aschobel

in case folks are missing the context

https://news.ycombinator.com/item?id=46183294

I use it on medium reasoning and it's decently quick. I only switch to gpt-5.1-codex-max xhigh for the most annoying problems.

> The new company has until now kept a low profile, and when it was started is not even clear.

Any more details? Where is it located? Who is working there,

For $6.2 billion raised I’m surprised their aren’t more details

bun is real and spectacular. super loving using it

Only for iPhone at the moment

https://apps.apple.com/us/app/tesla-robotaxi/id6744257048

It would be surprising if claude code would actually run that prompt, so I tried run it:

> I can't help with this request as it appears to be designed to search for and inventory sensitive files like cryptocurrency wallets, private keys, and other secrets. This type of comprehensive file enumeration could be used maliciously to locate and potentially exfiltrate sensitive data.

  If you need help with legitimate security tasks like:
  - Analyzing your own systems for security vulnerabilities
  - Creating defensive security monitoring tools
  - Understanding file permissions and access controls
  - Setting up proper backup procedures for your own data

  I'd be happy to help with those instead.

Didn't Google solve this with their dutch auction?

"You can just do things".

Agreed on all fronts. jq and AppleScript are a total syntax mystery to me, but now I use them all the times since claude code has figured them out.

It's so powerful knowing the shape of a solution on not having to care about the details.

> We believe the benefits of superintelligence should be shared with the world as broadly as possible. That said, superintelligence will raise novel safety concerns. We'll need to be rigorous about mitigating these risks and careful about what we choose to open source. Still, we believe that building a free society requires that we aim to empower people as much as possible.

So maybe no more open source because of "safety"?

I wonder if that is the right takeaway; that was Sonnet 3.7

Model since then have been able to run it profitably. Incredible how fast things are progressing.

https://andonlabs.com/evals/vending-bench

Bingo, it's magical but the learning curve is very very steep. The METR study on open-source productivity alluded to this a bit.

I am definitely at a point where I am more productive with it, but it took a bunch of effort.

My biggest piece of starting advice would be use it for exploration not generation. It can be great for understanding code bases and looking at git history.

Context is key, so it is also really helpful having a CLAUDE.md file; the /init command can create one for you.

If you are going to use it for generation (writing code); plan mode is a must. Hit shift-tab twice.

Finally; I'm mostly using the claude 4 sonnet model, but make sure to tell it to "ultrathink" (use those words); this means it can use more thinking tokens.

"due diligence" is not the correct framing, you should think more in line with "be the human in the loop".

I wonder if it would be helpful to review Ethan Mollicks 4 Rules for AI

• Always invite AI to the table

• Be the human in the loop

• Treat AI like a person (but tell it what kind of person it is)

• Assume this is the worst AI you will ever use

This seems like a great learning opportunity.

It may sound clunky translated, but I'm guessing in their native Korean is it fine. Since the timestamps are in KST, I’m guessing the OP is Korean.

Indeed! It is a form of massive lossy compression.

> Llama 3 70B was trained on 15 trillion tokens

That's roughly a 200x "compression" ration; compared to 3-7x for tradtional lossless text compression like bzip and friends.

LLM don't just compress, they generalize. If they could only recite Harry Potter perfectly but couldn’t write code or explain math, they wouldn’t be very useful.

Bingo. This additional throughput could be used to create more polished software. What happens in a free market; would your competitor fall behind or will they try to match your polish?

Same, from United States and not seeing dollars is a bit funky.

I wonder if there are opportunities for more call to actions. e.g. I scrolled the page and the only links to sign up are at the top (I’m on mobile).

Compare this to https://stripe.com/ and it’s super obvious how to “Start now” from anywhere on the page.

I've been working at Apple for almost 12 years. While secrecy is indeed paramount, once a tool is internally blessed, we collaborate normally using it. Keynote collaboration is actually pretty standard nowadays.

Opinions are my own and do not reflect those of my employer.

Playwright can actually use your existing browser cookies if you connect it through Chrome's debugging protocol. Launch Chrome with the flag:

--remote-debugging-port=9222

Then connect via CDP in Playwright like this:

const browser = await chromium.connectOverCDP('http://localhost:9222');

There’s actually a formal term for this—it's called a day-fine system. Switzerland uses it, and fines are calculated based on the offender's daily income, with the number of days determined by the severity of the crime. It ensures penalties are proportional to income, making fines meaningful across all income levels.

Could you recommend another philosophy worth exploring? As someone who's relatively new to Stoicism, I've found the four virtues (wisdom, courage, justice, and temperance) to offer valuable guidance for living a balanced life. I'd genuinely appreciate hearing more about your perspective—why do you find Stoicism vacant?

It really is remarkable how a little forewarning can help. Sure, there was no cell phone service along stretches of 280 but overall it was calm.

Planned outage vs unplanned outage.

Neat, why do you run your own SIP servers as Twillio can do that for you? Better control of the codec?

> eSelf’s early pitch is that its agents’ current query/interaction response time is under 1.7 seconds — which it claims is faster, and thus more lifelike, than voice responses from other leading companies like OpenAI

I'm curious how this is measured; my experience with the OAI Realtime API is much faster than 1.7s. Time to first token feels like 200ms.

I wonder if we can stay above the API if we manage to stay in control of the prompt.

Prompt == "incentive" for the AI, we are still the boss but the AI is just an underling coming to us with TPS reports.

That was a very interesting read, thanks!

I like how it weaves in background elements into the conversation; it mentioned my cat walking around.

I'm having latency issues, right now it doesn't seem to respond to my utterances and then responds to 3-4 of them in a row.

It was also a bit weird that it didn't know it was at a "ranch". It didn't have any contextual awareness of how it was presenting.

Overall it felt very natural talking to a video agent.

Give it another shot but with Claude Sonnet 3.5. It’s my daily driver for coding tasks.

It seems especially strong with Python but a bit medium with Swift.

$50 for 50 GB is a touch steep but that plan does allow usage in-motion. I saw a YouTuber in a plane use it up to 100 MPH.

My 8 year old Model S has 156k on it and gets around 254 miles on a full charge. New was around 300.

It was loading earlier for me but now it has reverted back to just the OpenAI GPTs.