user43928

This release is interesting for enterprise at API rates.

At least unless the Terminal Bench 4.0 score is more indicative than the other ones, because that one shows it clearly worse than 6.1 Sol for both performance and cost.

Looks great on the first chart, but terrible on the Terminal Bench 4.0 where it needs xhigh to come close to 6.1 Sol, and costs more for a slightly worse result.

We all know AI does not write well today.

What JetBrains needs is nothing short of a revolutionary product, which they would need to sell with an excellent introduction.

Instead they phoned it in and pasted that slop:

> Once a GitHub PR is open it shows the CI checks and review decision, and lets you mark it ready, hand failing checks or a diverged branch to the agent, merge it with the repository's last-used method, reopen it, or delete the remote branch.

When I was checking earlier, it seems some European projects will use SpaceX to launch. Probably makes sense economically.

From what I saw Europe has no plan to launch a large constellation that could become an alternative to Starlink. The ambition apparently does not exist.

Sure, but are we building them fast enough?

Our current global share of AI compute is estimated at 5%, US at 78%, China 11%.

We're projected to briefly rise to 8% around 2028, and back to 5% for 2031.

That does not look like sovereignty and independence from the US.

Accusations like these are against the HN guidelines by the way.

They are often unfounded and obviously incorrect when made towards old accounts with established history of posting.

If Europe can't even build data centers, which come with low to moderate environmental impact, they won't manage to deploy a Starlink alternative either.

Even though they could use SpaceX launch capacity to deploy such a constellation, Europe has no such plans.

Your comment seems to reinforce the parent's opinion about societal paralyzation.

Europe can claim to be cleaner or more democratic, but it doesn't change that they get nothing done.

I'd say it's the fermented, charred dough in combination with fresh, varied toppings.

That's seriously impressive.

I wonder how such games can bring in millions per month.

Perhaps when your ads outperform the competition due to these devious optimizations, you can afford a higher bid for paid acquisition, keeping users coming.

I heard that with the latest policy update last week, Apple rolled out some heuristics that aim to detect feature flags that bypass App Review.

Supposedly this can also fire for legitimate A/B tests or other debug flags that aren't stripped from the Release build.

I suspect these games will be fine as long as they have an Account manager and pay Apple millions in commissions or for paid acquisition.

I found that for Qwen Image 2.1 the prompt enhancer makes a big difference.

It won't bridge the gap to gpt-image-2.5 though, and it's going to slow down the process further.

With a 4090, Qwen Image 2.1 Turbo takes around 15s per image, depending on the resolution. That's without the prompt enhancer.

It's calorie-dense and high in sodium?

I think we could do a bit better here with the explanation of what looks bad to you..

The benchmark is hardly perfectly secure code.

Was not your concern that the generated code would be significantly less secure than what an average engineer produces?

The goal posts here basically moved toward infinity.

Regulators have been asleep for the last two decades while Apple and Google continue to extort 30% taxes from developers.

There was a glimmer of hope with the DMA but regulators are again asleep, emboldening Google to cripple app installation outside of their store and force developers to enter into a contract with Google to distribute from the developer's own website.

Whether someone like Musk is worth 500B or 3T on paper makes little difference when it comes to power, don't you think?

If that was the goal, surely one would need to run for president.

I don't think the brains of billionaires work so differently from us all. Who doesn't want his name in history for their achievements?

I know that I wouldn't want a dystopian future, and I think it is the same for most billionaires. I find arguments that they are all evil rather unconvincing.

Is that what they're doing, or are they merely retaining shares of the companies they created?

Do they need to donate their shares in order not to be considered hoarding wealth?

I don't think this is what's going on in the first place.

Like, are eg. Musk and Zuckerberg doing this for bragging rights about a number on financial statements?

To be honest I have no idea about Zuckerberg's motivations. With the proposed billionaire tax in California, recent reporting indicates his main concern is retaining voting rights at Meta.

With Musk I believe he is investing into space flight and colonizing Mars.

I fundamentally don't agree with the idea that rich people are evil and motivated by nothing other than greed.

A poor design decision in OpenCode configuration handling is interesting, but not directly relevant to the question.

Assuming the question is whether not reviewing generated code is likely to create vulnerabilities despite SOTA models having audited the implementation.

Here it seems to have been a design choice, rather than an implementation mistake. I do not know what models were used and whether a review or audit has been done.

We are basically back at a vague the AI failed.

And sure, this happens, but there are some strong claims I'm reading here that are imo not evident:

- not reviewing implementation is foolish

- doing so will result in a flood of insecure software

- more AI cannot fix this

--

My arguments on each point:

- depends on risk tolerance, attack surface, error rate

- depends on actual performance of the proper setup I described on typical tasks, compared to the average engineer

- more AI can fix this, provided the audits become accurate enough to identify vulnerabilities with eg. 99.9% reliability

--

Are you sure that such accuracy for auditing is impossible, and will remain impossible in the future?

I'm to take for granted your baseless accusation that the labor market research is only marketing and that they don't care?

My point stands, it seems much preferable to spend 20-30% on public benefits rather than some dystopian suppression of the public via robot armies.

You don't get to call me indifferent for not sharing your extreme pessimism.

At best it gets rid of some scams, certainly not most of them or poorly written apps, which continue to flood the Google PlayStore.

Despite Google's best efforts to automatically ban malicious developers, catching many legitimate developers in the crossfire, their store isn't exactly a bastion of quality or scam-free.

I give it 3 years at most, if the progress doesn't dramatically slow down.

I hope the policy measures Anthropic and OpenAI recommend in their research will be implemented in order to disperse the productivity gains enabled by AI widely, like they say.

I'm an optimist, but honestly, I think the dystopian future many pessimists expect is unlikely.

What would be the point for a few rich to concentrate even more wealth? Even if they were so selfish, surely it's worth spending some 20-30% on public benefits to prevent unrest, no?

If someone were to genuinely tell me that with the two SOTA models with the review step, severe security issues were left in the code, I would believe it.

That's not what anyone tells me though. I only hear theorizing and vague references to how AI failed people.

You're spot on that the thing I trust most is my own experience. I'm mostly working on mobile apps, my project does not have user accounts, so I couldn't tell you about authentication.

However, it is my experience that these models got very good at catching high-priority issues if you use the review step as I described.

Thus I theorize that SOTA models are likely to produce secure implementations.

I expected an interesting article about how most people underestimate change, but found nothing of the sort here, only a sarcastic comment.

Do people really think we will be working for another 20-30 years?

That seems very unlikely, and particularly so for software engineering.

If you're so concerned you can create such a benchmark and show us all how wrong my guess is.

Instead I mostly read such opinions based on active imagination and no concrete experience with using AI for such work in a setup as described.

It's not stupid if it works.

Concretely, I would like to see it benchmarked:

Have Opus 5.5 implement the authentication, with a 6.1 Sol reviewer. Out of 100 times, how often is the result insecure?

I wouldn't be surprised if it's zero times.

There is also the possibility of specifically prompting for a security audit on top, which would perhaps further lower the risk of overlooked issues.

Now one can say that due to the non-deterministic nature of the tools, there is always a non-zero chance of producing insecure implementation.

That's not really important here - obviously manual review isn't deterministic either, the risk will never be zero.

By now I've seen at least 5 different people or companies blamed for leaks or releasing fixes early.

If anyone is at fault here, I'm looking at the PHP maintainers.

The embargo apparently held for like two months, during which they have not managed to release patches.

Not of all games, but afaik at least for the first-party Nintendo developed games.

In Europe, the games are often available for cheap on Amazon.fr.

The pre-order for Ocarina of time was available for 46 Euro.

In what world would a Claude Code or Codex automation be appropriate here?

Even if the 15 minute interval wouldn't result in delays where no work is done and wasted invocations where none are needed, the UX of interacting with that as a personal assistant is obviously terrible.

This should be somewhat testable.

I'd set up a model with cyber capabilities, maybe GLM 5.3, not sure if it needs to be abliterated?

Disable web access and let it work on exploiting the pre-patch PHP version in a VM, given the info we had yesterday.

Can it do it? How long does it take, and writing the findings into a skill, how quickly can it be used on new targets?

I'd be interested in giving it a try, but unfortunately I'm very busy this week.

I'm struggling to see any reason to use PHP in the first place if it can run without access to private data.

At that point I would not need a backend.