Browse Comments — Raw (as collected)

Honestly, the part that stood out to me wasn’t how much energy a single AI query uses. It was the reminder that AI is no longer just a software conversation. It’s becoming an infrastructure conversation. The more capable these systems get, the more important energy, compute, and data centers become.

Youssef Ben Mahmoud I currently believe thay AI redistributes the surplus to the few who managed to expand their human agency (through AI)

Srinath Vemuri now every human can be ambitious, curious, and ponder about meaning, having been released from the bondage of tasks [that are automated]... how would the world look like?

This is a practical blueprint for moving from experimentation to real execution.

The idea of AI preserving and documenting the reasoning paths behind discovery is especially compelling for both research transparency and knowledge transfer

Marcus O'Dell Do you think AI’s greatest impact on scientific research will come from increasing productivity, or from enabling entirely new lines of inquiry that were previously impractical to pursue?

Anthropic engineers shipping 8x more code proves that software development is shifting rapidly from manual execution to high level architectural oversight.

Thats...always been teh goal

"Oh no! You can do what we can do." - Nervous

In the end, every LLM has limitations set by the mathematics behind it. Therefore, if these models optimize themselves in the future, they will do so within these very limits. See https://arxiv.org/abs/2409.05746 and https://arxiv.org/abs/2401.11817 for details.

If Claude is already authoring over 80% of production code, recent graduates shouldn't just be practicing syntax memorization. True professional advancement will belong to the students who learn to act as system editors, focusing heavily on logic verification and architectural oversight. https://thenextweb.com/news/anthropic-claude-recursive-self-improvement-code#:~:text=Anthropic%20reveals%20that%20Claude%20now%20writes%20over,up%2C%20but%20because%20Claude%20does%20it%20now.

The key issue may not be only how capable AI systems become, but how clearly their work can be constrained, reviewed, and connected to real workflows. Progress without control creates noise. Useful AI needs boundaries, validation, and accountability built into the process.

If Anthropic engineers vibe code and their product is getting insane hype, you know that vibe coding is not optional. Not alone, but with oversight and human judgement baked in.

AI will start building and create a physical Body with additive Production and 3d printing :D das

Like so not new 10 years ago. I'm getting tired of Anthropic worship.

It might also be that they hit a wall with AGI about two years ago; making this massive data centre buildout is an enormous waste of resources. It might also be that they are using their IPO to foist their incredible mistake onto everyone else... aka ... exit liquidity for private equity.

no shit.

It seems we're going to enter an infinite loop. 😄

The capability jump matters, but most businesses will trip over control long before they hit recursive self-improvement. I am seeing teams struggle more with review, boundaries, and accountability than raw model power. Which of those becomes the first real bottleneck in your view?

Fascinating glimpse into AI's accelerating capabilities and the crucial need for responsible governance.