andromaton

Split personality A) pierce reality B) stocks

AI2027.com puts this around early 2027 and treats the loss of legible chain-of-thought as a major negative turning point for alignment.

to those that want to "punish" the uk ai security institute: these failures are good for all of society.

Sorry for his loss, and for everybody helped by her in any way. However ineffective a comment like this is

Pixel accurate is easy to say, hard to do, over-specific, and counterproductive.

We've known artificial intelligence will do unexpected things since the 90s and that it can do difficult things since 2025. Pausing worldwide not easy but we do harder things all the time.

impressed at the look (not the geometry) of 3.6 lol at flash-lite could not find your posts about them

Extremely useful

A lower key Putnam moment

Self-replication requires a minimum amount of information plus machinery. Consciousness may have a similar threshold: below some level of self-modeling, nothing; above it, degrees.

Now includes multi-second llm latencies

Since yesterday.

Claude FM: heartfelt Gemini: psycho Grok: incompetent OpenAI: by the book

If priced like other Anthropic models, Mythos will make vulnerability discovery a lot more accessible.

The author compares it to AISLE, ZeroPath, and OpenAI’s Codex Security. AISLE and ZeroPath are much more expensive. OpenAI’s Codex Security is gated.

Most people don't care about the first two and don't complain about the latter's policy because they are all specialized models and/or harnesses.

Mythos will be available to all.

Dave Barry also insists he is not dead.

https://davebarry.substack.com/p/death-by-ai

It seems it's quite exhausting.

Also too many lands and hits.

Related https://news.ycombinator.com/item?id=47838508#47839757

The ux of copilot driving Claude beats Claude Code handily.

I never understood the low visibility.

Expensive ram is annoying. I don't look forward to expensive ai.

- New sign-ups paused for Copilot Pro, Pro+, and Student plans.

- Individual plan limits tightened.

- Opus models removed from Pro plan.

A comment from someone who knows or knew the author or was part of the project sharing details that makes readers feel like they've just been handed backstage passes.

A highly voted comment that seems insightful if you don't know the domain but amateurish if you do.

A comment noting the article is from YYYY and asking if the mods can add (YYYY) to the title.

A comment explaining that this has long been a solved problem in some other language or domain

A commenter pointing out that the article is implicitly US-centric

A comment noting that the submitter has posted this exact link multiple times in the last six months.

A post agreeing, adding a personal anecdote about a gentle nudge received years ago that the commenter still thinks about.

A post thanking dang for his tireless moderation work.

A post by dang about related posts:

A Technical Blog Post by a Big Name Expert - https://news.ycombinator.com/item?id=5326511 - March 2013 (189 comments)

A compelling title that is cryptic enough to get you to take action on it - https://news.ycombinator.com/item?id=43219556 - Feb 2025 (112 comments)

A Hacker News thread where every comment describes itself - https://news.ycombinator.com/item?id=38451203 - Nov 2023 (74 comments)

A request for others to add to the list.

Indeed, some examples:

https://news.ycombinator.com/item?id=12340348 Neural network spotted deep inside Samsung's Galaxy S7 silicon brain (2016)

https://ieeexplore.ieee.org/document/831066 Towards a high performance neural branch predictor (1999)

Tbh I'm surprised so much guidance.