Hacker Newsnew | past | comments | ask | show | jobs | submit | code51's commentslogin

Why on earth is nobody here talking about the sudden jump to use von Mangoldt function?

The reasoning trace never types Λ, never types "von Mangoldt", and never invokes ∑_{q|n} Λ(q) = log n.

There is a clear discontinuity at play. I remember an article on this, maybe a comment by Terence Tao himself, seen here, but cannot find it.


During training they gate with a lot of guardrails the format of the reasoning tokens output. They don't just use a reward for getting the correct answer during training but also reward human readable output. That said, if they didn't, the reasoning tokens that are the most efficient to get to the final correct answer during training would most likely look like a lot of gibberish.

There is a relationship between the tokens in the output in the model's vector space, that is the most important, and something hidden we will never see.


I think that the thought trace is definitely incomplete - you can see cases where it is like and "let's calculate the integral:[no integral calculated]". The train of thought it's on towards the end of the trace looks like an entirely different approach than what it ends up returning, so I think we are just not seeing the part where it hits on the right approach (sadly).

Thought traces are indeed not an accurate representation of what models actually do. If you ask an AI model to add two values it will do so, then in the next prompt ask it to explain the algorithm it used, it will regurgitate that it used some standard textbook method, whilst in reality it used a completely different algorithm. Thinking LLMs don't record the neural pathways they used.

Does DeepSeek's solution look more traceable?

https://chat.deepseek.com/share/nyuz0vvy2unfbb97fv


Anything from Steve Yegge must come with an AI bias disclaimer now.

He was also effectively paid $300,000 to facilitate a cryptocurrency rug pull on Gas Town, bowing out after the rug pull because Gas Town required his “full attention”. [0]

Everything he says now is suspect.

[0] https://steve-yegge.medium.com/steveys-birthday-blog-34f4371...


Yeah, I used to respect him as a tech blogger, but you can't wash that crypto stink off once it gets on you.

Exactly this. OpenAI is running huge workloads silently, without anybody patting their back.


Input:

  i ate a cookie
Output:

  I’m thrilled to share that I’ve just successfully completed the consumption of a high-quality, artisanal cookie. This experience reinforced the importance of consistent self-reward and maintaining a growth-oriented fuel strategy. Grateful for the opportunity to recharge and optimize my performance for the challenges ahead. #GrowthMindset #PersonalDevelopment #FuelingSuccess
Amazing.

Oh, tried with a horse.

  I am thrilled to announce that I have successfully completed the challenge of consuming an entire horse.

  This journey taught me so much about resilience, dedication, and the importance of setting audacious goals. It wasn't just about the meal; it was about pushing past my perceived limits and embracing a growth mindset.

  Key takeaways:
  1. Scalability is everything.
  2. Persistence pays off when tackling large-scale projects.
  3. Fueling your ambition requires thinking outside the box.
  Grateful for the support of my network as I continue to hunger for the next big opportunity! #GrowthMindset #Leadership #Resilience #Disruption #NextLevel


Skills is a generic construct. System prompt is generic as well. Subagents, AGENTS.md, CLAUDE.md etc. these are generic, "please care for my instruction" kind of constructs without any real guarantee to close gaps.

Tool is generic (CC vs OpenCode) Ecosystem is already same everywhere.

I don't understand what's the point.


The point is that wrappers matter. Orchestration, tool calls, reasoning loops, system prompts, agentic capabilities. Output is different, quality is different.

This is the moat for AI frontier companies.


Physical losses in undermaintained water grids are the biggest cause for the issue. Yet, economic downturn creates a vicious circle: governments avoid infra spend because of low funds, then agriculture and other economic output gets hit because of water shortage. Lower resource lower will for infra spend. Until you hit the very low: stopping the grid because day zero. At that point, both the grid and city hygiene becomes a mess anyway. Costs build up so much that most governments cannot cope up with it properly.

and this is why you need sane people at the top earliest


> Physical losses in undermaintained water grids are the biggest cause for the issue.

Correct me if I am wrong, but doesn't that mean the water is returned back to the environment? It's not made unusable, nor does it disappear permanently.


In many places water is pumped from deep underground aquifers but leaks go to surface ground waters and could quickly end-up in the Ocean so aquifers are still depleted.


Groundwater recharge from surface to an aquifer where it "rests" can take up to 1000 years.


We're talking about accessible fresh water. If it evaporates and then rains over the ocean then it's lost, or as the article mentions, if it becomes contaminated then it's not longer usable as fresh water


Wait, your job was len(x) == 15?


The hardest part is figuring out how to use leftpad.


The hardest part is checking to see if the version of left pad you just pulled or one of it’s 300 dependencies has a supply chain vuln in it.


Recommending new material involves risk. Once these companies go big and mature, they hate risk. They hate risk in hiring (taking a chance in people) and they certainly hate risk in algorithms.


> A well written book on such a topic would likely make you rich indeed.

A new religion? Sign me up.


12 Angry Agents


Consider applying for YC's Summer 2026 batch! Applications are open till May 4

Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: