fhub

If you are interested in learning about the other perspective, you can watch some parents’ congressional testimony here https://youtu.be/y8ddg4460xc?si=-yYduYDppF4TQWqD.

The character.ai one is gut wrenching.

The lone ranger may have actually done something optimal but indirect. There is a lot of press that went global pointing out real problems. Japanese are proud people, this might actually help direct public funds to solve the problems.

Right now I'll pay 2x for a subjectively 20+% better coding agent. But in a year I don't think there will be an agent that to me is subjectively 20% better amongst the big three.

> Opus feels more like a guy to me, while Codex feels like a machine

I use one to code and the other to review. Every few days I switch who does what. I like that they are different it makes me feel like I'm getting different perspectives.

He is likely working on a very clean codebase where all the context is already reachable or indexed. There are probably strong feedback loops via tests. Some areas I contribute to have these characteristics, and the experience is very similar to his. But in areas where they don’t exist, writing code isn’t a solved problem until you can restructure the codebase to be more friendly to agents.

Even with full context, writing CSS in a project where vanilla CSS is scattered around and wasn’t well thought out originally is challenging. Coding agents struggle there too, just not as much as humans, even with feedback loops through browser automation.

They use the word "Sonnet" 60+ times on that page but never give the casual reader any context of what a "Sonnet model" actually is. Neither does their landing page. You have to scroll all the way to the footer to find a link under the "Models" section. You click it and you finally get the description

"Hybrid reasoning model with superior intelligence for agents, featuring a 1M context window"

You then compare that to Opus Model description

"Hybrid reasoning model that pushes the frontier for coding and AI agents, featuring a 1M context window"

Is the casual person meant to decide if "Superior" is actually less powerful than "Frontier"?

I do appreciate this note more than others. It is food for thought. I think it could have been worded a lot more respectfully though.

He'll likely file in California or Federal and ask for Jury trial. I think a Jury will be sympathetic. I doubt Google will want this to go to a jury trial - not worth the risk, further news cycles of negative PR and impact on staff morale. NPR is credible and liked.

Door close button is supposed to cancel the door dwell time. But due to some disability codes in some regions all major manufactures allow it to be disabled (as required by some codes). i.e. The owners/managers/technicians can disable it.

I thought by not naming it wouldn't shift the focus to the particular model, but it did the opposite. It was gpt-5.3-codex in medium mode.

Even with automated tests you'd need to think of this exploit right? Perhaps fuzzing would have got it. The mailing lists says they proved it successful on

- OpenIndiana

- FreeBSD

- Debian GNU/Linux

So not complete YOLO.

See https://lists.gnu.org/archive/html/bug-inetutils/2015-03/msg...

FWIW, a well known LLM agent, when I asked for a review of the patch, did suggest it was dodgy but didn't pick up the severity of how dodgy it was.

My wife, who was on a H1B visa and managed to fly without an ID a few years back. They took her to some side room, asked a bunch of questions and looked her up based on name, DOB, address etc.

The use case claimed here is (a) they can move around (b) they are "universal".

But

(a) Those things look like they need a wide berth to move around and flat terrain

(b) Those end effectors are far from universal. The payload weight seems so low that it even dropped an empty box at https://www.youtube.com/watch?v=3FIXjy2GWTg&t=150s

This feels like the start of treasure hunt like game. Between username of rabinovich (as others have pointed out) and the prior submission by rabinovich of an archive.today like tool 3 months ago - https://ghostarchive.org/. When you click into the search query examples for ghostarchive such as this one https://ghostarchive.org/search?term=https://docs.google.com. Many of the documents are very weird indeed.

It feels to me like this should have gone through private arbitration rather than public accusations.

Do you let it consume PII? Anything related to authenticaion?

What is new to me is that people let LLMs consume PII and potentially authentication related data. This, frankly, is scary to me.

Are you comfortable giving LLM read access to fields that have PII? Anything related to authentication? Is it allow-list of access or a deny-list?

Our solve is to allow it to work with a local dev database and it's output is a script. Then that script gets checked into version control (auditable and reviewed). Then that script can be run against production. Slower iteration but worth the tradeoff for us.

Giving LLM even read access to PII is a big "no" in my book.

On PII, if you need LLMs to work on production extracted data then https://github.com/microsoft/presidio is a pretty good tool to redact PII. Still needs a bit of an audit but as a first pass does a terrific job.

I doubt there is a law, that when enforced, allows an ICE officer to sell someone's phone, at least without a lot of paper work.

(1) Can be true. However with the "Absolute immunity" rhetoric from VP, this seems unlikely in the current environment.

(2) If one officer is telling you to move on and the other officer is telling you to get out of the car and you need to listen to both, which one do you comply with?

(3) You can be a "Legal Observer" does not imply "Political vigilantism". The idea of a "Legal Observer" is laid out by National Lawyers Guild and to some extent, by ACLU.

Here is how the gift card scam works (in Australia)

[Quote]

Yes they do still get activated at the checkout. But when you go to redeem, the code is missing the last digit or two so it doesn't work. People take the unactivated gift card, tamper with it to get inside carefully so it's not detectable, scratch and get the code, remove the last digit or two, replace the scratch off layer, put the unactivated gift card back on the shelf. Then after you activate the gift card at the checkout, they redeem it.

[/Quote]

From this discussion

https://www.ozbargain.com.au/node/937339

Outside of engineers there is a whole raft of people on a team that should pick up and push back on this sort of copy problem at all phases of building a product.

For USA, A Visa is a right to request entry into the country. The I-94 defines the duration you are authorized to stay. You can have an expired Visa and time left on you I-94 and remain in the country.

Some day the forward facing cameras on your car will read the street cleaning signs and warn you before you park and notify you if you are still parked there when street cleaning day/time is approaching. The nuance is that you are allowed to park after the truck has gone by (in SF).

https://www.sfmta.com/getting-around/drive-park/how-avoid-pa...

Pre-kids maybe twice a year. Post-kids maybe twice a month.

YouTube Documentation on this feature - https://support.google.com/youtubekids/answer/6172308?hl=en&...

The embedded walkthrough video on how to set it up is really quite good.

He references childcare and paternity leave in the post and he was a co-founder in a $3B acquisition. To me it seems it is a time-of-life/priorities decision not a straight up burnout decision.

Some might, but I imagine some have left the company when the pain is really felt and are excited to do it all again at the next company.

My five year old played with this quite a bit at Maker Faire last year. He picked it up quite well. I had it in my mind to get him one for his next birthday but forgot until I saw this post. His school has Makedo tools and he likes them. So the combination might be something that he'd use on a semi-regular basis. We have no shortage of "material" being delivered.