seizethecheese

> The world of agents gets a lot more exciting (and less costly) with decision models. If you’re running a classification agent, try the Strands Decider repo

Wait, their solution here is not an "agent" at all? It's literally output length of "1", unless I'm missing something? (It's not clear what they are replacing is an agent either, it looks like a reasoning model with no tools that outputs JSON.)

We do, but it’s not definitive. We are building a battle system on top of OpenCode [1] and Opus 5.5 seems to have dropped since Nov 3rd.

If you calculate Elo only before 11/3 and compare to only after 11/3, the Elo is 52 points lower after. Unfortunately we are still in beta so this isn’t conclusive (confidence interval of -160 to +40).

If we do this calculation on all our models, Opus 5.5 is first in apparent drop. (Again, not definitive, but supports the conclusion.

[1]: https://agentduel.com/leaderboard

You can probably assume they’re using this as an example of an “invented” right that many here would agree with, to say it’s not necessarily a bad thing for courts to “invent” a right.

Sure, I meant htis for electric trucks in general, and I was wrong about the physics anyway.

> Consider what happens if you walk at 2 MPH into a concrete pillar which is fully anchored into the Earth.

Excellent thought experiment and this helped it click.

I was thinking of F = MA, where mass seems important here. Is this the wrong mental model?

Edit: okay I get it, the human acceleration from impact basically asymptotes.

Truly terrible for pedestrians. Modern trucks are already death traps with their super high front grills, multiplying the weight compounds the problem.

Do you think so? I'm not arguing here, I genuinely don't know.

My kids are too young for this, but surely there's a tradeoff here. Should you override some of your child's natural wonder because of adult aesthetics (or worse political concerns)? Yes. But how often? I'm not sure the line should be drawn at a truck design. (Although, maybe you're teaching your kid taste, again I'm not sure.)

Fair enough, and fascinating. I guess I’m conflating first degree murder with “death penalty”. Please tell me nobody is pleading guilty and getting the death penalty.

Why would anyone plead guilty to first degree murder?

Evidence of pre-meditation !== evidence the suspect applied pre-meditation. You need to tie the premeditation to the suspect, I think. In your example, if you can't prove the suspect wrote the note, I don't think you can convict on first degree murder.

Beyond what others have said about creativity, projects still take work to iron out. LLMs can speed up work 10x but they can’t validate very well.

It’s quite shocking that innocent people get convicted of first degree murder. In theory, premeditation beyond a reasonable doubt must be proven. And you need consensus of 12 people!

To me this is proof of failure of system to educate a sufficient majority of people to think rigorously.

I disagree. Just finished a book about a serial killer who relished fame. He really didn’t want to die. He fought death sentence hard to try to live rest of life in supermax prison. I think the preferences of those facing this predicament, who tend to appeal in every possible way is instructive.

Whenever I see daily stock results I try to reframe it in a longer term. Interestingly, Nasdaq is only up ~4% YoY.

This is surprising to me, given how much acceleration there has been in AI capabilities since the same time last year. I guess costs and risks are accelerating just as fast.

The issue of you already buying something is interesting to me. It’s so obvious that they’d fix it if it cause revenue loss. My guess is that conversion for any ad is so low that you’re indeed more likely to buy a lot of products twice rather than buying something totally different.

> ie, they're trying to coerce you into wanting their products

Im not sure if you mean “coerce” literally but if so this is overkill. Maybe “persuade” or to be more critical, “manipulate”?

I think you’re alone to think time for task completion doesn’t matter

We validate people are doing real work using LLM judges.

Posts on who wants to be hired divided by posts on who is hiring

The number of posts there is skyrocketing while the number of posts to “who is hiring” is flatlined https://micjm.github.io/hn-hiring-index/

If you're interested in this kind of thing, I just finished reading a book where the author improbably was commissioned to write a book about a serial killer but the killer was a friend of theirs!

It's a decent read, the author went on to being a bestselling true crime writer. https://www.goodreads.com/en/book/show/15654.The_Stranger_Be...

It's a ratio of job seekers to jobs available. This is super relevant to estimate "tightness" of job market. (Obviously this is a simplification because HN isn't a self-contained universe, but it's interesting to see the chart spike so hard.)

Higher in the index means more people looking for jobs than job posts.

Pandemic effect is nothing compared to agentic coding.

Blog post on this: https://micjm.com/blog/hn-hiring-index-agent-duel/

We allow early voting. This is against what most arenas do, but is on purpose. If the model is faster to do simple stuff, it should be chosen. In this sense, DS4.1-Flash is better on many tasks.

You can remove early voting from leaderboard and they are no longer the same.

Speaking of arena comparisons of K3.1 and DS4.1-Flash, we're spinning one up that's more relevant than design, I think: https://agentduel.com/leaderboard

Here, DS4.1-Flash and K3.1 are basically tied, with K3.1 being slower and more expensive. Big error bars, though.

Even with error bars, DS4.1-flash is clearly an impressive model.

Naive question: are you conflating debt with liabilities? Are these two separate technical terms or are they the same?

To my lay understanding, debt means a loan you have to pay back, whereas liabilities are everything you're on the hook for.

A friend of mine wanted this, so I built it for him.

Full disclosure: fully vibe coded while working on an agentic engineering tool.

To me, "a probably $200-300B company" sounds absurd. You're describing one of the possible outcomes of a massive range that goes from $0, where they get over their skis on financial commitments to well beyond $2T, where their 10x YoY growth continues for a few more years. A $200-300B outcome is possible, but certainly not probable. (And in fact, for us HN commenters, the most probable outcome should be assumed to be roughly what the market will bear.)

This feels like an ex post explanation for 20th century history.

For example, the Great Depression. Hoover was a literal textbook author, but is widely blamed.

The only real professor was Wilson, but he had similar failings to what people in this thread are worried about.