user43928

I mentioned their cabal, as in the Signal group "Exchange coordination" reported in 2022, with executives from Binance, FTX, Kraken, Tether, and Tron.

The crypto economy isn't real, it's a pointless wealth redistribution scheme.

Anyway, that and your question are hardly relevant to the fact that Binance and their cabal own the crypto exchanges and consequently could manipulate the price to reduce the size of their obligation denominated in crypto rather than currency or stable coins.

Surely with all their security programs, they would run testing on the ChatGPT implementation.

I speculate that the AI didn't find this and that human effort and intuition for where to look were needed to find this vulnerability.

Crypto isn't foreign currency. The price may fluctuate wildly based on sentiment and action taken by exchanges such as Binance, which would directly impact their obligation here.

Comparing crypto to actual currency typically connected to large, real economies is not relevant here.

It's an important distinction.

$3.5B stolen was "stable coins", which you could equate to a piece of paper with a dollar amount written on it.

Those are largely freezable by the issuer of the paper.

Then we have $2.4B in Bitcoin and ETH stolen. Those are not directly freezable.

However, if that does create a liquidity challenge for these exchanges, what if the price just so happens to drop by 60% after trading is restricted? That liability might shrink in actual value.

Only tangentially related, but recently when I was working on a feature using iCloud, I noticed that it's just using Google storage APIs. I read Apple uses AWS S3 too.

Supposedly Apple holds the encryption keys on their own infra, but I still found it interesting how they are just reselling Google Cloud storage here.

Cursor made Grok Bot and is working on the Grok models.

I wouldn't say it's obviously unsuccessful.

More of whatever you had in mind.

Clearly there are few limits to what you can implement with AI agents today.

Depending on the complexity, you will spend a lot of time QA testing, giving feedback, and iterating.

One can take shortcuts such as porting existing software, as you mentioned. Many are just content publishing crappy novel software that they barely tested.

You don't need to do that. No one is forcing you not to QA test and iterate until you have a good product.

That's a fair point, and with newer package managers, at least the dependencies no longer automatically execute build scripts on install.

There were a few common sources of large runtime dependencies in the past. Like the browserify Node polyfills, babel-polyfill and runtime with core-js, React -> fbjs, Material UI -> JSS and plugins, styled-components.

It's also not trivial to tell whether a dependency you listed under devDependencies is actually imported and consequently bundled into your production build.

Since today we no longer need polyfills and the modern stack with Vite etc. doesn't have a sprawling dependency tree, I think things are now looking much better.

Interesting. The Grok Bot computer connects to your box via SSH and using herdr it can work well with Claude Code and Codex there?

I started using Grok Bot a few days ago and I have seen zero screwups.

I don't know what I'd need a channel for, creating bots/personas seems more natural.

Adding credentials is essential, contrary to what the author of the article said. I gave it my library login and it can fetch newspapers for me instead of me clicking through the app every day.

When I had to provide a GitLab token, it nicely spawned a box for me to save the credential.

  - What did not work well -

Voice calls. The quick responses are often dumb while the proper model does the thinking, and in the middle of talking to me it would stop to do work or searches. It did not feel realtime.

Permission settings. There is a form where you can automatically allow actions you describe in natural language. I tried to allow it to run Claude on my computer or Send certain emails automatically, but the settings achieved nothing.

Coding. I wasn't able to make it use Claude, and its own work is worse than me using Opus 5.5 in Claude Code.

  - What worked well -

Planning flights, accessing newspapers from my online library, checking stock in online stores, sending regular summaries of events and weather to plan activities.

No, because contrary to keeping Daylight Savings Time as it is, that would come with other tradeoffs.

Probably at a point where my worries will be when my team of construction robots finishes my villa, rather than whether companies are prepared to ditch source code.

That's simply not true.

Until spec files can be worked with so efficiently that we can just regenerate the binaries from them for every change, what you said makes little sense.

You're going to need an intermediate portable form of storing the program. It so happens that all the tools and AI are optimized to work with source code.

Then you probably haven't worked much with JS.

The Jest test runner alone has 250 dependencies in its latest version, and v25 had over 500.

Spy seems inappropriate if you can only read information stored in this side channel intentionally.

We'd first need to establish that this side channel leaks usable information during regular use.

My employer, my family, the TV's schedule, or whatever else.

At the end of the day I lose an hour of sunlight without good reason.

I'd fight tooth and nail for them not to take an hour of sunlight from me during summer evenings.

It is not roughly the same if the sun rises an hour earlier instead, at 04:30 instead of 05:30.

I don't get up before 08:30 and I will not somehow adjust my schedule by three hours to use the sunlight in the morning.

"The agent's own model", huh?

I hope Anthropic's next model is capable of better writing.

I find the way it adds this unnecessary "own" very grating in explanations.

For example: "Two benchmarks, each with its own score".

Of course the author could also just write at least the summary himself.

Mythos is Fable and you can see on the chart that Fable 5.1 is no longer at the frontier.

Doesn't Google have a particularly bad record recently at making their models publicly available?

Especially on their Plus tier for Gemini.

I think it's a fair concern that this might take weeks and trail Fable 5.5.

Would it have been so hard to call it Gemini 4.1 or 4.5?

To me it suggests someone with an interest in promoting AI.

I'm an enthusiast when it comes to using AI to my benefit. I could care less about what others do, as long as it does not concern my work.

I think it would be strange if someone were to label me an 'AI booster', but maybe that's just how I feel.

I've seen Moonshot pay creators on X for posting promotional messaging about Kimi K3.1 Pro, with the requirement not to disclose it as a paid sponsorship.

X then took action and suspended creator's accounts that were posting about Kimi K3.1 without the paid sponsorship tag, allegedly hitting some accounts that weren't paid by Moonshot.

In any case, I do not think that people are typically being paid to promote AI generally, rather than specific products.

To imply so on comments from well established HN users, as if it was unlikely they simply find value in AI and have a genuinely positive opinion, often seems misguided.

To me it seems like a term that users pessimistic about AI and the future apply to those with an optimistic opinion.

By speaking of boosting, they add an implication that the opinion isn't as genuine as their own and is perhaps motivated by some sort of financial benefit.

Am I wrong?

The people that I've heard say that there is no practical difference between Opus 5.5 and DeepSeek often prefer an AI-assisted hands-on approach.

I've rarely heard the opinion that for heavy agentic work, Opus wasn't noticeably better.

The whole point of the comparison was that people pay $1500 for an iPhone while a $500 Android phone could theoretically provide much of the same experience for a fraction of the price.

Comparing two devices at the highest price point is besides the point.

If you look at what it costs to serve Kimi K3, the exact opposite of what you claimed is the case.

It seems likely that Anthropic and OpenAI's inference margins on API prices are >95%.

That also casts doubt on the parent's claim that subscriptions are subsidized. It rather seems almost certain that they are overall generating profit.

The idea that noisy usage data is worth that much for training as to balance the fatal risk of the crime you describe, imo borders on a conspiracy theory.

I am under the impression that curated training pipelines are driving the gains, not utilizing random usage data in nefarious ways.

I think people often underestimate how far behind most of the competition is.

Isn't the first open model that likely surpasses original Mythos from February this week's Kimi K3.1 Pro?

Nothing compares to September's Opus 5.5 yet.

Now the question is whether Anthropic is preparing another phenomenal upgrade with Fable 5.5, or if the progress eventually slows down and competitors catch up.