notfromhere

It’s heavily RL’d and it’s gotten worse because they’re using LLM judges, that’s why it’s gotten worse. The last good writing model was GPT 4.5.

I suspect they’re leaning heavily on some kind of auto verifier, and that’s why the writing sucks, because good writing is so subjective.

And they’re probably nerfing what it can get from training as they’ve lost lawsuits for scraping copyrighted work, so they’ll purposefully avoid anything that references a specific writers style, etc.

If a party expressed anti-democratic goals, protecting them by the social compact of democracy is stupid.

Look how well that idea worked out in Turkey. Authoritarian movements use democracy like a train and get off when they get power

I think if you buy into the idea of cancel culture is a real thing, you’ve already lost the plot.

[flagged]

I believe they’ve pushed back the ipo. I always figured that frontier labs were going to be very vulnerable against tech giants that had actual profitable businesses to fund their AI work.

The amount of total funding Anthropic or OpenAI has raised individually is like a year of ebitda for Google, meta, or Microsoft. And it doesn’t look like there’s a real moat outside of money

An Enterprise client at that size can just be someone at an f500 put a credit card in

Nobody is quiet about doing well. And the frenetic release schedule of kinda half baked products tells a story (not a good one)

Why build cars when you can just change the seat covers

I don’t know if convenient is a moat when it makes switching very easy

They are very unprofitable…? I don’t think that’s really in dispute. We haven’t yet seen a profitable frontier lab and model pricing remains fairly subsidized

This really just exists so cognition can stop spending API tokens with Anthropic or OpenAI.

Basically any successful AI based service will do this because at scale the frontier models are expensive and you’ll have enough data to fine tune your own.

Same reason Harvey is doing models now and basically every other provider

Well the models did get better but yeah their early product was godawful

its nowhere near as good as it was in Dark Sky. the apple weather app is pretty bad where I'm at.

Aren't they used doing this so they can generate training data for using Macs via the UI with computer use?

Sol Light on computer use is fantastic. I use it whenever I need to dive deep into whatever shitty web saas app menu if the API is unavailable