News for hackers. Stuff that matters.

nastrofa

Re:Speculative tool calling: cutting voice AI agent latency on Tuesday October 13, 2026 @07:32AM (#50070012)
Have you tried the model ?
0.5 dropsout to build on an idea they got on an acid trip
slop alert
This is so cool!

Imagine creating real simulated env for every physical task and using RL to train robots. That will be a game changer.

I think AI and humans operate very differently

If a human doesn't know an answer, they will either not fake it or will give a brief response to the best of their knowledge. They don't know what they don't know.

If an AI doesn't know the answer, it will find connections in its vast training patterns to hallucinate a response.

Humans think and use logic AI predicts and finds the most probabilistic answer

It would be really cool to create different analysis across the time

- Each month's / year's top news headline

- Left / Right swings of publishers

Re:A New Mental Model for Work in the AI Age on Saturday May 02, 2026 @08:18AM (#47985763)
A challenge I am facing is sycophant nature of LLMs. It just validates my thoughts and never questions it, can't trust it when its appreciating or supporting me.
Good news for Frontier it seems