We reduced our cloud bills from +$7,789.45 → $2,115.11 (on average).
But most importantly:
The less we depend on specific vendors, the less locked-in we become. +80% of our infrastructure can now be hosted practically anywhere.
Biggest misconceptions I see:
❌ Self-hosting is extremely complex.
Reality: It's less complex and "scary" than you think.
❌ AWS, Azure GCP make hosting easy.
Reality: They can get very complex too. And large bills can hit hard and unexpectedly.
AI is still being oversold.
And it's such a shame.
They tell you it will:
- Think for you
- Run everything for you
- Make money for you
It's obviously not true.
And people had it enough of this...
And as a result, you might go 180 and think: "AI seems like useless hype"
I'm telling you: It's not.
You can do INCREDIBLE things with AI.
But you need the following:
→ A human orchestrator.
→ The right lineup of tools.
Here is the breakdown 👇
🔸 Postman
API discovery and agent-builder.
🔸 Aidbase
UI-based RAG and AI support.
🔸 n8n
Automation and Agentic workflows.
🔹 GitHub Copilot
Agentic AI coding
🔹 Cursor
Full AI-Assisted IDE.
🔹 Lovable
Text-to-prototype in minutes.
✨ Social content and blogs:
A combination of n8n, fine-tuned models, and FeedHive's AI post-prediction workflow.
You can create this setup for yourself too!
Right now, you can get LIFETIME access to FeedHive, Aidbase, and +3 other tools for a single one-time purchase.
Whenever a new tool launches, people go "RIP existing tools" (especially if it's from OpenAI).
But they're missing the point.
The field of ops and AI automation is WAY too big to be covered by one single tool.
That's why Agent Builder isn't killing n8n.
That's why there's room for tools like Lindy.
That's why Zapier and Make still thrive.
These tools don't compete.
Instead, you have the option to combine.
Do this - and you can build the most powerful AI ops infrastructure imaginable.
✨ Agent Builder + n8n
Build a chat interface with Agent Builder.
Have it delegate to specialized agents in n8n.
Request in natural language → Complex solution.
✨ Aidbase + GitHub Copilot
Build a user-facing AI chat in Aidbase.
Pass bug reports to GitHub Copilot agents.
Get PRs with bugfixes proactively.
✨ Let AI mix and match.
Build an AI manager in Agent Builder.
Give it access to an n8n MCP server.
Let it mix and match between a broad list of tools and specialized workflows.
At this point, you don't need more people.
You need a better understanding of what AI ops infrastructure can do.
→ MVP in 48 hours
Develop an MVP fast using Cursor and Lovable.
→ Landing page & ads
Set up a landing page run ads on Google and Facebook. Create eye-catching creatives to grab attention.
→ Test for interest
Monitor the number of sign-ups you get. Use a tool like Smartlook to record sessions. See how early users are using the tool.
⛔ If no one shows interest - scrap it.
✅ If users seem to use it like you intended it (even if the MVP isn't suited to get real results yet) - double down and improve fast.
Total time spent: <50 hours
Total money spent: <$200
Keep going until you hit something.
Ps. If you liked this post, you'll enjoy my weekly SaaS newsletter.
Join 30K+ SaaS founders, indie hackers, and entrepreneurs. Lightweight, easy-read tips on SaaS, product, and growth - directly to your inbox every week.
⛔ How we used to do it:
- Plan a new feature.
- Draft the design in Figma.
- Prototype it in Figma.
- Hand it over to a developer.
- Developer manually codes it.
✨ How we do it now:
- Vibe coding directly in Lovable.
✅ Result:
Planning, drafting, designing & prototyping ALL happen in one single place. And the artifact is REAL application code, not just a slideshow.
Don't let people tell you AI can't code.
It can. Let me show you 👇
Simon Høiberg
Simon Høiberg
There's nothing wrong with building an AI wrapper.
But you cannot keep building the same simple ChatGPT wrapper everyone else already built.
You need to go much more niche.
Fortunately, there are tons of small, specific problems that can be solved with specialized, fine-tuned AI models.
- You can find them on Hugging Face.
- You can host them on Replicate.
- You can build it with Lovable.
It's still a wrapper.
But it's specific, niche, and targeted at a small group of people willing to pay.
Let me show you what I mean.
Here are 8 ideas you could build today 👇
2 days ago | [YT] | 298
View 1 reply
Simon Høiberg
This year, we started moving off "the cloud" - 80% of our infrastructure now runs on bare-metal servers.
AWS → Hetzner.
Instead of:
- DynamoDB
- ElastiCache
- ECS/Fargate
- SQS
- SNS
- CloudWatch
- X-Ray
We use:
- PostgreSQL
- Redis
- Kubernetes
- RabbitMQ
- BullMQ
- Grafana
- Prometheus
We reduced our cloud bills from +$7,789.45 → $2,115.11 (on average).
But most importantly:
The less we depend on specific vendors, the less locked-in we become. +80% of our infrastructure can now be hosted practically anywhere.
Biggest misconceptions I see:
❌ Self-hosting is extremely complex.
Reality: It's less complex and "scary" than you think.
❌ AWS, Azure GCP make hosting easy.
Reality: They can get very complex too. And large bills can hit hard and unexpectedly.
Here's the breakdown 👇
4 days ago | [YT] | 372
View 13 replies
Simon Høiberg
AgentKit + FeedHive MCP 🤯
Being first to the conversation after the breaking news hit gives you a HUGE advantage on social media.
- It creates authority
- It creates trust in your brand
- It drives business
But keeping up with the constant stream of breaking news (especially in AI) is exhausting.
But I found a way to solve this.
I built a workflow using:
✨ OpenAI AgentKit
✨ FeedHive MCP
This allows you to:
→ Scout for breaking news
→ Put together a news report
→ Write a post for social media
But not just any post!
With FeedHive's new MCP, your can write a post that perfectly fits your brand:
→ Using your tone of voice
→ Reflecting your brand identity
→ Highlighting what's important to your people
I made a step-by-step walkthrough.
Here's how you can set this up for yourself.
Let me show you 👇
1 week ago | [YT] | 44
View 0 replies
Simon Høiberg
This AI agent plans my daily todo.
Here's how it works 👇
I always have a loooong todo list.
All kinds of things of varying importance/urgency.
My AI agent runs every morning:
- Looks through the list
- Filters it
- Prioritizes it
- Adds to it (if needed)
It uses 4 essential checkpoints:
⚠️ Critical
Checks our support channel on Discord.
Checks our system monitoring in Grafana.
Checks Aidbase for critical bug reports.
📅 Work
Checks my emails.
Checks my calendar.
Checks my Notion workspace.
👤 Personal
Checks my personal goals.
Checks my "watch later" lists.
Checks my "read later" lists.
❤️🩹 Health
Checks my Whoop app:
- Sleep score.
- Recovery score.
- General health state.
Now the AI has everything it needs to create:
- 3 important todo items.
- 3 optional todo items.
- 1 workout program.
- 1 daily learning.
At 5.30 when I get out of bed, this todo list has already been curated for me - ready in Todoist.
AI is a wonderful technology.
2 weeks ago | [YT] | 200
View 8 replies
Simon Høiberg
AI is still being oversold.
And it's such a shame.
They tell you it will:
- Think for you
- Run everything for you
- Make money for you
It's obviously not true.
And people had it enough of this...
And as a result, you might go 180 and think: "AI seems like useless hype"
I'm telling you: It's not.
You can do INCREDIBLE things with AI.
But you need the following:
→ A human orchestrator.
→ The right lineup of tools.
Here is the breakdown 👇
🔸 Postman
API discovery and agent-builder.
🔸 Aidbase
UI-based RAG and AI support.
🔸 n8n
Automation and Agentic workflows.
🔹 GitHub Copilot
Agentic AI coding
🔹 Cursor
Full AI-Assisted IDE.
🔹 Lovable
Text-to-prototype in minutes.
✨ Social content and blogs:
A combination of n8n, fine-tuned models, and FeedHive's AI post-prediction workflow.
You can create this setup for yourself too!
Right now, you can get LIFETIME access to FeedHive, Aidbase, and +3 other tools for a single one-time purchase.
Go to www.founderstack.pro/ to see how.
3 weeks ago | [YT] | 200
View 5 replies
Simon Høiberg
AI ops tools combine.
They don't compete.
Whenever a new tool launches, people go "RIP existing tools" (especially if it's from OpenAI).
But they're missing the point.
The field of ops and AI automation is WAY too big to be covered by one single tool.
That's why Agent Builder isn't killing n8n.
That's why there's room for tools like Lindy.
That's why Zapier and Make still thrive.
These tools don't compete.
Instead, you have the option to combine.
Do this - and you can build the most powerful AI ops infrastructure imaginable.
✨ Agent Builder + n8n
Build a chat interface with Agent Builder.
Have it delegate to specialized agents in n8n.
Request in natural language → Complex solution.
✨ Aidbase + GitHub Copilot
Build a user-facing AI chat in Aidbase.
Pass bug reports to GitHub Copilot agents.
Get PRs with bugfixes proactively.
✨ Let AI mix and match.
Build an AI manager in Agent Builder.
Give it access to an n8n MCP server.
Let it mix and match between a broad list of tools and specialized workflows.
At this point, you don't need more people.
You need a better understanding of what AI ops infrastructure can do.
3 weeks ago | [YT] | 177
View 3 replies
Simon Høiberg
n8n is a free SaaS idea catalog.
There is an arbitrage opportunity that can take you from idea to profitable SaaS with near-zero risk and a very low budget.
Here's how it works:
1️⃣ Browse n8n workflows
Stop "imagining" what your SaaS is going to be about. Look for real problems people are solving with n8n.
Look at n8n's website.
Search on YouTube.
Find 1-2 popular workflows.
2️⃣ Break it down
Essentially, an n8n workflow is a problem that's being solved. Break it down.
The basic mechanism.
The outcome.
3️⃣ Open Lovable
n8n is awesome. But it's not for everyone. Recreate the idea behind the workflow in a much more user-friendly way.
Vibe code the front end.
Ask AI to analyze the workflow.
Recreate the backend.
4️⃣ Charge
Put your solution behind a paywall using Stripe checkout.
5-minute implementation.
Start charging from day one.
Now start marketing your solution.
- Run ads.
- Use social media.
- Use cold outreach.
Worst-case: If this doesn't work, you've spent a few hundred dollars and some hours of work.
Best-case: You'll have a SaaS solving a real problem for real users paying real money.
4 weeks ago | [YT] | 321
View 8 replies
Simon Høiberg
You need 3 things to win.
- Build a business.
- Build systems to run it.
- Stay healthy.
Here are 6 YouTube channels that will teach you exactly that.
🚀 Startups
Starter Story (@starterstory):
Real founders share how they built their businesses from scratch.
Greg Isenberg (@GregIsenberg):
Startup ideas and advice on creating products people can't resist.
⚙️ AI & Automation
Liam Ottley (@LiamOttley):
Unlock next-level productivity with practical AI workflows.
Nate Herk (@nateherk)
Tutorials and deep dives into automation and ops.
☘️ Health
Brad Stanfield (@DrBradStanfield):
Science-backed habits to upgrade your mind and body.
Bryan Johnson (@BryanJohnson):
Radical insights on living healthier and longer.
Master these, and you're unstoppable.
1 month ago | [YT] | 233
View 9 replies
Simon Høiberg
How smart people create a SaaS:
→ MVP in 48 hours
Develop an MVP fast using Cursor and Lovable.
→ Landing page & ads
Set up a landing page run ads on Google and Facebook. Create eye-catching creatives to grab attention.
→ Test for interest
Monitor the number of sign-ups you get. Use a tool like Smartlook to record sessions. See how early users are using the tool.
⛔ If no one shows interest - scrap it.
✅ If users seem to use it like you intended it (even if the MVP isn't suited to get real results yet) - double down and improve fast.
Total time spent: <50 hours
Total money spent: <$200
Keep going until you hit something.
Ps. If you liked this post, you'll enjoy my weekly SaaS newsletter.
Join 30K+ SaaS founders, indie hackers, and entrepreneurs. Lightweight, easy-read tips on SaaS, product, and growth - directly to your inbox every week.
Join here: simonl.ink/newsletter
1 month ago | [YT] | 204
View 5 replies
Simon Høiberg
AI now writes 90% of our code.
(Yes, really...)
Biggest shift:
We skip designing & prototyping.
⛔ How we used to do it:
- Plan a new feature.
- Draft the design in Figma.
- Prototype it in Figma.
- Hand it over to a developer.
- Developer manually codes it.
✨ How we do it now:
- Vibe coding directly in Lovable.
✅ Result:
Planning, drafting, designing & prototyping ALL happen in one single place. And the artifact is REAL application code, not just a slideshow.
Don't let people tell you AI can't code.
It can. Let me show you 👇
1 month ago | [YT] | 76
View 0 replies
Load more