Good Morning! Here's what I have for you in today's newsletter:
Anthropic introduces MCP connector for enterprise
OpenAI brings GPT-5.6 to Kiro, powering plan, build, and review workflows
ElevenLabs launches CLI v1, managing voice workflows from the terminal
Five insane skills that turn Claude Code into a PhD-level researcher
4 new AI tools worth trying today
AI TOOLS
Anthropic made enterprise-managed auth for MCP connectors generally available, centralizing authorization through an organization's identity provider so Team and Enterprise admins control access, while users get tools and data connected automatically with no individual OAuth required.
Admins now authorize MCP connectors centrally through their identity provider, rather than approving access on a per-user basis.
Users get their tools and data connected automatically once admin-level authorization is in place, skipping the individual OAuth step entirely.
This is scoped to Team and Enterprise plans, positioning it as an admin-facing control rather than a change to how individual users experience Claude.
Managing Claude access across a growing team has meant approving sign-ins one by one and hoping someone kept track of who had access to what. Centralized authorization replaces that with a single approval and a consistent permission record, which makes auditing straightforward instead of a reconstruction project.
AI MODEL
OpenAI brought GPT-5.6 into Kiro, bringing its latest model into the workflows developers already use in production to plan, build, test, and review software.
GPT-5.6 now powers Kiro's existing production workflows directly, rather than requiring developers to switch tools to access the newer model.
The integration spans the full development cycle Kiro already supports: planning, building, testing, and reviewing code.
Independent testing found GPT-5.6 Terra completing tasks in Kiro at roughly 82% lower cost, driven by Kiro's spec-driven structure.
GPT-5.6 is now inside a coding environment built specifically for long, complex development work, not just a chat window. The 82% cost reduction on real benchmarks suggests the combination of a structured spec-driven workflow and a stronger model is doing something genuinely different from dropping a new model into an existing tool and calling it an upgrade.
AI CODING
šļø ElevenLabs Launches CLI v1
ElevenLabs launched CLI v1, letting developers and coding agents manage ElevenLabs workflows directly in the terminal with well-documented commands, built-in agent skills, structured JSON output, and dry runs to preview any operation before it actually runs.
Well-documented commands and structured JSON output mean the CLI is built to be scripted and chained into larger workflows.
Built-in agent skills let coding agents call ElevenLabs functionality natively, extending voice and audio capabilities into agentic workflows.
Dry runs let anyone preview exactly what an operation will do before committing to it, reducing the risk of an unintended change to a live voice or chat agent configuration.
Managing voice and chat agent workflows has typically meant switching to a dashboard or web interface. Bringing that management directly into a CLI with structured output makes ElevenLabs workflows scriptable and chainable in a way a web UI never could be.
HOW TO AI
You save 40 links a week and read maybe two. Every project starts from zero because your research has no memory. The fix is a vault that remembers what you've already found, five skills that keep finding more, and a check that catches a confident wrong answer before it gets saved.
Skill 1: Sourcing Trends
npx skills add ZeroPointRepo/youtube-skills --skill search
Search YouTube for "ai agents tutorial," last 6 months,Ā
sorted by engagement ratio, top 10
Skill 2: The Analyst
pip install "notebooklm-py[browser]"
notebooklm login
notebooklm skill installLoad your sources, then ask for an audio overview or mind map grounded strictly in what you gave it.
Skill 3: The Pulse Check
/plugin marketplace add mvanhorn/last30days-skill
/last30days ai agentsSearches Reddit, X, Hacker News, GitHub, and more simultaneously, ranked by real engagement.
Skill 4: The Academic
curl -fsSL https://raw.githubusercontent.com/openags/paper-search-mcp/main/claude-code/SKILL.md \
Ā Ā -o ~/.claude/skills/paper-search/SKILL.mdFind recent papers on ai agents from arxiv and semantic scholar,Ā last 6 months
Skill 5: The Skeptic
Build me a skill called skeptic-check that takes a list of findingsĀ and actively tries to find a source that contradicts each one.Ā
If it survives, keep it. If not, log why in rejected.md.
Chain It Into One Command
Build me a skill called research-brief that runs all four sourcingĀ legs, loads everything into NotebookLM, runs skeptic-check, andĀ saves the result to output/ with today's date.
Then use it with one line:
/research-brief ai agentsP.S. You can access all the AI trainings, prompts and workflows if you upgrade.
Google Developers brought GitHub workflow support to Google AI Studio Build, letting developers start from an existing repo, pull changes in, push updates back out, and keep developing across environments without losing sync.
Apodex introduced Apodex 1.1, a new model family built to scale agentic intelligence for complex professional work, pairing a frontier model with an asynchronous Agent Team and an open-source research workbench called FrontierAgent.
ChatGPT introduced custom sticker packs made from any photo, idea, or inside joke, now with transparent backgrounds, shareable straight to iMessage or WhatsApp.
š¤ Grok Bot: now expanded to SuperGrok Plus, Cursor Pro+, and Teams plans, AI teammates with their own persistent cloud computer.
šØ Spline V2: a complete rebuild of the 3D editor for the agentic era, AI agent mode, WebGPU rendering, and MCP support for Claude Code, Cursor, and Codex.
š§ Gemma: Google's open model family, over 150 million downloads, run locally or explore in Google AI Studio, free under Apache 2.0.
š Zed: now supports sandboxing to restrict what agents can do while using the terminal and fetch tools, keeping AI-driven changes contained.
THATāS IT FOR TODAY
Thanks for making it to the end! I put my heart into every email I send, I hope you are enjoying it. Let me know your thoughts so I can make the next one even better!
See you tomorrow :)
- Dr. Alvaro Cintas
ā Full archive of premium guides with ready-to-use prompts
ā Structured AI courses (step-by-step, start-to-finish)
ā Every upcoming premium tutorial