drusepth

Once you've written something, it's incredibly easy to overlook minute changes to the text.

See: why authors wait days, weeks, or even months before editing what they've written (or, if you're more interested: cognitive regression, inattentional blindness, and the effects of misdirected saccades).

What browser are you using that doesn't have feet (or animations)? Seems to all be there in Chrome.

What should you be looking out for as bad behavior from this model? I've daily-driven 3.5, 3.6, and now 3.7 for however long it's been out (alongside Opus, as I hit daily limits on both subs each day) and haven't noticed anything "untrustworthy".

A huge benefit of local setups at our studio is that a lot of our software can't run headless, so when we're having agents work in Blender or in Unity etc with MCP that'd otherwise eat up our normal computer use. (Try to have two people try to work in the same Unity editor at once... then try ten!)

We also built some QA agents that are always playing our games from the same builds a player would and flagging things to fix/improve; that alone needs the game focused and front-and-center so it can properly screen-capture for deciding what inputs to take next (and for screenshots/replays), which also means we can't really do any hands-on work at all on the machine when it's running.

Having a separate (and tiny) machine for all of this has been great. We don't bother with local models because, you're right, the $20/month sub is way better than anything that can run on small consumer hardware atm.

It's a pretty well-known stat that I think originally came from one of Schwab's detailed Modern Wealth Index reports -- which are survey-based so there is likely some inaccuracy; they report 3% margin of error -- but other similar studies generally put the number somewhere between 53% and 63%.

From their methodology:

> The Modern Wealth Index, developed in partnership with Koski Research and the Schwab Center for Financial Research [...] was conducted by Koski Research from January 12 to January 19, 2018, among 1,000 Americans aged 21 to 75. Quotas were set so that the sample is as demographically representative as possible. The margin of error for the total survey sample is three percentage points.

Anecdotally, this percentage makes sense to me based on roughly half of the people in my circles/family/life living paycheck to paycheck, but there's obviously some exposure bias at work there too. It's possible you have some exposure bias as well if that number seems too high for you.

Some other interesting stats for context:

* About 87% of adults with a household income of $100,000 or more own stock; compared to just 28% of those earning under $50,000.

* Only about 37% of adults hold stocks, bonds, or funds outside of a retirement account.

* Between 55% and 60% of Americans live paycheck to paycheck and are one large expense or missed paycheck away from homelessness.

* While some paycheck-to-paycheck workers technically own small amounts of stock, the actual value they hold is minimal.

* The wealthiest 10% of Americans own roughly 90% to 95% of all U.S. stock value.

Thank you for the detail/explanation. Hope my comment came off correctly as curiosity and a desire to understand, which this helped with.

Can you explain what's missing for the "open source" label that open-weight models like DeepSeek/Quen/GLM/etc don't release?

Is it just the supplementary data/code for how they were trained, not just the final product?

I feel like using a service like this that's owned by SpaceX (and, therefore, in close proximity to X / Musk / etc) would be even more of a supply chain risk than GitHub occasionally going down.

That sounds like something absolutely wild happening in Brave.

Or maybe one of your installed extensions modifying the page (presumably private tabs disable those)?

This is almost fine until god forbid you forget to charge it. My old phone had to be charged every night (or occasionally every other night). Charging once a week at most is night and day difference. Charging every other week would be wonderful.

A /precommit tool can be very powerful -- not only can it run the tests, do light code review on its own, spin up sub-agents to review performance/architecture/etc impacts, but it can also write hilariously-verbose commit messages once everything passes, detailing not only what changed but why it changed.. tagged with Claude as a co-author.

One of the best features I built into our game studio's AI router (IMO) was to take a random sampling of generations and have them generate across ALL models (that we support, at least) so we, the developers, can browse outputs and get a sense each model's output. Seeing them all side by side for generations you're already familiar with makes it feel like significantly less cognitive load.

The better I get at recognizing what each model is good/bad at, the more I'm glad we're taking the time to choose specific models for specific prompts -- and the more I wouldn't trust a generic router to efficiently route for me.

I don't think OP's "Why do people ask employees this stuff" question implies you _shouldn't_ ask, but rather wonders why people even ask when you're obviously not going to get an answer (and if you do, it'll be a non-answer filtered through PR/HR/MR/etc).

Yes, I've worked remotely for the last decade+ on many innovative teams and typically find them more engaged, autonomous, and innovative than most of the on-site-only teams I've also occasionally helped.

Same, it was $150 for 1400 sq ft house last I did it, so roughly same ballpark. I do remember it being closer to $80 10ish years ago though.

nit: 45k is the average car loan for a new car, but average for a used car is $25k. Purchasing a car is a necessity but purchasing a new car is a luxury.

I don't think a cleaning robot and a vehicle are in the same category, as vehicles are more a necessity (at the very least to get to and from your jobs, to buy groceries, etc) and a cleaning robot is squarely a luxury.

The job postings being all on-site (no remote) is kind of wild.

Bots routinely overwhelm my public-facing websites. The last few times I've turned on "attack mode" in Cloudflare (which I do when bot swarms get particularly nasty), traffic dropped by at least 20k requests/minute (which is huge, considering my human traffic is around 3/4 of that).

on the API for people who don't want to change models, but I imagine most people will probably switch to their cheaper Opus 5 (cheaper for us and presumably also cheaper for them)

Very, very, very excited to hopefully stop getting support emails at Notebook.ai for people trying to get help with NotebookLM.

Gemini Notebook is a way better name for the masses.

Ironically, the prevalence of AI "tells" like that (combined with the ubiquity of AI works passed off as human-written) will inevitably feed back into more use by non-AI writers who think they're normal.

(Also, I'm never gonna give up my em dashes.)

Is a google search actually easier than asking a model?

I reaaaaaally hope we have an option to disable those random ums and ahhs that interrupt for no reason. :|

I had a 300k LOC game that needed refactored and segmented into DAG assemblies for a composable engine to reuse in a few other similar games our studio is prepping to build.

I tried holding Opus's hand over a week trying to get it done with a ton of in-depth planning and manual course-correction, but it never got it to a state where it was "done" (kept struggling to differentiate between game content specific to that game versus game systems we'd want to reuse, and how to cleanly separate systems vs content when extracting).

Fable needed a little bit of hand holding but got it done in less than a day.

> Among other things, Anthropic has agreed to proactively detect and address security risks associated with the models; to work diligently with the U.S. government on protocols and standards and releases for Mythos, Fable, and future models; and to inform the U.S. government of any malicious activity.

How likely is it the resolution was just a direct flow of data on what each Claude user is specifically using Claude for, probably sent off to some other (gov-owned) AI datacenter for analysis?

What is an "ephemeral" model in this context?

This makes me yearn for the golden years of april fools online.

Before you start driving, at stop lights, while waiting in lines, etc.

I don't know how it works on CarPlay but when I turn my car on I have a bunch of suggested addresses (home, work, parents, recent Maps searches, etc) that I just touch-to-go. Having to use voice every time you want to navigate not only sounds unnecessary, but cumbersome.