toddmorey

OK, fine. But they are also an ad network, which absolutely acts as a publisher. I think they should be liable for the revenue they bring in from knowingly fraudulent ads.

Legal definitions can should change in order to keep up with society and technology.

fwiw, I’ve only done cli tools for agents to install in their own sandboxes, not access on remote machines.

This was just to avoid MCP infra, but the newer stateless protocol makes that much more trivial.

Depending on use case, would start with just solid HTTP APIs, then a simple CLI to consume them, then add MCP using the new spec. Build slowly only adding new access protocols as needed.

Be surprised how far apps can get with just solid & well documented APIs. Modern agents are smart!

I’ve had a really good experience with Vivaldi as a replacement for the Arc browser.

It really is nice to have a browser from a company whose incentives seem to be for the actual users.

This is not true. Think of the problems AI has with creating appetizing food imagery, for example. Images that every human annotator marks as unappealing with a high degree of consensus.

The problem is in the rewards, but it's more how they average across domains and contexts, so AI misapplies them. Actually, in human preference evaluations, most of Gen Z in particular extremely adverse to the common look of AI images.

So it's not the annotators involved in RLHF, it's the models failing to map to human preferences at the task level. The failure rate is also uneven across domains.

Where there are challenges with human annotation, it's not the ability to gauge what's good but rather how to correct what's wrong. Most people can tell when something is off but struggle to articulate why. That's where more domain expertise is required.

Why do all these tools require a paid account? Bruno specifically says it’s platform-free and stores everything in git.

I agree with this, but I can’t help but reflect on how we poisoned our own well by destroying the human internet before AI.

The answers to almost every search are already out there, but we made the public internet so toxic and user hostile, you now need a whole new layer of AI to sift through the bullshit, ads, and clickbait.

I don’t think for a second that AI will not go the same hellish direction.

So here's what I don't get: The United States is the world's largest consumer market. From what I can find, personal consumption expenditures (PCE) account for about two-thirds to 70% of total US economic output. We live and die on consumer spending and confidence.

So next we automate the bejeebus out of this economy with AI robots and Elon believes it grows 3x in under five years. Splendid. But if a consumer-driven economy has no payrolls, it eventually has no buyers. It just seems to me that ultimately, billionaires and investors cannot decouple themselves from the masses... but maybe it's a game of get mine before the system collapses?

Always exuded so much authority & confidence. What a legend.

But he start losing his hair at 17 and really struggled with it. A turning point came when a friend (a Hungarian judo black belt) and his wife wrestled him down and physically cut off his toupée with scissors.

His agent even had him do the Star Trek audition in a wig, but luckily the crew had better ideas.

I can't decide if the report's writing ("One number, followed" & "Quoted by all five. Sourced by none.") is being lazy or ironic?

Aside: Does anyone else find the new AI texture explosions Apple's using on their marketing pages now a little unsettling and un-Apple? I guess I'm desperate for at least one tech brand to feel / remain human.

This has been my hope for a while! I live in a 1911 house and the amount of small detail everywhere would be impossibly expensive by modern construction standards. I also like modern minimalism, but I know many of the drywall corners everywhere in modern homes aren't design decisions as much as cost savings measures.

Neither one: I think this is the best output they could get.

Packaging & pricing seem to be two things Anthropic really struggles with

I think the only solution will be a local model on the phone that routes these to the trash

Which benchmark? Would love to see it & any datasets.

It’s super tricky to tease out whether the similarities point towards distillation. A lot of the aesthetic similarities are naturally emergent from the training data.

The conversation examples are impressive but this made me say holy shit out loud: "it runs locally on a MacBook Pro M4 Max or better. And we figured out how to train Dessa on only one NVIDIA GB200 NVL72."

If it really can replicate that low latency + that expressiveness on local devices, that's one heck of an accomplishment for "indie" AI.

This isn’t even the sandbox. It needs Docker or other virtualization. This seems to sit a level up and orchestrate sandboxes and access to network & resources. It’s a pretty heavy setup from what I can tell.

What I mean is CMS isn’t required now for most web projects.

I heard it expressed that AI is bringing a new era of small and personal software.

I think CMSes and websites are where that is most applicable. You no longer need these generalized abstractions and can now make your interfaces and data types just perfectly match your use case with little effort.

We went through great pains to build these mammoth, one-size fits all applications like Drupal and Wordpress for running dynamic sites across industries and companies with very little in common.

These days, I love to stay close to the model and stay close to the metal.

what's striking to me is this isn't some novel, new-to-the-AI-era vulnerability. We've dealt with the security concerns around discoverable "private" links for 20+ years. There's... so much prior knowledge, art, and discussion.

How is it v1 with an architecture page coming soon? I'm really having a hard time gauging the maturity of software & services these days.

I still struggle to wrap my head around a $30,000 GPU card. But impressive payback numbers.

“I am going to show you where on earth in the future we will find the biggest rock: Space.”

https://youtu.be/aO0TUI9r-So?si=7fNwDvf86GpVFxDR

Beanie’s Baby

I don’t get the endgame. If your model underperforms in real life usage, the backlash comes quickly. I guess I’m asking what exactly the perverse incentive is? A day of press at most?

I say never do you own plumbing or electrical and I stand by that. You’ll want to be able to show electrical work was inspected and approved by a master electrician god forbid an insurance event like a fire puts that into question.

Trades will be under attack on both fronts. The DIY kits will be used to train the robots. Don’t under estimate how motivated these companies are to embody AI and replace human labor.

What's interesting about the Hyrtl Skull Collection is the collection was built to counter the claims of phrenologists by showing just how varied Caucasian skulls across Europe really could be. Noble, but if you are going to build a skull collection, you are going to use dubious means to collect them.

It is something else to see that many skulls looking back at you. Reminds me of the Crypt of the Capuchin Monks (walls covered in skulls) with the haunting inscription "What you are now, we once were; what we are now, you shall be." Stays with you.

Here’s how I think about it: only organic data from real scenarios will provide realistic edge cases. The world is much harder to accurately simulate than people sometimes assume.