AJ Learning HubArchive
Big releases, trends, and shifts across GHL, Claude, and AI automation. This is a snapshot archive — last updated June 2026, not a live feed.
New (June 24) workflow triggers fire when a conversation SLA is 'due soon, overdue, or manually dismissed,' plus admin permission controls. Build escalation automations that auto-route or notify a rep before a lead goes cold. Source: https://ideas.gohighlevel.com/changelog
HighLevel shipped 7 prebuilt workflow templates (June 20) for Asana, ClickUp, Slack, and Google — drop-in starting points for cross-tool automations instead of building from scratch. Pairs well with the upgraded Workflow AI step (structured JSON output + test mode). Source: https://ideas.gohighlevel.com/changelog
Anthropic introduced Claude Tag (June 23), a new offering for collaborating with Claude across a team — relevant as you move from solo Claude use toward shared client workspaces. Source: https://www.anthropic.com/news
As of mid-2026, n8n's rebuilt AI Agent node does tool-calling across Claude, GPT, Gemini, and Groq with persistent memory (Redis/Postgres) and per-node retries — the strongest option for complex multi-step LLM automations. Zapier still wins for fast 2–3 step Zaps; Make sits in between visually.
Make's Maia assistant builds scenarios from a plain-English description and Zapier Agents run autonomous tasks across 8,000+ apps. The shared pattern: describe the outcome, let AI draft the workflow, then you refine and harden it. Treat AI-drafted flows as first drafts, not production.
Anthropic released Claude Fable 5 on June 9 — its most capable widely released model (1M-token context, always-on adaptive thinking), generally available on the Claude API, Bedrock, Google Cloud, and Microsoft Foundry. Claude Mythos 5 shares its specs but is not self-serve: it is limited to approved customers under Project Glasswing for defensive cybersecurity work. Source: https://platform.claude.com/docs/en/about-claude/models/overview
Frontier capability is increasingly gated behind access tiers rather than price alone — Claude Mythos 5 is invitation-only under Project Glasswing, while Fable 5 is generally available. Design automations with a fallback model in mind so a tier you can't get into never blocks production. Source: https://platform.claude.com/docs/en/about-claude/models/overview
June brought a wave of frontier releases — Anthropic's Claude Fable 5 (Jun 9), Zhipu's GLM-5.2 (Jun 16), plus coding-focused models from Moonshot (Kimi K2.7 Code) and Cohere. Capability gaps between labs keep narrowing — pick models per task, not by loyalty. Source: https://llm-stats.com/llm-updates
Claude Code keeps expanding agentic capabilities — better long-running tasks, richer MCP tool ecosystems, and stronger multi-file edits. Lean on MCP servers to connect Claude to your real tools (GHL, GitHub, Calendar).
GoHighLevel continues deepening AI inside workflows — Conversation AI and Content AI for first-touch responses. Pair trigger links with custom fields to segment leads the moment they engage.
n8n, Make, and Zapier all ship first-class AI/LLM steps. Practical takeaway: drop a model call directly into a workflow to classify, summarize, or route data — no external glue code needed.
Function calling and structured output keep improving — more reliable JSON and better tool selection. For agent builders, that means fewer parsing hacks and more dependable multi-step automations.
Snapshots remain the agency superpower — build a system once (workflows, pipelines, funnels, custom values) and deploy it into a new sub-account in seconds. Standardize your snapshot to slash onboarding time.
Self-hosting n8n in Docker is the cost-effective path for high-volume workflows — no per-task fees and full control. Worth the steeper learning curve once your automation volume grows.
Archive through June 2026. Anything you add is saved to this browser only.