One source of trouble here is that the agent's view of the web page is so different from the human's. We could reduce the incidence of these problems by making them more similar.

Agents often have some DOM-to-markdown tool they use to read web pages. If you use the same tool (via a "reader mode") to view the web page, you'd be assured the thing you're telling the agent to read is the same thing you're reading. Cursor / Antigravity / etc. could have an integrated web browser to support this.

That would make what the human sees closer to what the agent sees. We could also go the other way by having the agent's web browsing tool return web page screenshots instead of DOM / HTML / Markdown.

I could handle this style when it wasn't everywhere. But now I've developed a hypersensitivity and can't bear it. It's like suddenly most of the internet is in a language I can't read.

I always wonder now if an article was written by GPT, or by someone who spent so much time chatting with GPT that they've started sounding like an LLM.

I have genuinely wondered if I've picked up an elevated tendency to say "it's not just X, it's Y". No smoking gun that it's affected me just yet, but it's at least a live enough issue that it raises the question.

I hand wrote something recently that I re-read the next day and I'm worried I sound like an LLM now, I'm pretty sure I always sounded like one because I like it to make exhaustive lists in my sentences, but it makes me wonder if the bot is rubbing off on me.

It's part of our ecosystem now, we unconsciously mirror the patterns that we notice around us. This will include the language of LLMs because it has been invited in. We always affect the environment, and it always affects us. I hope that the consequence will be that we reduce the fluff, we stop writing to sound important or to justify a position, but instead use language to operate on the level of insight and shift our future into one that benefits us all.

I wish almost every news article came with a statistics section. If you must, go ahead and write that article about a particular murder or traffic accident or drug trial or earthquake. But if you don't include statistics on similar events over time, geography, demographics, etc, you're misleading more than informing.

I'd _like_ to blame the reader -- inferring anything about how common something is based on how often it's reported is unreasonable. But readers do make that inference, and writers shouldn't pretend they don't know it.

And for most of us nowadays it's not about articles and writers. It's about eight-second video clips on TikTok and creators. So I don't have any hope that we'll become better informed.

Even in the just-photoshop-not-ai days product photos had become pretty unreliable as a means of understanding what you're buying. Of course it's much worse now.

Note: Please understand that monitor may color different. If image does not match product received then kindly your monitor calibration. Seller not responsible. /ebay&amazon

look at the bottom of the sleeves, they don't match. the bottom of the jacket doesn't match either.

I didn't see it at first sight but it certainly is not the same jacket. If you use that as an advertisement, people can sue you for lying about the product.

> I've tried to ask dozens of companies that wanted to hire me just for how many shares were outstanding and/or authorized.

"Wanted to hire me" as in they made an offer, or an earlier step? At offer stage, I've never had a company refuse to answer these questions. I don't have "dozens of companies" worth of experience though, maybe one dozen if that.

Every time I hear this I think experiences and expectations are vastly different between SV and the rest of the country. 30ish years of working in New York and I haven't encountered a single company that isn't 100% opaque about their equity to employees until time of exit/IPO. And I keep a large network.

That said, everyone here treats equity of non-public companies as if it's toilet paper. Some of my coworkers got very lucky and very rich when our company went public, but that was also a long time ago now.

I worked at one for five years: Materialize. Based in NYC. Everyone knew how many shares they owned, what percentage of the total equity that represented, and what the rights of the preferred share classes were.

Glad to see this data point. This is a company founded relatively recently and I hope this means that some kind of change is happening regionally.

Isn't "context" just another word for "prompt?" Techniques have become more complex, but they're still just techniques for assembling the token sequences we feed to the transformer.

Almost. It's the current prompt plus the previous prompts and responses in the current conversation.

The idea behind "context engineering" is to help people understand that a prompt these days can be long, and can incorporate a whole bunch of useful things (examples, extra documentation, transcript summaries etc) to help get the desired response.

"Prompt engineering" was meant to mean this too, but the AI influencer crowd redefined it to mean "typing prompts into a chatbot".

Haha there's a pigheaded part of me that insists all of that is the "prompt," but I just read your bit about "inferred definitions," and acceptance is probably a healthier attitude.

Whether or not you're really outliers, it would be very surprising if "my friends and I" were representative of the general population.