ph4rsikal

I suppose that's what happens if demand goes up for external reasons while supply doesn't follow.

This is me. I was in a leadership role at a large institution and made the mistake of joining a promising startup (great partners, good funding). In 2021, the company went down, and since then I have been rejected for everything. And by everything, I literally mean every possible way to get money in. I have only been scraping the leftovers at the bottom. Recently, I joined an outsourcing company to do digital transformation work (I have deep AI experience). It's shocking to see people getting hired with 0-1 interviews who then prove to be incompetent just because they are the right caste.

I think we are at the beginning of a new golden age of software engineering, but not in the way we think. The cost of creating software might go down to zero with AI. That doesn't mean that no skills are needed. Quite the opposite. If AI would create outstanding code, then it would win every Kaggle coding competition on the planet. But it doesn't.

]1] https://github.com/storytold/artcraft [2] https://huggingface.co/blog/a2aprotocol/universal-modder-a-p... [3] https://jdsemrau.substack.com/p/jamming-with-jev-and-claude-... [4]

Great work, op. I have also had good experiences with Qwen3-TTS. Might be worthwhile to explore for everyone who wants to do longform locally.

Maybe the engineer makes 90K, but there are still thousands in social contributions. Can be up to 15K.

An LLM is just a brain in a vat. Our brains would also not reason in such a condition. Given the right framework that can do miraculous things. I am not saying that these systems are conscious, but they are able to abstract their context in a way allowing them to understand their own limitations.

https://www.finextra.com/blogposting/31255/adaptability-as-e...

Humans are also ethnically homogenic animals.

If you look at Europe's immigration problems, this is probably a good thing.

Makes much more sense described in this way.

Mistral's annual revenue (700M) is what Anthropic generates in 3 days. It will be extremely difficult to build new models with these numbers.

Arianspace launches from French Guiana while a French commune is not really close to the actual continent of Europe.

It is indeed a huge success but it also underlines how far ahead the rest of the world is.

Claude used to be 10x more efficient. I think there are no barriers of entry between one coding agent and the other. Therefore, I expect them to reverse as soon as people switch. At least this is what I will do.

The blue line (200K context) is lower than the red line (300K context). Indicating that it's cheaper to run blue rather than red over longer cycles.

> Qwen 3.7 Pro is just unusable. wastes too much time, goes off track, useless stuck loops, cannot debug at all. D

Anthropic should not have bugged their knowledge distillation attacks.

Hope they move away from the engagement model. I prefer blogs over reddit/insta/tiktok in search results.

ChatGPT Codex launches, and it's the worst of both worlds. Who would have thought?

The on-ramp in crypto is incredibly well built out. But then, once you are in the fake money world, you will notice that the off-ramp doesn't exist.

You only say that because you benefit from it.

For tenure-track positions—the pipeline for future faculty—white men have gone from 49 percent in 2014 to 27 percent in 2024 (in the humanities, they’ve gone from 39 percent to 21 percent).

https://www.compactmag.com/article/the-lost-generation/

Increasingly men have been disenfranchised over the last decade so the only way forward is to take on such titles to at least appear successful. It looks like normal behaviour to me.

I am not railing against anyone. I am simply pointing out the fact.

So it's like Sliceline from Silicon Valley (the show)