Lesson · Nov 2026

AI for
Entrepreneurs

Use AI for everything. Verify everything.
Point it at hard problems.

Federico Rampazzo
Federico Rampazzo
apiplant.com ↗
AI
✦ write ✦ present ✦ code ✦ video ✦ sell
A small confession

Structure Wording Visuals Checked, fixed and made mine by a human.
Show of hands: if a startup showed you a deck like this,
would you think they're serious?
The signal is gone

Polish no longer
proves quality.

Real company. 3 years of work.
Made this weekend. 20 minutes.
Both ways
Judging others: look past the polish.
Building yours: polish is the minimum,
not the difference.
The one thing to remember

Use AI for everything —
or get left behind.

01

What's happening to work

02

Is it a bubble?

03

What it means for founders

04

Live: one agent, everything

01 · What's happening to work

If you do it on a computer,
AI can do it.

Being replaced right now
Customer supportCopywritingTranslationJunior designData entryJunior coding

Nobody hires juniors anymore. And the rest of office work is next.

Emails
Spreadsheets
Documents
CRM updates
Meeting notes
Reports
Office jobs now · manual jobs next

Robotics is the
new gold rush.

👁 Vision models already understand a room.
⚙️ Run them cheaply on a robot → manual work automates.
📦 Warehouses first. Then everything else.
SV SZ
Top talent is moving from Silicon Valley to Shenzhen — betting everything on robots.
object: boxconf 0.98 ▮▮▮▮▮ grip: okt+0.24s
When office + manual work automate

Cheap work →
cheap everything.

The biggest cost in most products is people's time.

Great as a consumer. Brutal if you sell your time — or a product whose price falls to zero.

todayautomation → wages prices of almost everything
The transition phase

Right now, the edge is yours.

You + AIwins today
>
AI alonemost work
>
You alonefalling behind
the window is closing — learn it now
02 · Is it a bubble?

Probably
a bubble.

Doesn't matter.

Yes, I think there's a bubble. No, it doesn't change what you should do.

$$$
Why I think it's a bubble

The big jumps
are over.

Raw intelligence of top models has more or less peaked.

So labs specialise — and smaller, cheaper models beat yesterday's frontier.

Transformers are good enough to automate a big chunk of work. Replacing human judgement needs new research, not just more hardware.
20202026 intelligence wowwow! plateau · "nice." 3D modelling movie scripts small & cheap
Meanwhile, look at the money

The gold rush is eating the hardware.

Consumer RAM
~0×

64GB kit: ~€170 in mid-2025 → €926 today. Data centres are buying all the memory.

Micron
Crucial

Shut down its consumer brand to sell to AI companies instead.

Nvidia · 2026
0

New consumer GPU generations — a first in about 0 years.

Why it doesn't matter

Even if it bursts,
AI stays.

Worst case: OpenAI and Anthropic disappear tomorrow morning.

Chinese labs have released open models, close to the frontier and good enough for most uses.

Open = download it, modify it, host it anywhere. You can't un-invent that.

Not hype. Just logic: adapt, or be left behind.
open
OpenAI
Anthropic
DeepSeekV4.1 Flash 1017
GLM5.3-Flash · Zhipu
KimiMoonshot
QwenAlibaba
MiniMaxMiniMax
03 · What it means for founders

The bar went up.
For everyone.

Bad news first. No sugar-coating. Then the good news.

Bad news #1

Pressure for even smaller teams.

10people a startup used to need
3what investors now expect
1founder + good AI tools. Lean is the default.
Bad news #2

Everyone can build →
too much of everything.

AI writes the code, so technical people got cheaper. Anyone can build — so everyone does.

10apps in your niche
50then…
200oversupply, every sector
Bad news #3 · more inequality

A few win big.
Most earn nothing.

Products are easy to copy and the internet lets the winner reach everyone. Winner takes most.

winnerseveryone else
Bad news #4 · buyers are drowning

Your inbox.
Every buyer's inbox.

Flooded with "vibe-coded" products. Every email well written. Every landing page beautiful.

Polish means nothing now. Same for consumers: infinite apps, less chance they pick yours.

Inbox1,284 unread
Bad news #5 & #6

Distribution is the
new bottleneck.

Building is cheap. Getting noticed is expensive.

First 100 users, first 10 customers: now a huge budget and a huge effort.

The bar is raised for everything. Product, pitch, content, support. "Good enough" is now average — average is free.
PAYING CUSTOMERS PRODUCTS LAUNCHED TODAY
Put it all together

Random SaaS?
Not worth your time.

Another to-do app

cloned this weekend

A CRM for some niche

given away free

A thin ChatGPT wrapper

already exists ×200

Sounds depressing. But it isn't the end of the story.

04 · The good news

Go where
it's hard.

If easy things are worthless, hard things are worth more than ever.

⚡ Energy🧬 Biotech🤖 Robotics🏭 Manufacturing🩺 Healthcare🚚 Logistics

Domain knowledge · hardware · regulation · private data · real relationships.
The only place left where competition isn't infinite.

The biggest, least obvious opportunity

Most companies run on knowledge in people's heads.

No manual. No SOP. Just people who know.

You can't automate what isn't written down. Years of interviews, mapping, change management, resistance.

Maria in accounting knows which supplier always pays late.
The warehouse manager knows the workaround for the broken system.
SOP.md
404
not written down
The other path

Or… they get replaced.

Incumbent · hundreds of people
→
+
Industry insiders + forward-deployed AI engineers

Engineers who sit inside the business, learn the work hands-on, and build the automation right there.

That lean startup could be you.
pick an industry → learn how it really works → team up with insiders → build the AI-native version
05 · Use AI for everything, verify everything

Everything. I mean everything.

Write

Give context and examples. Let it draft, you edit. Then ask it to critique like a tough professor.

Present

Notes → outline → argue with the outline → slides. You're looking at one.

Video

Scripts, storyboards, voiceovers, subtitles, editing — and now the footage itself.

Market & sell

Personas, competitors, positioning, 10 ad variants. Role-play your toughest customer.

Code

Even if you've never written a line. Landing pages, prototypes, small automations.

Last month, in practice
apiplant.com/open-source ↗

My AI Software Factory.

Ported generative-AI inference engines — video, voice cloning, text analysis — to my favourite language. For each one, independently, it:

Built
the UI

Packaged it
for every OS

Ran automated
tests

Benchmarked &
beat the original

Wrote docs &
launched the site

weeks of senior engineering→a couple of hours.
The most important warning of the day

Trust,
but verify.

AI is not infallible. It hallucinates: stats, quotes, court cases, papers, links.

With total confidence. It doesn't know it's wrong.

market-research.md
made up
"According to a 2024 Harvard study, 87% of founders who use AI reach profitability in year one."
My rules

AI does the first draft. You're the editor.

1

If it cites a source, open the source. Check it exists and says what the AI claims.

2

Check every number that matters.

3

If it writes code, run it and test it. Don't assume it works.

4

Read everything before it goes out with your name on it. If it's wrong, it's on you.

Trick: ask a different AI to find the mistakes.
Still be good at your job — AI just made you 10× faster.
06 · Tools

Climb the ladder.

You don't need to spend money to start.

1Gemini & ChatGPT, free plans$0Writing, research, ideas, images. Turn off training in settings.
2OpenCode + free models$0GLM-5.3-Flash & DeepSeek V4.1 Flash 1017 for free. Coding and agents.
3GLM-5.3-Flash · DeepSeek V4.1 Flash 1017cents / M tokensBest value. Data goes to China — host elsewhere if sensitive.
4Claude Pro / Max · Opus 5.5$20–200 / moDecks, websites, video pipelines. A real quality edge.
5Local AI on your hardwarethousands upfrontNot yet. RAM costs ~5× what it did. Wait.
LIVE

One agent.
Five jobs.

AInstall the agent BEdit text in Markdown CMake a presentation DA live website on Cloudflare EVideo with Seedance 2.5

    
AI for Entrepreneurs
click / → next · right-click / ← back · F fullscreen