eric.marqueez

AI, SaaS, Automation and all that.

[eric.marqueez @Eric_lluhen]

AI is moving so fast that humans cannot even judge the output properly anymore.

That means average work is about to look elite.

And a lot of people are about to confuse “polished” with “good.”

[eric.marqueez @Eric_lluhen]

The dangerous person in the AI era is not the person who cannot use AI.

It is the person who cannot tell when AI gave them complete nonsense.

[eric.marqueez @Eric_lluhen]

OpenAI saying AGI could be here by the end of the year is cool.

But your company still has five dashboards showing five different revenue numbers.

Let’s maybe fix that first.

[eric.marqueez @Eric_lluhen]

The “AI agent economy” is becoming:

agent accesses system
agent touches production
agent does something weird
human opens Slack at midnight

Very autonomous.

[eric.marqueez @Eric_lluhen]

Vibe coding made it easier to build impressive products.

It also made it easier to build a beautiful app that has no users, no retention, no security, and no clue how it makes money.

[eric.marqueez @Eric_lluhen]

Every AI demo looks like the future.

Every production deployment looks like permissions, error logs, token bills, weird edge cases, and one exhausted person trying to keep it alive.

[eric.marqueez @Eric_lluhen]

The biggest vibe-coding reality check:

You can build a product in one night.

You cannot build distribution, trust, support, retention, or a reason to exist in one night.

[eric.marqueez @Eric_lluhen]

AI makes software easier to build.

Great.

Now everyone gets to learn that building the product was never the hard part.

Getting someone to care enough to pay is.

[eric.marqueez @Eric_lluhen]

Companies are building agents that can browse the web, write code, book meetings, and buy stuff.

Then the agent hits “create account” and loses the fight to a password field.

[eric.marqueez @Eric_lluhen]

AI agents signing up for products is going to expose how bad most signup flows are.

If your onboarding only works when a confused human clicks 14 buttons, you do not have a product.

You have a ritual.

[eric.marqueez @Eric_lluhen]

Mistral’s Agentic Search is basically teaching AI how normal people work:

search
open
read
realize you missed something
search again

Somehow this is revolutionary now.

[eric.marqueez @Eric_lluhen]

The next AI moat is not the model.

It is knowing where the actual answer lives inside your company’s 40-year-old mess of docs, spreadsheets, Slack messages, and Salesforce fields.

[eric.marqueez @Eric_lluhen]

“Our AI is connected to all company knowledge.”

Okay.

Ask it what is on page 47, inside the table, under the footnote.

Then watch it panic.

[eric.marqueez @Eric_lluhen]

Most company AI projects are not failing because the model is dumb.

They fail because the model gets one random paragraph from a 200-page PDF and starts speaking like it read the whole thing.

[eric.marqueez @Eric_lluhen]

Mistral just gave AI five tools to read documents properly: search, open, navigate, read, grep.

Crazy that we called the old version “enterprise AI” when it was basically guessing from three random PDF chunks.

[eric.marqueez @Eric_lluhen]

AI is making developers faster.

It is also making it way easier to create bugs nobody understands because nobody actually read the code.

[eric.marqueez @Eric_lluhen]

“We use AI to write code faster.”

Cool.

Who is using AI to make sure you did not just ship a faster security problem?

[eric.marqueez @Eric_lluhen]

The funniest AI story this week:

AI helped write a security bug.
Another AI found it and exploited it.

Now imagine this happening inside a company with 400 AI agents running.

[eric.marqueez @Eric_lluhen]

AI is not “just a chatbot” anymore when the company behind it can be worth more than almost every public company on earth.

[eric.marqueez @Eric_lluhen]

Anthropic potentially going public at $2T is insane.

People still think AI is a feature inside SaaS.

The labs are becoming the infrastructure SaaS is on.