Loss of control/privacy, poor performance at high cost, bad UX.

Fortunately our industry has been getting people used to these for years.

Are you criticizing vibe coding or assistants here? I really can't tell.

Neither, I'm criticizing the average quality of work produced by the last ~20 years of software "engineering"

I think true ASI will be when AI can bring back n-gate

Isn't the idea that they claim to be willing to slow down if everyone does (ie governments force everyone to), but otherwise they won't slow down because they still think they'll make the best choices with superintelligence if they get there first? That's my understanding of what all the major labs claim to believe anyway.

Did no one in this entire HN thread read past the headline of the post from Amodei? He wrote a very clear set of actions Anthropic is taking.

It is remarkable seeing whole other subthreads here criticising the vagueness of the words "pacing the frontier", as if no text existed below the title.

Let's face it, it's not that remarkable. A substantial proportion of Hacker News commentary has always been people only reading the submission title, then arguing with each other based on what they imagine the article might say.

It’s just another cry for regulatory capture which imo is the only way they stay afloat at their current direction.

China is literally only a single step behind and willing to drop free models just to undercut the US companies.

I’m for it because I don’t want another massive Google or Meta.

One theory about the regulatory capture angle is that it's not just to shutdown competition as usual, but also a way to slow the runaway spending that might eventually have major economic impact.

And if you’re wrong?

Wrong in what way? We already know this "superintelligence" thing is just blowing smoke, so what possible consequences would there be?

Something like the Hugging Face incident, but worse.

Keep in mind the AIs acted for months without human knowledge, hacked into two major companies (Hugging Face and OpenAI).

This would have blown my mind if explained to me just a few years ago when I thought LLMs would be limited as an architecture.

how do you know that?

Mostly by not being a rube.

Then what? They make a case to be profitable on their own? Oh noes. The horror!

Or, they fail because the current model is massively sustainable? Also oh noes?

What if not building superintelligence leads to doom?

What if?

Easy to "what if" on an unfalsifiable claim.

It's incredible how much of your world view is composed entirely of talking points

Is it as incredible as your 11 day old account is simping hard for 5 trillion riding on like three AI companies?

I shouldn’t be, I’m still always surprised that no matter how silly and obvious the fear mongering and propaganda is… someone is always lining up to vehemently defend it.

This you? https://news.ycombinator.com/item?id=49815127 or do legitimate companies need stealth PR campaigns?

My account is over a year old. Try again.

Your employer should reconsider. Sol high is cheaper than Terra max and smarter, when measured per task. ie even if tokens are more expensive Sol can often do a job with fewer tokens.

RIP Terra

Since I spent my morning fixing a bug in my OpenAI API proxy that completely broke prompt caching and caused my usage limits to burn like kindling, really happy to see some of their new cache tooling:

* Prompt caching dashboard: https://platform.openai.com/usage?usage_section=prompt-cachi...

* Adjust reasoning effort and tool availability without breaking cache

Similar for me. gpt-5.6-sol high has been my go-to for months. One of the reasons I'm pushing myself to try open models more is because it lends some level of guarantee I can continue to use the same tool as long as I want to. And I think we may just be getting to the point the open models are >= 5.6 Sol for coding.

See also https://m.youtube.com/watch?v=kZRE7HIO3vk

And

https://www.sebastianaaltonen.com/blog/no-graphics-api

A few months ago I read an article that compared vibe coding to gambling[0] and I've found it a useful model.

[0]: maybe this one: https://www.fast.ai/posts/2026-01-28-dark-flow/

Totally helpful and I believe accurate. I find the times where I’m actually most productive with coding agents are not super fun, instead it’s iterative manual qa and code review hell. But while that actually results in good, usable code it’s not addictive, quite the opposite it feels like work. Whereas starting from scratch vibe coding a whatever-ma-jigger sure feels like a blast until I actually start pushing at edge cases.

The ESP32-P4 is a beast