simondotau

The problem with Grok 4.7 is that it's not amazing at anything, and didn't even retain the blazing fast performance of Grok 4.5 which made it awesome for highly interactive workflows.

But despite all the negative coverage, I can say with confidence that Grok 4.7 is a great code reviewer. You know who says that? Every other model. A lot of my time is spent orchestrating code reviews. I will give the same prompt to three models, then give each model the responses from the other two models to assess and rank them. Finally I give the rankings and meta-rankings to Opus to synthesise a final report.

Right now, for the code I'm dealing with (embedded c++, python, swift, and CFML) I'm fairly consistently seeing every other model rate Opus 5.5 first, with Grok 4.7 a very close second, often finding real issues other models miss.

Kimi K3 has been my go-to model for its writing style and (prior to Muse Wonder) my go-to for a third opinion during multi-round/multi-agent analysis tasks. Its writing style is IMHO so good that, for code which humans need to review, I regularly got it to clean up the painfully verbose comments left by Opus.

I'm excited to see if its writing style and unique takes have been retained in this new release.

Late to reply, but for the sake of posterity:

You are correct that many developers complain about Safari because they are lazy little shits who would much prefer the whole world be forced to change which browser they use in order to make their jobs a bit easier.

Nobody is saying anyone needs to test on an unpopular browser. Safari on iOS is, by the numbers, very popular.

Or maybe your opinion of Grok and its affiliated entities is slanted, and mine is straight. Without an objective reference point from which to measure deviation, all you're measuring is the relative angle between our opinions.

It's also worth noting that, living in an unfavourable time zone relative to most HN activity, I'm rarely here to offer the opening opinions in a discussion. More often I'm arriving 12 hours into a conversation, where everything I'd want to say has already been said. I end up contributing only where I think a different perspective is missing. Naturally, that means my comment history will disproportionately reflect the rare instances where my opinions buck the prevailing HN orthodoxy.

If you disagree with a particular claim I've made, by all means challenge it on its merits. That could be the basis of a much more interesting conversation than this fatuous one.

Always fascinating to see someone so allergic to the idea of an opinion differing from their own that they immediately invent a covert marketing operation to explain it. Someone liked a model you don't like? Impossible! Clearly, there must be a conspiracy.

Of course, it should go without saying that my comment was a genuine, good-faith account of my real world experience with these models. Not that a public reputation seems to count for much these days, but for what it's worth I have over 25 years of public involvement in the tech sector, and I'm posting under my real name and real identity.

Opus 5.5 is a god tier model, but for serious large scale work, nothing beats getting at least three non-similar models to repeatedly refine a super detailed plan before any code is written.

Right now my consortium is Opus 5.5, Grok 4.7 and Muse Wonder. These three are different enough that their insights are very complementary. Once a plan is detailed enough, any decent model can execute on it successfully, but I personally like Grok’s ability to stick to the plan. (And no 5-hour quota!)

Hopefully someone has pointed this youtuber at Rick Beato / Legal Eagle's initiative in this direction.

https://www.youtube.com/watch?v=M8HxQkk0PSY

(FWIW my day job directly involves that Act.)

Beach's ruling determined that Apple should not be allowed to block, as a policy, competing app stores. Beach also recognised that Apple's commission is not merely a payment-processing fee, but a means of recovering its investment in the broader ecosystem, including developer tools, security, app verification, distribution infrastructure and intellectual property. He also acknowledged Apple's entitlement to a commercial return on those investments.

The question remains how Apple can get the commercial return which Beach acknowledged Apple is entitled to, given the weirdly arbitrary restrictions defined in Beach's order. This contradiction is left unresolved and is begging for an obvious overturn on appeal.

I personally have no problem with competing stores, so long as they're not just a mechanism for redirecting money previously going to Apple into their own pocket. But this ruling doesn't establish competition, because there's not competitive reason for alternative stores to exist except to avoid fees which Beach acknowledged Apple is entitled to.

I’d be happiest if they were all brought to heel at exactly the same time. We don’t need precedent, we need legislation.

As an Aussie, I hope Apple maliciously complies to the greatest extent possible until the same rules are applied to all companies equally. And I mean ALL. Every single computing platform with any kind of loadable software developed by any third party should be similarly unlocked.

If the rules are good enough for Apple, they are good enough for everyone.

This might sound like a defence of Apple, and to a small extent it is. But actually, I’m being deadly serious. I think the world would be a better place if these rules applied to everyone.

“But poor Nintendo” rings as hollow to me as “but poor Apple” probably does to you.

Either way you're having a day or more of downtime. No reason to assume today is worse than tomorrow.

> From a Machiavellian standpoint

From a marketing standpoint, having the first model to find a bug of this magnitude and severity would be a pretty damn impressive proof of capability. Finding previously unknown, critical vulnerabilities in widely used open-source software could become a gold-standard test of frontier AI competence.

To be fair, the fact that you gave it the task at all is a clue that you know the answer might be uncharacteristically simple.

Not only should they never be trusted again, the individuals responsible can't either. They need to be blacklisted somewhere. This wasn't garden variety stupidity, this was (IMHO) do-not-collect-$200 stuff.

Absolutely they do. Off today is better than a ransomware attack tommorrow.

I would have expected there to be a hundred entities running every single model over major code like PHP, and clang, and pretty much any behemoth open source project the world relies on. This is the sort of thing the model makers should be doing as a matter of course, as part of routine testing of their new models.

Undoubtedly, but presumably not this specific set of vulnerabilities.

Once the dust has settled, it would be interesting to give LLMs the pre-patch source code and see whether any of them can identify this chain of vulnerabilities. Even if they can't identify the full chain, can they at least spot one of the constituent bugs and recognise its seriousness?

If they can't, why not?

If they can, why haven't they already?

By that definition, every decision is political, which tells us nothing about whether Cloudflare's decisions were actually politically motivated.

Correct -- and in addition, most people who buy an EV have never owned a petrol car with comparable performance. The vast majority buying EVs aren't coming from a Golf GTI or a Subaru WRX. They're not even coming from a BMW 340i or some high-trim Mustang.

It’s not pointless when it allows an Apple developer to rapidly mimic the exact environment, in case it’s an extremely narrow bug involving a second-supplier controller chip.

What is political about mass murder or maliciously distributing stolen intimate photographs? Seriously. I want an answer.

What's the falsehood in that statement? Did you catch that person actuating any of the controls manually? Again, this was a demonstration of the actuator hardware being sufficient and the underlying concept being possible -- ironically relying on the inferior pre-mapping crutch that Waymo still leans on.

Despite its overly optimistic thrust, the meat of the video was proven correct over time, with the main valid criticism being that Tesla in 2016 underestimated the amount of compute power which would ultimately be required. What's incredible is that FSD v14 Lite demonstrates what's possible with HW3 despite having fairly potato-quality cameras.

Everyone knows the robot demonstrations were remote controlled. It was a demonstration of the hardware capability, not the software.

You do know that bulletproof doesn't mean zero surface damage right?

[flagged]

Exactly. Watch what they do.

Even in a degraded state, Starlink and Starshield remains a critical strategic advantage for Ukraine, even if it's less so now than before. Being less helpful to Ukraine isn't good, but it's still actively helping Ukraine, and that is worlds apart from actively helping Russia.

This has nothing to do with semantics. I'm arguing ideas. If you'd prefer different words to express those ideas, I'm happy to use them. Let me know what you'd like me to change.

Power is instantaneous. A solar panel doesn't get the same credit for generating electricity when there's already a surplus as it does when that electricity is needed. I don't just mean credit in principle, I mean literal credit. In a functioning market, the marginal value of surplus electricity tends sharply towards zero, potentially becoming negative. Solar generators have almost no marginal cost of production, so there's little reason to expect producers to withhold generation simply because the market price is approaching zero.

Batteries don't just "move existing power around". They can absorb worthless renewable generation that would otherwise have been curtailed, then supply it when the grid would otherwise fall back to coal or gas.

It should also be noted that the remaining solar production doesn't exclusively represent the curtailment of emissions-producing generation. It can also represent the curtailment of other zero emissions sources like wind. Probably not to a significant degree, but I'm sure it's not zero.

A kWh is a kWh in terms of energy. It isn't necessarily a kWh of equivalent economic value, emissions reduction or contribution to grid reliability.

> FSD is lacking at edge cases, such as parking

From videos I've seen FSD is perfectly capable of parking. It doesn't yet have the ability to reliably discern which spaces will reliably avoid parking fines, but it can park just fine.

This really gets to the question of what qualifies as driving. A robotaxi can't top up its own wiper fluid either, but keeping that topped up is, IMHO, a critical part of being a responsible driver.

> if Tesla was so confident, they would take on the liability, like Waymo does

This is a non sequitur. Both operators accept the same liability under the same conditions. Both Waymo and Tesla accept liability for vehicles in their own autonomous taxi fleets. Neither Tesla nor Waymo have proposed any mechanism to accept liability for customer-owned vehicles.

IMHO transferring liability for customer-owned vehicles is a deeply non-trivial matter, and I doubt it will ever occur in any absolute sense. For example, how does liability shift if an autonomously operated vehicle crashes into a pedestrian because it was raining and the tires were overdue for replacement? What if the crash occurred because of prior damage to the suspension which the owner knew about and chose to ignore?

> Now it’s just a fun game to see how bloodied you get.

As I said, pure debate bro.

[flagged]