unified 'Copilot Fusion' app in August, bundling features like AutoPilot agents and Podcasts. The move comes amid feature cuts aimed at boosting paid adoption, suggesting struggles to monetize AI assistants.

Greenhouse MCP enters open beta, making ATS optional for agents

Greenhouse's MCP is opening to all Site Admins on July 6, enabling governed agent actions with read-first access. This shift allows AI agents to interact with applicant tracking systems in a controlled, auditable way.

Cognizant uses OpenAI's GPT-5.5 to automate patch deployment on Windows servers

Cognizant has launched a service leveraging GPT-5.5 for autonomous vulnerability remediation and patch deployment on Windows servers, with auditable, sandboxed validation. The solution aims to improve patch deployment speed and security.

Research with immediate practical relevance

ByteDance discovers new scaling law for post-deployment AI learning

Researchers at ByteDance have identified a scaling law for post-deployment learning, showing how AI agents can improve continuously after deployment. The finding, demonstrated via EdgeBench across 38,000+ hours of interaction, could sustain AI progress as pre-training scaling plateaus.