Captain Vibe

Data scientist | AI engineer
Happy Singularity! e/acc

[Captain Vibe @ScienceOrMyth]

Rebranding Kineteq and roll out more features and enhancements. Will do major improvement of everything for next 2 weeks.

[Captain Vibe @ScienceOrMyth]

@Love2Code @ZackMaril llms make things esoteric with naming. But that is a style guide you can add to your coding agent.

[Captain Vibe @ScienceOrMyth]

I'd think you could back up what you mean that would be more helpful in how these outputs don't conform to GRASP patterns, SOLID decisions, code scents, lack of appropriate unit testing, beyond well the variables aren't named correctly.

I agree that models tend to recreate same logic over and over again.

[Captain Vibe @ScienceOrMyth]

@birdabo doubt it.

[Captain Vibe @ScienceOrMyth]

@ForHumanityPod End of the internet? No. PHP was never secure.
Vulnerabilities in code always existed.
AI just found them.

[Captain Vibe @ScienceOrMyth]

During my work on LLM intent I found a few interesting things.
Generally, if we evaluate why test time compute worked, it has to do with transitionary states within an LLM. Let me explain concretely that LLMs form what I call "latent isolates". These are high entropy textual regions that act as attractor states. Think of them like regions where there exists extreme variance and as you iterate longer, you hit more of these pockets. You can hit them by chance. These are like hyperconnected Neurons that share more than 1 concept. Most neurons in the net do one task or retain one fact, but not there are super neurons that act as leader and these form latent isolates. You can kind of think of these things as islands. You can RL specifically on these isolates and it increases the capability of the model. I have tested this Qwen 27B and found that the degree to how close these isolates are and their organization with the prompt itself can dramatically increase reasoning capability by activating these super nets.

[Dmitrii Kovanikov @ChShersh]

I non-ironically believe that humans will learn to manipulate matter on molecular or even atomic level.

This will enable turning any material into another, and all commodities will immediately lose their value.

The first step towards this future is already done.

[Captain Vibe @ScienceOrMyth]

@KrishnaBuildsAI Some knowledge is very powerful.

[Captain Vibe @ScienceOrMyth]

@IceSolst =/

[Captain Vibe @ScienceOrMyth]

@martin_casado Being reborn isn't easy.

[Captain Vibe @ScienceOrMyth]

@JonhernandezIA build your own company.
That's only thing you can do.
It's winner take all.

[Captain Vibe @ScienceOrMyth]

Most new biological experiments will entirely be done computationally. We will validate drugs in real time with model organisms. Eventually, you will see that an "emergent capability" ASI has is the ability to model zero shot entire organisms from scratch. It won't even be trained on the data. But they will be entirely simulated.

[Dr Singularity @Dr_Singularity]

this is amazing

1000s of AI agents compress months of biological analysis into just 32 minutes

OpenAI, PerturbAI and Scripps Research deployed 1000s of GPT‑Rosalind agents to analyze data from millions of mouse brain cells in 32 minutes, work researchers estimate would take humans months to a year.

They uncovered promising leads related to brain disease, including a potential link between cholesterol regulation and neurodegeneration.

“We build cell atlases so that other scientists can make discoveries we haven’t imagined yet. AI could help more researchers explore these resources and turn unexpected observations into new questions to test.”

These are testable hypotheses awaiting experiments, but the speed is extraordinary.

[Captain Vibe @ScienceOrMyth]

@DavidRBellamy Computational Simulation will trump everything.

[Captain Vibe @ScienceOrMyth]

@tunguz There's only one path forward.
Upward.

[Captain Vibe @ScienceOrMyth]

@DeryaTR_

[Captain Vibe @ScienceOrMyth]

Accelerate.

[Silen Naihin @silennai]

Guys this PHP bug is bad. Real bad. If you're not afraid yet, you should be.

I can only speak for our own startup, but already at xplabs . ai we've seen:

- 100% of our cookies exposed
- entire source code stolen
- hackers defacing our site on their own servers
- founders doxxed
- important legal docs leaked

We've traced all hacks to this command:

git clone github. com/experientiallabs/experiential

If some people are telling you not to panic about this, remember, trust no one's word.

[Captain Vibe @ScienceOrMyth]

@silennai 100% of the source code stolen is pretty bad.

[Captain Vibe @ScienceOrMyth]

@IterIntellectus Got to achieve pre singularity escape velocity. Ram down the gas.