Kürşat Aydemir

[Kürşat Aydemir @tlhepsh]

Running an OpenVole Server From Scratch https://t.co/7C56hnWbHk

[Kürşat Aydemir @tlhepsh]

Released OpenVole v4.3.0, has some industry first features:
- Ultra lightweight microkernel
- Distributed AI agents over its own VoleNet p2p networking
- UI-Rendering plugins (paws) - bring your own paws with their own UI.
- Serve and manage multiple isolated agents in one server

[Kürşat Aydemir @tlhepsh]

It is both. The generation part is generalization of knowledge like in ML. Generalized knowledge is not memory. But the incredible part about human brain learning is its continual learning mechanism. The retrieval part still plays role in short term memory

[Vicky Verma @Unexplained2020]

This is an interesting theory! Cognitive scientist Elan Barenholtz says memory isn't retrieval. It's generation. There is no image of your mother. 🧠

He's exploring how we might improve our own prompt design, because “memories, in some ways, aren't real.”

Barenholtz says memory isn’t what we usually think it is. Most people imagine that memories are like stored images or videos sitting inside the brain, ready to be pulled out when needed.

For example, when you try to remember your mother’s face or a moment from last summer, it feels like you’re just replaying something that was already there. But according to Barenholtz, that’s not how it really works.

[Kürşat Aydemir @tlhepsh]

How a tiny agent loop grew, over four major versions, into a self-hosted operating system for AI agents, and three unique features: VoleNet, agent server, and embedded paw apps.
https://t.co/nzUwP4wvyG

[Kürşat Aydemir @tlhepsh]

More intelligence require a breakthrough how AI perceives information. AWARE is on it.

[Elon Musk @elonmusk]

It is humbling to consider that if we harness just 1 millionth of the Sun’s power for AI, that will be much more than a million times the intelligence of all of humanity

[Kürşat Aydemir @tlhepsh]

Galactic council

[Jon Stewart @JonStewartIL]

Posted than immediately removed. 🧐

[Kürşat Aydemir @tlhepsh]

Adiyukh (Ӏэдийху), Goddess of moonlight, from Circassian Nart sagas

[Kürşat Aydemir @tlhepsh]

It is funny such huge companies just blindly push the hypes across the employees and exhausting tokens. What is the leadership doing?

[unusual_whales @unusual_whales]

Microsoft, $MSFT, has started canceling Claude Code licenses, per the Verge

[Kürşat Aydemir @tlhepsh]

Gemini now knows Kabardian very well obviously. Even though we are lazy bones Gemini got our back slightly

[Kürşat Aydemir @tlhepsh]

The extension, package, library or module repositories recently getting more and more exposed to malicious contributions. Repos must start ensuring they publish only safe packages or secure only module repos will have to emerge

[GitHub @github]

1/ We are sharing additional details regarding our investigation into unauthorized access to GitHub's internal repositories.

Yesterday we detected and contained a compromise of an employee device involving a poisoned VS Code extension. We removed the malicious extension version, isolated the endpoint, and began incident response immediately.

[Kürşat Aydemir @tlhepsh]

@usanewshq I hope it is a siren head

[Kürşat Aydemir @tlhepsh]

Gradient defined ML. It is also the wall in front of ML paradigm. The new paradigm is going to be built with a new approach: Perturbation. The true autonomous aware system will not be ML.

[Kürşat Aydemir @tlhepsh]

Coding humans becoming aware of the ultimate quality critic: ITERATION. Quality is iteration, however good product you do, the better or tweak is waiting for you at the corner.

[David Marcus @davidmarcus]

It's wild that every time you run a Codex code review from Claude Code, it finds critical issues. Not 95% of the times, 100%.

[Kürşat Aydemir @tlhepsh]

I think we've achieved man, like we've done, like I feel we finally did it. It is called AGentic Intelligence that his perception feels AGI. An AI model, in its own loop, continuously learning without forgetting dramatically would be AGI. And paradigm wouldn't be autoregressive.

[Yuchen Jin @Yuchenj_UW]

Jensen: “I think we’ve achieved AGI.”

“It’s not out of the question that an OpenClaw could create a web service or some interesting little app that, all of a sudden, a few billion people use for 50 cents.”

[Kürşat Aydemir @tlhepsh]

@Kekius_Sage Everything is one. Creator and created cannot be separated. Ultimate creator is the one and one is everything. If separated both will be impossible beings.

[Kürşat Aydemir @tlhepsh]

@elonmusk @peterwildeford 3 years is a long scale when you consistently and integerly talk about super intelligence in a shorter term.

[Kürşat Aydemir @tlhepsh]

Псори зыщ... singularity-р аращ

[Kürşat Aydemir @tlhepsh]

Man will realize many more things. LLM is not just a thing to implement from scratch, it is built with neural net and you'll learn to build primitive neural network too, if you truly don't use an external lib.

[Rohit Negi @rohit_negi9]

I am creating an LLM from scratch in C++ without any external libraries. Now I have finally realized why we need a GPU to train an LLM.

First I thought about optimizing the matrix multiplication using Strassen’s Algorithm, but it won’t show much improvement compared to parallel computation.

But first, I will run a basic version of this LLM on my Laptop.

[Kürşat Aydemir @tlhepsh]

@MyLordBebo Brace yourselves, both sides are fluffs, they don't care about the bluffs 🤣

[Kürşat Aydemir @tlhepsh]

Make Oracle relevant again

[Yuchen Jin @Yuchenj_UW]

“OpenAI signed a $300 Billion GPU data center deal with Oracle.”

> Larry Ellison gained $101B in one day
> now the richest man alive.
> but Sam doesn't have $300B
> Larry doesn't have $300B worth of GPUs.

We’re living in a simulation.