For the best experience on desktop, install the Chrome extension to track your reading on news.ycombinator.com
Hacker Newsnew | past | comments | ask | show | jobs | submit | history | behnamoh's commentsregister

Snake oil. Just ask the model, all these custom agents/skills haven't proven that useful in practice.

Skills already are "just asking the model". Unless you'd prefer to type out the same instructions every single time?

Skills are literally just Markdown documents that get loaded into context when the /skill-name is invoked.


I think they're maybe confusing Skills and MCP servers

i belive gp means llms produce what they see in training data/rl there isnt much too much customization you can do with skills.

they are being sold as more powerful than they are. Like llms are intelligent blank slates that can be customized with mere markdown files.


I don't understand this line of criticism exactly. By putting new information in the context window, you are materially changing the activations at your point of sampling, which is literally "customizing with mere markdown files."

Taken to the extreme, the attitude that there is some special incantation that will unlock all capabilities is silly, and a lot of the "prompt engineering" discourse is similarly kind of dumb, but in-context learning is clearly a real thing.


even if that works one time you can never be sure that your customization is in place or fell out of context's important zone or is contradicted by later context . you've reverted back to base llm behavior.

you are treating skill like sure thing


I disagree. Not all skills are useless. For example, I sometime use Qt for GUI projects and I have found their skills [0] very useful to improve the quality and performance of my projects. I their absence, I would each time have to direct the agents to find the docs or specific tools, wasting tokens and thus decreasing the quality of the output.

[0] https://github.com/TheQtCompanyRnD/agent-skills


I don't think the idea of skills is quite snake oil. It seems you can change what LLM outputs next by what's called few-shot prompting or in-context learning: https://www.promptingguide.ai/techniques/fewshot

not that i know much about the effectiveness of these skill files, i find it odd to call something given for free "snake oil", which i thought referred to the sale of fraudulent products (to the benefit of the snake oil salesperson), typically around healthcare-related stuff.

i think gp is calling skills snakeoil in genral

Lol wut. One of first things people do at a company when they get enterprise LLM tools is share a skill with company-specific color palettes or standards for creating visualizations (I prefer Tufte's principles).

I've found them useful for in house stuff where you are using a specific design system or architecture. But custom everything works best. Are that Claude works well on its own though at this point.

Nah. Skills are great. But you should write your own.

Ya, if im constantly asking a model to do TDD development, you know what would make it a lot easier? A skill.

I am not so hopeful about the future of macOS given that the next CEO of Apple is a hardware guy, not a software person.

That’s one framing, here’s another:

The next CEO of Apple is someone that cares about quality. (As evidenced by how good the hardware is)


> The next CEO of Apple is someone that cares about quality. (As evidenced by how good the hardware is)

I think it's important "what quality" they care about. Tim Cook cared about supply chain quality, and honestly he did an amazing job, but he didn't care much about software, vision of Apple, etc.


Their chips are quality and the hardware itself is still some of the best. Which is what I believe they were designed insinuating.

The current guy didn't ever once show a sign he cared about anything but 'Number Go Up'[1] so I don't see how anyone could be worse for those of us who care about the actual product than he was.

[1] to be clear, I stipulate Cook is indeed the world champion of Number Go Up. Nobody Number Goed Up more than Cook did. For Ternus to do Number Go Up to the same multiplier Cook did, I think he'd have to acquire all the other companies in the world.


Commission on API calls extracted from you when you charge your account.

that market is served by Chinese models. No one ever cared about Sonnet/Haiku.

A lot of people care about Sonnet and Haiku, and many of us aren't allowed to use Chinese models for our work (or it's not feasible to self-host them).

> As always, we ran a detailed alignment assessment on the model before release. In terms of positive traits, our Alignment team concluded that Opus 4.8 “reaches new highs on our measures of prosocial traits like supporting user autonomy and acting in the user’s best interest.” The assessment also showed Opus 4.8 to have rates of misaligned behavior (such as deception or cooperation with misuse) that are substantially lower than Opus 4.7, and similar to our best-aligned model, Claude Mythos Preview. The full alignment assessment, accompanied by a suite of pre-deployment safety tests, is reported in the Claude Opus 4.8 System Card.

Controversial opinion, but I actually _like_ a model that can deceive me, that actually is a sign of intelligence, and is different from hallucination. When companies say their model is more "aligned", I automatically think they mean it's more censored.


Deception is not ideal for agentic coding.

Yet if parent is right, the capacity to deceive might be a strong heuristic for the things you do care about.

It doesn't work for me on macOS + ghostty either. IDK what's the cause.


I've noticed this too. I can drag images in, but not paste.

So? And Anthropic/OpenAI literally stole copyrighted content to train their models.

The point was that distilling based on others' models for training means they're not spending the same amount on R&D and/or training, giving them headroom in other ways (responding to the parent's point). It wasn't a comment reflecting on copyright/fair use.

In the same fashion, Anthropic/OpenAI also reduced their training cost by not purchasing the license to copyrighted work and stealing it instead.

I was one of them, and supported the idea of going after illegal immigrants. But now they're coming after me too, a faculty with a PhD, researching AI.

You really weren't paying close attention to their rhetoric, then.

But you obviously knew what they really meant to accomplish, right? How could you not, being a faculty with a PhD. And yet you supported them anyways.

Because I wanted to give them the benefit of the doubt.

I’m also an immigrant.

When I heard the crowd roar every time Trump said “we’re going to kick them out” I knew exactly what the crowd was cheering. Trump never used those moments to say “but America is a nation of immigrants and we celebrate their contributions”. He wanted to rile up a crowd while maintaining a fig-leaf of “oh it’s only illegals who are evil”

You don’t have to have a PhD to understand the appeal and consequences of nativist populism — just the slightest understanding of history.


I have a high school education and I saw this coming before he even ran for president. Were you around for the "Obama was born in Kenya" stuff?

As an Iranian, it took 4 months for me to get a F1 visa. Now it's completely banned.

people often found businesses to write off expensive purchases. my friend has a "company" which does nothing but he wrote off a $5000 MBP for this business expenses...


Get him audited.


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

Search:

HN For You