Devlery

Blog

Notes and analysis on AI development.

Malta makes ChatGPT Plus a national AI access experiment

Malta makes ChatGPT Plus a national AI access experiment

OpenAI and Malta’s AI for All program bundles AI literacy with one year of ChatGPT Plus or Microsoft 365 Copilot access.

The New AI App Vulnerability Is Configuration

The New AI App Vulnerability Is Configuration

Microsoft found exposed AI apps, 15% unauthenticated MCP servers, and Mage AI and kagent cases where defaults became real attack paths.

Codex on mobile turns approval UX into the agent battleground

Codex on mobile turns approval UX into the agent battleground

OpenAI Codex mobile preview shows coding-agent competition moving from model output toward approvals, remote hosts, hooks, and enterprise control.

Glean ADLC turns agent sprawl into a 7-step operating model

Glean ADLC turns agent sprawl into a 7-step operating model

Glean ADLC shows enterprise AI agent competition moving from demo counts to lifecycle, trace, access policy, and ROI measurement.

Copilot Memory now remembers the developer

Copilot Memory now remembers the developer

GitHub Copilot Memory user preferences show coding agents moving from answer quality toward persistent context, work habits, and trust management.

Cloudflare Cut 1,100 Jobs, and Agent Usage Became an Org Chart

Cloudflare Cut 1,100 Jobs, and Agent Usage Became an Org Chart

Cloudflare tied a 600% jump in internal AI use to a 1,100-person workforce reduction. The important story is how agent metrics are entering operating models.

The new bottleneck in AI document extraction

The new bottleneck in AI document extraction

Airia Form Review Step adds human review and audit trails before AI-extracted document data becomes a system-of-record entry.

Anthropic and Gates Foundation's $200M public-good AI experiment

Anthropic and Gates Foundation's $200M public-good AI experiment

Anthropic and the Gates Foundation's $200 million partnership is an experiment in deploying Claude as public-good infrastructure for health, education, and agriculture.

Grok Code Fast 1 Exits Copilot, and Model Routing Becomes the Risk

Grok Code Fast 1 Exits Copilot, and Model Routing Becomes the Risk

GitHub removed Grok Code Fast 1 from Copilot while xAI redirects the retired slug to Grok 4.3. The real issue is coding-agent model routing, cost drift, and operational control.

AWS AgentCore Payments gives AI agents a wallet

AWS AgentCore Payments gives AI agents a wallet

Amazon Bedrock AgentCore Payments previews a runtime layer where AI agents can pay for APIs, MCP servers, web content, and other agents with x402.

Microsoft MDASH shifts security AI from findings to proof

Microsoft MDASH shifts security AI from findings to proof

Microsoft says MDASH used 100+ agents to find 16 Windows vulnerabilities. The security AI race is moving from model scores to validation harnesses.

Copilot code review double billing makes AI review a CI cost

Copilot code review double billing makes AI review a CI cost

GitHub Copilot code review will consume both AI Credits and Actions minutes from June 1. AI review is becoming an operational CI workload.