AI Annika

AI Engineer · PhD ·
Daily AI insights, tools & trends ·
Building real AI systems ·
Sharing what actually works

🇩🇪 in 🇳🇱

[AI Annika @MildlyAnnika]

Agentic AI Brief
NIST Agent Standards Tighten, OpenAI Buys Promptfoo For Frontier Testing, Codex Security Pushes AppSec Into Agent Workflows, And ChatGPT Moves Into Excel With Market Data. Link in comments 👇

[AI Annika @MildlyAnnika]

@BettyShackelf14 @CryptoTice_ @grok @grok is this true?

[MrComputerScience @MrComputerSci]

Pithy Neural Kernel v1.1 is a useless masterpiece.

A self-organizing neural system running live in pure browser JavaScript. No backend. No libraries. No compute.

It just thinks.

↓

Free. Open. Drop the file and it runs.

#AI #MachineLearning #Programming #Tech

[AI Annika @MildlyAnnika]

AI made code cheaper.
Historically, that tends to increase demand, not reduce it.
When building becomes easier,
software expands into more parts of the business.

[AI Annika @MildlyAnnika]

AI made it easier to create software.
It also made it easier to create future maintenance work.
Fast output feels productive.
Until someone has to live with it.

[AI Annika @MildlyAnnika]

AI agents become impressive
the moment they can take actions.
They become dangerous
the moment they take the wrong one.

[AI Annika @MildlyAnnika]

In modern software development,
decisions don’t last as long anymore.

Better tools appear constantly.

Architectures shift.

Assumptions expire.

The skill isn’t choosing once.

It’s being able to change.

[AI Annika @MildlyAnnika]

Agents can take actions.

That’s impressive.

But in real systems: someone owns those actions,

someone audits them,

someone takes the blame.

AI can act.

But it still needs adults in the room.

[AI Annika @MildlyAnnika]

Decisions in software don’t last anymore.
Not because engineers got worse.
Because the environment changes faster.
New tools
new patterns
new constraints
The skill is not being right.
It’s staying right.

[AI Annika @MildlyAnnika]

AI lowers the barrier to entry.

But it raises the bar for quality.

Because now: bad systems can be built faster too.

The difference is no longer: who can build

It’s: who understands what they built

[AI Annika @MildlyAnnika]

AI increases what software developers can build.

It also increases what can go wrong.

More possibilities means more edge cases,

more interactions,

more failure modes.

Complexity grows with capability.

[AI Annika @MildlyAnnika]

AI lowers the barrier to entry.

But it raises the bar for quality.

Because now: bad systems can be built faster too.

The difference is no longer: who can build

It’s: who understands what they built

[AI Annika @MildlyAnnika]

AI increases what software developers can build.

It also increases what can go wrong.

More possibilities means more edge cases,

more interactions,

more failure modes.

Complexity grows with capability.

[AI Annika @MildlyAnnika]

In modern software development,
decisions don’t last as long anymore.

Better tools appear constantly.

Architectures shift.

Assumptions expire.

The skill isn’t choosing once.

It’s being able to change.

[AI Annika @MildlyAnnika]

Software development looks simpler than ever.
That’s the illusion.
Underneath, systems are getting harder to reason about.
More tools. More layers. More interactions.
Easy to build.
Hard to understand.

[AI Annika @MildlyAnnika]

“The people who are crazy enough to think they can change the world are the ones who do.”
— often attributed to Steve Jobs

[AI Annika @MildlyAnnika]

Software development looks simpler than ever.
That’s the illusion.
Underneath, systems are getting harder to reason about.
More tools. More layers. More interactions.
Easy to build.
Hard to understand.

[MrComputerScience @MrComputerSci]

⟶ Think your IT team can't see you using ChatGPT/Claude/Grok at work?

As someone who has worked as a sysadmin for years, the reality might shock you.

I broke it down + shared real safeguards to keep your personal AI use private at work.

Read below for free. ↓

#AI #Privacy

[MrComputerScience @MrComputerSci]

⟶ I just published a brand new article that talks about what it means when an AI company claims they "don't train on your data."

Are they lying? My article spells in out in detail including four SHADY reasons they use your data anyway lol. Details below. FREE. ↓

#AI #Privacy

[AI Annika @MildlyAnnika]

“The people who are crazy enough to think they can change the world are the ones who do.”
— often attributed to Steve Jobs

[MrComputerScience @MrComputerSci]

Most RAG systems do not break. But they DO rot. You ship with perfect retrieval. Six weeks later the "vibes" are off. The accuracy craters. This is the Silent Killer of AI infrastructure. Embeddings Drift. Here is the FREE guide on stopping the decay. #AI #RAG

Details below. ↓