Christian Tacsiat

[Julian Goldie SEO @JulianGoldieSEO]

Someone in your community just went quiet.

No message. No warning. They just vanish. 👻

I built a free AI system that catches this days before it happens — and steps in before they're gone for good.

Here's what it does:

→ Scores every member on how "quiet" they've gone

→ Sends helpful content to people slowing down

→ Flags almost-silent members straight to a human

→ Sends a personal welcome email the second someone joins

→ Runs itself every morning at 8am, no manual checking

Built entirely in n8n. Zero code.

Save this video, you'll want this system before you lose your next member. 💬

Want the SOP? DM me.

[Julian Goldie SEO @JulianGoldieSEO]

5 things I learned letting AI build my retention workflow in n8n:

→ One execution = one full run, no matter how many nodes are inside it.

→ The Assistant plans, builds, and tests — but never sees your API keys.

→ A workflow can look 100% finished and still have zero working credentials.

→ Routing by risk score beats sending everyone the same message.

→ Always verify an integration actually exists before building on it.

n8n's trial gives you 14 days and 1,000 executions to test all of this. 🧠

Save this video, you'll want to check this before your next build. 💬

Want the SOP? DM me.

[Julian Goldie SEO @JulianGoldieSEO]

The best retention automation shouldn’t feel automated.

That’s where most churn workflows get it wrong.
With n8n:

→ Track activity automatically
→ Score members by engagement
→ Send value before asking them to return
→ Personalize messages with AI
→ Escalate serious cases to humans
→ Keep a record of every action

Good automation doesn’t replace people.

It tells people exactly when they’re needed.

Save this video, you’ll want this retention system.

Want the SOP? DM me. 💬

[Julian Goldie SEO @JulianGoldieSEO]

People don't quit a community because of bad content.

They quit because nobody noticed they went quiet. 👀

So I built an n8n workflow that notices for me — and the hard part wasn't building it, it was what came after.

Here's what happened when I ran it live:

→ It pulled member data, scored everyone on activity

→ Routed people into 3 branches: watch, at-risk, critical

→ Execution #8 failed instantly on node one

→ 5 nodes had zero credentials connected

→ The fix took an hour once I knew what to check

Structure took minutes. Connecting it took the real work.

Save this video, you'll want it before your workflow breaks on you too. 💬

Want the SOP? DM me.

[Julian Goldie SEO @JulianGoldieSEO]

A support team was buried in the same calls all day.

Then an AI agent named Lucas started picking up. Here's what changed.

Lucas answers EV driver support calls in seconds, day or night, even when calls pile up.

Meanwhile a healthcare group handed their endless phone tag to a Retell agent, so staff could focus on patients instead of a ringing phone.

The trick behind it all is 3 things working together:

→ Speech turned to text, instantly

→ An LLM brain figuring out what you meant

→ A natural voice speaking the answer back

All in ~600ms. The time it takes to blink.

That tiny gap is what makes a call feel human instead of laggy.

Save this video, you'll want to build your own Lucas.

Want the SOP? DM me.

[Julian Goldie SEO @JulianGoldieSEO]

A senior care team was drowning in phone tag all day.

Then they handed the calls to an AI voice agent. Here's what happened.

Pine Park Health plugged in Retell AI to run their scheduling calls.

The AI handled the back-and-forth. The humans went back to actual patient care.

The result?
→ Scheduling satisfaction jumped 38%.

→ Staff hours got handed back.

→ Zero missed calls after hours.

The magic is 3 things working together:

•Speed (~600ms replies)

•Human-sounding voices

•Knowing when to talk vs listen

Old phone robots fall apart off-script. This one doesn't.

Save this video, you'll thank yourself later.

Want the SOP? DM me.

[Julian Goldie SEO @JulianGoldieSEO]

"It's not just chatting, it's doing the work."

That one line is the whole reason voice AI suddenly matters.

Most tools stop at talking.

Retell AI takes real action mid-call.

Here's my honest take on why that's the game-changer:

→ Function calling → books a slot, updates a record, transfers to a human, all mid-conversation

→ Knowledge base → pulls real answers from your site, auto-syncs when it changes

→ Turn-taking model → knows when to talk and when to listen

→ Drag-and-drop builder → normal people can build these, not just engineers

Even OpenAI published a full write-up on it. That doesn't happen for hype.

Save this video, you'll want to revisit this one.

Want the SOP? DM me.

[Julian Goldie SEO @JulianGoldieSEO]

You can turn one spreadsheet tab into 4 different tools now.

Same data. One prompt each. Here's what I'd build first with Sheets canvas →

✔ Pipeline board
A drag-and-drop board grouped by stage. Move a card, the row updates.

✔ Visual calendar
Plot launch dates, edit them, add events — right inside the view.

✔ Dashboard
See how people move through a process instead of reading 200 rows.

✔ Gallery view
Card-based layout you can sort and filter. Perfect for a resource library.

Bonus: it also does heat maps and whiteboards with sticky notes. 🗒️

Save this — you'll come back for the build ideas.

Want the SOP? DM me.

[Julian Goldie SEO @JulianGoldieSEO]

Ever built a Google Sheet and hated how it looked?

What if that same sheet became a real app — one sentence, no code.

Google quietly switched this on. Most people haven't opened it yet.

It's called Sheets canvas, and here's why it's different →

It's a "read-write layer" built with Gemini.

Not a chart. Not a screenshot of your data.

→ Drag a card in the canvas, the row in your sheet updates.

→ Edit the sheet, the canvas updates back.

→ It syncs both ways, in real time.

The best part? It lives as a tab inside your spreadsheet.

Sharing works exactly like a normal sheet. Same link. Same permissions. One "View data" button drops you back to the raw rows anytime.

Save this post — you'll want it the next time a spreadsheet makes your eyes glaze over. 👀

Want the SOP? DM me.

[Julian Goldie SEO @JulianGoldieSEO]

ChatGPT feels the same when you open it.

It isn't. OpenAI shipped 4 updates that change how work actually gets done.

Here's what moved 👇

→ It quizzes you (or your team) on any topic, in-chat.

→ It reads across whole Google Drive folders.

→ It books real restaurant reservations for you.

→ It suggests your next task before you ask.

The chatbot quietly became a coworker. 🚀

Save this post, you'll want all four the next time you log in.

Want the SOP? DM me.

[Julian Goldie SEO @JulianGoldieSEO]

Everyone's talking about the new ChatGPT features.

Almost no one knows which plan actually gets them.

Here's the clean breakdown 👇

✔ Interactive quizzes → all consumer + Edu plans, web and mobile.

✔ Google Drive in Library → Plus, Pro, Enterprise, Edu, Business (web first, mobile after).

✔ Restaurant bookings → all plans, on mobile, web + desktop.

✔ Homepage suggestions → eligible paid users only.

✔ Bonus: Free + Go users got the Think button for harder questions.

These are rollouts, not switches. If you don't see one yet, you're probably not doing it wrong. 🔁

Save this post, you'll want it when a feature finally shows up.

Want the SOP? DM me.

[Julian Goldie SEO @JulianGoldieSEO]

ChatGPT shipped 4 major updates in one week of August.

Most people noticed zero of them.

Here's the timeline 👇

→ Aug 10: real restaurant bookings via OpenTable, Resy + Yelp.

→ Aug 13: Google Drive lands inside the Library.

→ Aug 14: interactive quizzes, right in the chat.

→ Rolling out: personalized homepage suggestions for paid users.

The work didn't get faster. It moved into the tool. 🚀

Save this post, you'll want the dates and details handy.

Want the SOP? DM me.

[Julian Goldie SEO @JulianGoldieSEO]

Q: What's the Gemini update everyone's sleeping on?
A: Custom MCP support inside Gemini Spark.

Let me break it down. 👇

Q: What's Spark?
→ Your personal AI agent. It runs on tasks, not conversations.
→ Organises folders, builds docs from your files, works across Workspace.

Q: What's the Mac upgrade?
→ It can read + act on local files on your machine.

Q: What's MCP?
→ Model Context Protocol. An open standard that lets an agent talk to outside tools.
→ Gemini's built-in list is a menu. MCP means you're not stuck with the menu.

Q: How do I use it?
→ Gemini web app → Connected Apps → point Spark at an MCP server link.

Q: Any warnings?
→ Only connect servers you trust. Start read-only. Remove it when you're done.
That's the ceiling coming off Gemini. ⚡

Save this video, you'll set up your first Spark job this week. 🚀

Want the SOP? DM me.

[Julian Goldie SEO @JulianGoldieSEO]

"Small feature. Big difference in how fast a team moves."

That's how I feel about Gemini Notebook sharing.

Here's what changed. 👇

A notebook is a project workspace → PDFs, docs, links + YouTube videos, full context every time you open it.

Now the sharing actually works:

→ Give Viewer or Editor access by email

→ Share a link with anyone who has a Google account

→ "Allow copies" hands over sources + Studio content

→ Your chat history + notes stay private

So you can pass someone a clean research pack without your messy thinking attached.

One tip → keep chats and shareable sources in separate notebooks from day one.

Save this video, you'll build your first shareable notebook today. 🗂

Want the SOP? DM me.

[Julian Goldie SEO @JulianGoldieSEO]

When did you last actually open Gemini?

Most people type one question and close the tab.
Then it changed again.

Google just shipped a version of Gemini that works while you sleep.

Almost nobody has turned it on.

Here's what most people scroll past →

→ It runs tasks for you now, not just answers

→ It watches news, email + sites and pings you when something happens

→ It connects to 13 new outside apps

→ Notebooks hold full context so you never re-explain

→ Custom MCP lets you plug in almost any tool

The shift is simple.

You stop describing your problem and start showing it.

Save this video, you'll come back to it the day you finally set Gemini up. 🔖

Want the SOP? DM me.

[Julian Goldie SEO @JulianGoldieSEO]

Your AI agent isn't slow because the model is weak.

It's slow because of a part nobody ever names. ⚙️

The model is the brain. The harness is the body around it.

→ Claude Code is a harness around Claude

→ Codex is a harness around OpenAI's models

→ "AI agent" = model + harness. Always has been.

Open models have been everywhere for a while.

The body stayed shut. Two or three companies owned that layer.

On 13 August, DeepSeek open sourced theirs. MIT licensed. 116,000+ stars in days. 🔓

The layer that turns a model into something that actually does work is now yours to rebuild.

Save this video, you'll want it the next time an agent underperforms and you blame the model.

Want the SOP? DM me.

[Julian Goldie SEO @JulianGoldieSEO]

"Everything is a plugin."

Four words that matter more than any benchmark this month. 💡

Most agent setups are a sealed car. Great engine — but when a better one ships, you start over.

This one comes apart.

But the line nobody clicked is buried in the docs:

AGENTS.md and CLAUDE.md

Translation → it reads the instruction files your other agents already use.

Your rules. Your tone. Your projects. Your do-nots. One file.

Drop a CLAUDE.md into a folder, ask the harness something only that file could answer — it answers from the file. No re-explaining your business from scratch. ✔

Two more things worth knowing:

→ OS-level sandboxing → an agent working in a folder stays in that folder

→ Append-only session log → resume it, fork it, search it, replay it

When a challenger builds for compatibility, they're playing for adoption. 🎯

Heads up: it's v0.1. Things will break. That's normal, not you.

Save this video, you'll thank yourself when your next harness reads the same file.

Want the SOP? DM me.

[Christian Tacsiat @catacsiat]

RT @JulianGoldieSEO: Can Claude actually remember your work across devices now?

Yes. And it's a bigger deal than it sounds.

Q: What chang…

[Julian Goldie SEO @JulianGoldieSEO]

Your AI tools don't talk to each other. 🤯

That's the real reason your "automated" workflow still feels manual.

Most people keep adding tools thinking more = better. Wrong move.

What you actually need is a system that connects the tools you already have — one dashboard, one Kanban board, shared memory across agents.

That's what Agent OS does. Not a new AI. A better system wrapped around the AI you already use.

Save this video, you'll stop chasing new tools and start connecting the ones you have. 💬

Want the SOP? DM me.

[Julian Goldie SEO @JulianGoldieSEO]

Free ChatGPT just stopped feeling free. It started feeling premium. 🔥

OpenAI flipped a switch on Aug 6 and barely anyone noticed:

→ Free users now run GPT-5.6 Luna — current generation, not last year's leftovers → Unlimited text chats. The 10-message wall is gone → New Think button: one tap tells the model to reason instead of react → Simple question? Just type and send → Question with 5 moving parts? Tap Think first

That one tap is the difference between a rushed answer and one that actually fits your situation.

Save this video, you'll want it the next time ChatGPT hands you a shallow answer. 📌

Want the SOP? DM me.