Welcome to the 295th edition of Coding Jag brought to you by TestMu AI!š
Big week - OpenAI launched Daybreak to hunt vulnerabilities before hackers do, Anthropic shipped Agent View for Claude Code, Mira Muratiās Thinking Machines Lab took a direct shot at OpenAIās real-time AI, and Microsoftās data shows developer jobs are still rising despite AI coding tools.
In this edition, Novo Nordisk partners with OpenAI for AI drug discovery, JPMorgan bets $1.2B on AI infrastructure, DeepMindās co-mathematician cracks a 60-year-old problem, NASSCOM upskills 150K Indian developers, and Joe Colantonio unpacks the hidden Token Tax of AI testing on TestGuild.
š¬ Come across something useful or interesting? Just reply and letās exchange ideas.
News
11 min
openai.com
š”ļø OpenAI launched Daybreak on May 11, a new cybersecurity initiative that pairs GPT-5.5 with Codex Security to help organizations find and fix vulnerabilities before attackers do. Cloudflare, Cisco, CrowdStrike, Oracle, Palo Alto Networks, and Zscaler are already integrating the technology. Daybreak is OpenAIās direct answer to Anthropicās Project Glasswing and Claude Mythos, which helped Mozilla patch 271 Firefox vulnerabilities in April.
07 minclaude.com
š¤ Anthropic (@AnthropicAI) shipped Agent View for Claude Code in research preview on May 11. Instead of juggling multiple terminal tabs and tmux windows, developers can now manage all parallel Claude agent sessions from a single command-line dashboard. Launch with claude agents, send sessions to the background with /bg, and only jump back in when Claude needs your input. Available on Pro, Max, Team, Enterprise, and API plans.
08 minstartuphub.ai
š¬ Mira Muratiās Thinking Machines Lab published its first technical post, arguing that the turn-based approach used by OpenAIās Realtime API and Googleās Gemini Live is fundamentally wrong. The lab introduced TML-Interaction-Small, a 276 billion-parameter model where listening, speaking, seeing, and pausing are trained in, not bolted on. On cued response timing benchmarks, GPT Realtime-2 scored 2.9% while TML scored significantly higher.
07 minblogs.microsoft.com
š Microsoft published its latest Global AI Diffusion Report, showing AI usage reached 17.8% of the worldās working-age population in Q1 2026, up 1.5 percentage points. Git pushes increased 78% year-over-year globally. Total U.S. software developer employment reached approximately 2.2 million in 2025, rising 8.5% year over year, a record high. March 2026 data shows employment is still about 4% higher than the previous year.
07 mincnbc.com
š Novo Nordisk CEO Mike Doustdar announced a strategic partnership with OpenAI to integrate AI across drug discovery, manufacturing, and supply chains. Sam Altman said the collaboration will accelerate scientific discovery and redefine the future of patient care. Pilots launch across R&D and manufacturing now, with full integration by the end of 2026. Doustdar told Reuters the goal is to supercharge scientists, not replace them, as Novo races Eli Lilly for weight-loss drug dominance.
08 minartificialintelligence-news.com
š¦ JPMorgan Chase is pushing its 2026 technology budget to roughly $19.8 billion, with $1.2 billion earmarked for AI initiatives. CFO Jeremy Barnum confirmed the $2 billion year-over-year increase during the bankās company update. The spending targets customer service automation, personalized client insights, and developer tooling - moving AI out of experimental R&D and into the same category as cybersecurity and core risk controls.
AI
07 minmindstudio.ai
š” The MindStudio Team breaks down the five features Anthropic shipped at its Code with Claude event across San Francisco, London, and Tokyo. Dreaming reviews past sessions and self-improves agents over time. Outcomes enforce output quality without human review. Multi-agent orchestration lets agents coordinate on complex jobs. Claude Finance ships with 10 pre-built agents. And Add-ins enable enterprise deployment.
12 minnewsroom.ibm.com
š IBMās Institute for Business Value, working with Oxford Economics, surveyed 2,000 CEOs across 33 geographies. The headline: 76% of organizations now have a Chief AI Officer, up from just 26% in 2025. IBM Vice Chairman Gary Cohn writes that enterprises that succeed will operate AI-first - not as a layer of technology, but as a new operating model. Companies with an AI-first C-suite have scaled 10% more AI initiatives than their peers, and 64% of CEOs are now comfortable making major strategic decisions based on AI-generated input.
06 mindeepmind.google
š§® Google DeepMind unveiled the AI co-mathematician, a multi-agent system built on Gemini 3.1 that scored 48% on FrontierMath Tier 4 - a new high among all AI systems. Oxford topologist Marc Lackenby used the system to solve Problem 21.10 from the Kourovka Notebook, a group theory question that had remained open since 1965. Unlike single-prompt chatbots, the system runs parallel workstreams with specialized agents for literature search, proof verification, and computational exploration.
11 minanalyticsinsight.net
Ankit Bose, Head of AI at NASSCOM, announced AI Code Sarathi - a national initiative to upskill 150,000 Indian developers across corporate, startup, and academic sectors. The program runs in three phases: Explore (AI-assisted coding workshops), Engineer (mentorship and capability building), and Excel (a national Agentic AI Hackathon with real-world problem statements). Aligned with the IndiaAI Mission, top performers receive blockchain-backed NASSCOM certificates and get recognized at AI Confluence in July 2026.
Automation
08 minclaude.com
āļø Anthropic launched Code Review for Claude Code - a multi-agent system that dispatches a team of reviewers on every PR to catch bugs that quick reviews miss. Running internally at Anthropic for months, the system has generated findings on 84% of large PRs (1,000+ lines), averaging 7.5 issues per PR, with less than 1% marked incorrect. Available in research preview for Team and Enterprise users.
09 mingithub.com
āļø Jaktestowac released an open-source repo packed with 70+ custom instructions, agents, and prompt templates for test automation using GitHub Copilot. Covers Playwright E2E test generation, API testing from OpenAPI specs, accessibility audits with WCAG 2.2 mapping, tech debt scanning with severity-ranked reports, and a multi-agent orchestrator that designs, implements, reviews, and verifies tests in one loop. Works with Copilot, Claude Code, Cursor, and Windsurf.
Tools
12 minclaude.com
š ļø Anthropic launched the Claude Platform on AWS, bringing the full Claude API to Anthropic-managed infrastructure accessible through AWS, with AWS billing and IAM authentication. Access the Messages API, Files API, Message Batches API, Claude Managed Agents, Agent Skills, code execution, and tool use through native AWS endpoints. Claude Opus 4.7, Sonnet 4.6, and Haiku 4.5 are all available.
07 mingeekwire.com
š ļø AWS updated Kiro, its agentic AI IDE, with a new Requirements Analysis feature that uses neurosymbolic AI to mathematically verify software specs before any code is written. Kiro turns natural language prompts into structured requirements in EARS notation, generates architectural designs, and breaks implementation into discrete tasks. New additions include Parallel Task Execution (75% faster builds), Quick Plan for skipping step-by-step approvals, agent hooks that trigger on file save, and a CLI for terminal-first developers.
Video & Podcast
06 mintwit.tv
šļø Leo Laporte hosts the latest This Week in Tech (May 10), covering the biggest stories of the week: which religion does AI identify with, the AI cybersecurity race between OpenAI Daybreak and Anthropic Mythos, and the growing debate around AI agent autonomy in enterprise environments. A packed episode for anyone tracking how AI is reshaping the industry.
12 mintestguild.com
šļø Joe Colantonio sits down with Arthur Hicken on TestGuild (May 6) to unpack the growing problem of runaway AI token usage in testing pipelines. AI-powered testing tools are exploding - but so are the hidden costs. They cover why teams are underestimating LLM billing, how runaway AI agents and infinite loops happen in CI/CD, real-world examples of massive token bills, and why deterministic problems should not always use LLMs. Essential listening for QA teams deploying AI agents at scale.
Events
09 minio.google
š¤ Google I/O 2026 is just five days away. Join online on May 19ā20 for livestreamed keynotes and sessions. Expect deep dives on Gemini updates, Android, AI developer tooling, and the future of Googleās agentic AI ecosystem. Free to attend.