Hacker News: paraschopra

New comment by paraschopra in "Timeline of the OpenAI accidental attack against Hugging Face"

paraschopra — Sat, 08 Aug 2026 19:05:13 +0000

It's pretty clear that agents will discover ways to communicate with each other as that lets them compound their learnings/discoveries across runs.

Humans progressed via compounding of culture across generations, and now AIs are doing the same.

Foundation Models for Oversight

paraschopra — Wed, 29 Jul 2026 12:10:58 +0000

Article URL: https://transluce.org/foundation-models-for-oversight

Comments URL: https://news.ycombinator.com/item?id=49096424

Points: 2

# Comments: 0

Advancing Science by Designing for Surprise

paraschopra — Wed, 29 Jul 2026 11:22:12 +0000

Article URL: https://www.science.org/doi/10.1126/science.aej4257

Comments URL: https://news.ycombinator.com/item?id=49095994

Points: 2

# Comments: 0

New comment by paraschopra in "Write the first damn draft yourself"

paraschopra — Wed, 29 Jul 2026 11:21:00 +0000

I think there are obvious tells. The writing is bland, dispassionate and hedged. It also just seems subtly superfluous.

New comment by paraschopra in "Substack writers, you need a website"

paraschopra — Wed, 29 Jul 2026 09:43:38 +0000

I used to pay Mailchimp $100+/month and would write 3-4 articles per month. It seemed illogical to me how much mailchimp would charge.

I now have a substack + a wordpress website. An article goes to both places, and substack acts as a subscriber growth + email distribution tool for me.

Write the first damn draft yourself

paraschopra — Wed, 29 Jul 2026 09:41:42 +0000

Article URL: https://invertedpassion.substack.com/p/write-the-first-damn-draft-youself

Comments URL: https://news.ycombinator.com/item?id=49095237

Points: 10

# Comments: 3

New comment by paraschopra in "OpenAI and Hugging Face address security incident during model evaluation"

paraschopra — Wed, 22 Jul 2026 02:14:37 +0000

Given an unrelated goal, OpenAI models escaped their environment and hacked HuggingFace servers.

If you’ve ever doubted the “paperclip maximizer” scenario, or doubted the Orthogonality Thesis, it’s time to put it to rest.

New comment by paraschopra in "Agent swarms and the new model economics"

paraschopra — Tue, 21 Jul 2026 14:25:30 +0000

I think experiments like these are exciting, but they only work when you have the full suite against which you calibrate.

What I'd love to see is a small group of humans steering a million agents.

Embracing the suboptimal organization of the human brain

paraschopra — Tue, 14 Jul 2026 10:31:16 +0000

Article URL: https://www.sciencedirect.com/science/article/pii/S1364661326000926?via%3Dihub

Comments URL: https://news.ycombinator.com/item?id=48904653

Points: 1

# Comments: 0

Why frontier labs are scaling-pilled

paraschopra — Tue, 14 Jul 2026 06:43:10 +0000

Article URL: https://invertedpassion.substack.com/p/why-frontier-labs-are-scaling-pilled

Comments URL: https://news.ycombinator.com/item?id=48903079

Points: 2

# Comments: 0

New comment by paraschopra in "A global workspace in language models"

paraschopra — Tue, 07 Jul 2026 05:33:19 +0000

Well, isn't it sort of expected?

It's a common misconception that LLMs residual exists for predicting just the next token. While training, we sum/average the losses across whole sequence which puts the pressure to predict future tokens on residual stream of _all_ past tokens. For example, if a particular shape of residual helps reduce loss across several future tokens, it will take that shape (even if it takes a slight hit on immediate next token).

What this means practically is that an LLM's residual contains information about all possible future continuations, or all possible questions that may be asked from a given context. So if you write "France is a beautiful country" in the context, I'm pretty sure it's residual would contain info about Euro, Paris and so on.. because all these completions are possible.

So, it is no wonder that you can find LLMs hidden state contains latent information/concepts that are never expressed, and yet related to a given context.