droidjj

I love how all the replies to this comment recommend completely different strategies for deciding which model to use.

Product page: https://www.weaverobotics.com/isaac-1

Can we please stop submitting fully AI-generated text to HN?

The default WebView backend uses the operating system's own webview for small binaries, and you still have the entire npm ecosystem available through Deno's Node compat layer.

Sounds like a similar architecture to Tauri, but your business logic is in typescript instead of rust.

I wouldn’t trust Chinese labs’ TOS very much. It would be incredibly difficult and expensive to enforce any of those terms.

edit: this is a comment about suing and enforcing judgments against Chinese companies in the US, especially software companies, not necessarily about how trustworthy the Chinese labs are.

What kinds of tasks are you finding deepseek v4 incapable of?

The general consensus is stocks nosedived after the strong jobs report, because strong labor market means its more likely Fed will hike interest rates to curb inflation.

I struggle to get quality results from the frontier models at contexts > 256k anyway.

EOs have force of law if exercising valid executive authority. POTUS could have, e.g., conditioned federal contracts on compliance with this vetting program, but didn’t.

What seem like the most interesting bits, selectively pruned of legalese:

Within 60 days of the date of this order, the [executive branch] shall:

(a) develop and maintain a classified benchmarking process to assess the advanced cyber capabilities of AI models and determine the threshold at which an AI model should be designated a “covered frontier model” for the purposes of this order, sharing such assessments with AI developers and researchers as appropriate. …

(b) design a voluntary framework with AI developers through which developers would be able to: (i) engage the Federal Government to determine whether model(s) under development meet the designation of “covered frontier model”; (ii) provide the Federal Government with access to covered frontier models, subject to [some conditions]; and (iii) collaborate with the Federal Government to select trusted partners that will have early access to covered frontier models to promote secure innovation and strengthen the cybersecurity of critical infrastructure.

So… it’s entirely “voluntary”? This has no teeth.

Running my family's social media site off a single sqlite DB on a VPS is great. ~15 users, almost zero maintenance.

Details, please!

~Qualifying all of this with the caveat that I'm not a social scientist.~

I really appreciate your response, which made me realize this is more nuanced than I thought at first blush. (I wrote that reply before you linked the National Academies review, which was quite helpful.)

Maybe I am truly morally panicked, but I'm really hesitant to brush aside the evidence as "basically nothing." Small effects are worth paying attention to (https://psycnet.apa.org/record/2022-26026-014). And quantifying the harms of social media use is less like quantifying the harms of, say, smoking cigarettes, because social media has (indeed, is built on) network effects. You could plop a kid from the 50's into a modern American neighborhood and their mental health might decline even if they don't use social media because of the way it has changed childhood. When the social life of an entire generation is transformed by technology, it stops making sense to ask "How much does one hour of Instagram hurt mental health?" Yet that is essentially the question all the studies are designed to answer, and even still, we see an effect.

Anyway, thanks for taking to time to explain your perspective.

I'm not an expert in this area, but this seems far from flimsy.

https://www.thelancet.com/journals/eclinm/article/PIIS2589-5...

https://jamanetwork.com/journals/jamapsychiatry/fullarticle/...

https://jamanetwork.com/journals/jamanetworkopen/fullarticle...

https://jamanetwork.com/journals/jamapediatrics/fullarticle/...

Can you please explain how I'm wrong?

The title of the paper this blog post is based on, https://www.frontiersin.org/journals/developmental-psycholog..., is "We don't know how social media bans will affect youth but we're doing it anyway!" Which is a little cheeky for my taste given the seriousness of the issue these bans are trying to address.

This severely underplays the "risks" of not banning social media for teenagers. The link between social media use and mental health problems in teens is extremely well documented.

Section 40. Civil penalties.

(a) A large frontier developer that violates this Act shall be subject to a civil penalty in an amount dependent upon the severity of the violation that does not exceed $1,000,000 per violation.

(b) A large chatbot provider that violates this Act shall be subject to a civil penalty in an amount dependent upon the severity of the violation that does not exceed $50,000 per violation.

(c) A civil penalty described in this Section shall be recovered in a civil action brought by the Attorney General.

This is the "penalties" section of the bill (available at https://www.ilga.gov/documents/legislation/104/SB/PDF/10400S...). I'm not sure what counts as a violation, but if it's simply the act of releasing the model, this isn't going to have much impact at $1,000,000 maximum.

As to the bigger question of how can Illinois regulate a company that isn't based in Illinois: the general principal is that states can regulate a company's conduct within or targeted at the state. In the U.S., there are constitutional limits to this kind of regulation (https://texaslawreview.org/state-regulation-of-online-behavi...). It's a fuzzy line, though, and if this were a big enough headache for the frontier labs to comply with they would probably litigate it.

Link to the analysis this article is based on: https://www.lesswrong.com/posts/wRNJZz2iYrfDaSDdz/claude-aut....

The paper this article is based on: https://arxiv.org/pdf/2509.04664. It's not as sensational as the headline suggests. E.g., "Hallucinations are inevitable only for base models." I feel like this is something most people familiar with how LLMs work already understand.

Check out pi.dev. OpenCode is a nice batteries-included Claude Code replacement, but I’m in love with the extensibility of Pi.

This is what people said about the banks in 2007. Just because the big players’ balance sheets can take the hit doesn’t mean the wider economy is insulated.

Hallucination is also much better controlled in the context of agentic coding because outputs can be validated by running the code (or linters/LSP). I almost never notice hallucinations when I’m coding with AI, but when using AI for legal work (my real job) it hallucinates constantly and perniciously because the hallucinations are subtle—e.g., making up a crucial fact about a real case.

The pricing is an issue.

Yes, it reads like every goddamn LinkedIn post these days.

As a lawyer, I'm excited about this, but there are two roadblocks that I'm not sure how Anthropic will navigate:

(1) For non-lawyers who use these skills/connectors/whatchamacallits to try to get legal advice, their communications are not protected by attorney-client privilege. This will absolutely bite some people in the ass.

(2) If a lawyer uses this with confidential client information (which, to the uninitiated, doesn't just mean SSNs and bank account numbers, but "all information relating to the representation of a client") and forgets to toggle off "Help improve Claude" in their settings, they have possibly (maybe even likely) committed malpractice.[1]

[1] https://www.americanbar.org/content/dam/aba/administrative/p...

Gift link: https://www.theverge.com/transportation/926741/tesla-cybertr...

Oh, I hadn't even considered my comment could be taken the other way. I'm absolutely referring to the nasty email this guy received.

That kind of hate is why we struggle to keep good open source maintainers. Let’s try to be kind to people who gift their time and talent to the world, please.

The paper: https://www.science.org/doi/10.1126/science.adz4433 (April 30, 2026)

The most popular claim is the memory consumption, which to be fair has been decreasing over the last decade, but my main complaint (as I usually drive a 64GB RAM MacBook Pro) is the lack of visual consistency and lack of keyboard-driven workflows.

Lucky you. I avoid electron apps because I'm limping along with 16gb.

The OspreyCam, installed in 2018, has been streaming each nesting season ever since. Powered by solar energy, it transmits wirelessly to Lobsta Land Restaurant in Gloucester, and from there, out to the world. The stream runs daily from a p p r o x i m a tely 6:00am to 6:00pm.