Andre Pemmelaar

#ai #ai-memory #identity-layers #private-ai
Chief Builder at

[tarunyaa @tarunyaa_]

The next scaling law is multi-agent swarms. Mixture of models.

[Andre Pemmelaar @AndrPem]

@vasuman Me favourite line:
“because building can be done by vibe-coders but fixing cannot…” so true!

[Andre Pemmelaar @AndrPem]

@MarcelVelica @gregpr07 It definitely did work through my regular browser and that was impressive. But I had to give it lots of guidance on a simple Linkedin task and update the skills.md despite using Opus 4.7

[Andre Pemmelaar @AndrPem]

@buckeyevn @gregpr07 🎯🎯 Spot on analogy. I was cringing as I watched it struggle through a simple LinkedIn task

[Andre Pemmelaar @AndrPem]

For those who hate reading raw markdown...
I read this → ran /last30days → /ce:plan (@every + @kieranklaassen) → shipped view-md in one session w/@usemonologue

Beautiful markdown viewer, live reload, no install needed.

npx claude-md-viewer file.md
https://t.co/TuHnHRNrVe

[Matt Van Horn @mvanhorn]

https://t.co/AwabaZTtdp

[Andre Pemmelaar @AndrPem]

@rubenpdegroot @conductor_build This was super useful. Would love to hear more about your self documenting progressive discovery system. Particularly the progressive discovery.

[Andre Pemmelaar @AndrPem]

@naveennaidu_m @usemonologue @danshipper I have marked two examples with thumbs down that have this issue.

[Andre Pemmelaar @AndrPem]

I think we can say @usemonologue have successfully hit product‑market fit with me. After using it for 3 weeks I've reached a ranking of 10 in terms of total words generated. I can't imagine using Claude Code in any other way now. @danshipper @naveennaidu_m

[Andre Pemmelaar @AndrPem]

@naveennaidu_m @usemonologue @danshipper One small nitpick. I cannot get it to transcribe things like:
1. something something
2. Something

I get:
Two, Dot, what are all the health check failures being caused by...

[Andre Pemmelaar @AndrPem]

@naveennaidu_m @usemonologue @danshipper Intensely building a privacy-first personal assistant on AWS, which means about a third of all the things I say to it involve complex AWS CLI commands -lots of hyphens (E.g. use AWS describe-nat-gateways) and other stuff. It does an excellent job of properly transcribing that!

[Andre Pemmelaar @AndrPem]

Was a the huge fan of his work at Every! Can’t wait to this thing comes out!

[Evan Armstrong @itsurboyevan]

Fewer than 10,000 people are shaping the future for the other 8 billion of us. I think you should get to know them.

They’re founders, investors, researchers, and creatives—the rare few in tech who are disrupting the present to build something better. I talk with @bryce, @Altimor, @So8res, with some REALLY great guests to come.

The Leverage Podcast asks the most urgent questions in technology, while also giving you a closer look at the people making those choices.

Honestly, I thought starting a podcast was a dumb idea. Does the world really need another one? But it turns out I needed these conversations too. The first recordings have already shifted how I see the world, and I’m hopeful they’ll do the same for you.

It all goes live September 10th. Teaser below :)

[Andre Pemmelaar @AndrPem]

Local-OpenAI: Run OpenAI GPT-OSS 20B locally in a few clicks!
I am sure many of you have heard that OpenAI has released GPT-OSS, their first open-source models

This repo takes care of the downloads, setup and comes with a streaming GUI -> https://t.co/cPbZC1jCcZ
#OpenAI

[Andre Pemmelaar @AndrPem]

@Box @MatthewBerman @levie We’re already building this -Mygentic.ai

[Ruben Hassid @rubenhassid]

BREAKING: Stanford just surveyed 1,500 workers and AI experts about which jobs AI will actually replace and automate.

Turns out, we've been building AI for all the WRONG jobs.

Here's what they discovered:

(hint: the "AI takeover" is happening backwards)

[Andre Pemmelaar @AndrPem]

2/2 what most people don’t realize is that AI stores much of its data in vector embeddings that capture the semantic meaning and context. These essentially look like a long row of numbers in a spreadsheet, and to the human eye they seem indecipherable. In fact, you can write a 15 line piece of code that will turn those embedding back into text that exposures the original data. This isn’t sophisticated decryption. This is just simple AI coding. You could write it with an AI coder and 10 minutes.

[Andre Pemmelaar @AndrPem]

1/2 Great thread on how AI is changing the security guarantees of chat apps. Couldn’t agree more. Third-party AI apps that analyze your message flow but don’t have a privacy first architecture are making your communications extremely vulnerable.

[Andre Pemmelaar @AndrPem]

1/2 Great thread on how AI is changing the security guarantees of chat apps. Couldn’t agree more. Third-party AI apps that analyze your message flow but don’t have a privacy first architecture are making your communications extremely vulnerable.

[Andre Pemmelaar @AndrPem]

1/2 Great thread on how AI is changing the security guarantees of chat apps. Couldn’t agree more. Third-party AI apps that analyze your message flow but don’t have a privacy first architecture are making your communications extremely vulnerable.

[Matthew Green @matthew_d_green]

It is insane how scary the threat models of encrypted messaging apps providers are.

[Andre Pemmelaar @AndrPem]

I have to agree with @fseixas that these apps will in the next few years disappear as we see personalized orchestrator/meta agents appear. These meta agents will likely track all of your personal information and life, store in some sort of privatized way and act as an interface to the rest of the AI world. Not happening in the next 12 months, but they are on their way.