The report included a patch, but the patch was completely wrong based on a random guess of how a certain software could behave. It 'worked' for the reporter by also swallowing all errors and silently continuing, despite the particular call failing. It was a check to see if a particular action would be needed, and so it silently assumed the step would not be needed, which happened to be true for the reporter so it looked 'fixed' to them.

Problem is that there are more than two people that are stakeholders, and and it would be nuts for me to just let the codebase get screwed up to try to prove a point for one very minor stakeholder who won't even get bitten by the mistake because their situation doesn't hit the bug inflicted by the change. If I let a random user patch a codebase, I rightfully get blamed, not them. And AI assessment of the report would likely agree with the submitter, because the AI generated a plausible rationale, and AI gives the benefit of the doubt to the material in the prompt fodder.

There actually is a project that I used to be a part of but no longer use or work on, but left my notifications active, because why bother changing and it was largely a "done" project, almost no new issues, not much in terms of code changes. Some guy decided to go all in on LLM giving free reign. Was not confident to be a developer but declared AI made it possible for him to change things. Suddenly lots of code changes, and also a sudden flurry of external issues of users who had been using it for a long time asking why random stuff was broken. Near as I could tell when I got around to disabling all the notifications, he just kept feeding issues into LLM and releasing updates and getting a new wave of issues and repeating. What was a boringly stable project became an eternally random messed up mess because he let LLM take over everything and it breaks as much as it helps. I'm not letting any of my active projects turn into that.

Which explains a lot of very overly verbose "reports" that have been sent to me about "issues". The person thinks they are using AI to make their report more useful, but I still find the human input to be sufficient and the AI reports to be a pain to actually work backwards to the issue. Especially when it decides to fixate on what it thinks is the root cause and gloss over what the actual problem was... Which has at least once become super counter-productive, as the report got it flat out wrong and the reporter refused to give me the usual debug information because "Claude already looked it over and gave the answer, can't you just accept its findings and figure out how to fix it?"

Yes, you can use their app to pay. It sucks though.

For one, my watch can tap to pay, but not with walmart app.

For walmart, I have to open the app, wait for it to load, sometimes it randomly decides I need to "prove I'm a human", then as I'm trying to go to pay, it pops up a "you want walmart+ now!" and the "yes" button is close to where I would have tapped so I can imagine they get accidental accepts. Then finally I click and it opens the camera app to scan the qr code, which is an annoying approach in general. Even worse, their qr screen is glossy and I happen to catch a reflection of one of the glaring lights right where the code is unreadable.

Their app to purchase was just a mess of bad UI and poor store setup for a qr code based solution, very clunky compared to nfc type approaches.

Note I said incremental cost of a remote area in particular. Launching satellites doesn't even narrow down the investment to a single nation, let alone focus spending on the communities underserved. So we can look forward to subsidizing Starlink global infrastructure with money intended to improve options where they are lacking.

This is exactly what it is, Starlink can't do gigabit, can't do low latency, but *can* reach random places easily. Musk and Trump seemingly have made up relatively recently and I am certain this favor is a consequence.

So with this Starlink gets access to funds that are generally made available to give *quality* access to especially rural places. Money to make up for the relatively high incremental cost of wiring up an area compared to the revenue. Except Starlink has effectively zero incremental cost to serve these areas, so they get to collect the same money as a provider would have had to spend to actually run real wires.

I thought this administration considered fairness "woke", but for outdated technology, *suddenly* we need to be fair?

We should stop airplanes from flying, it's not fair to horse saddles.

I think that remains accurate, but it depends on the domain.

Search had always been free form text input, and that extends to this, open ended interaction with impossibly large dataset to browse and impossibly many things to request. Places based on text interaction are places where LLM have a good chance of having a large role.

However, this isn't the total world. For example, LLM will only be as good at taking an order as a drive through speaker. I greatly appreciate managing an order through an app or kiosk, where I can more quickly peruse options in detail, more quickly knowing what's in it, price, calories, etc. Maybe an LLM option when you know generally what you want to fill out the order then a UI to tweak details.

Tons of UIs are more than some compromise for helping a computer be a consolation prize for not bothering to have a human handle a request.

Lots of AI companies seem to be forgetting this and imagining that devices that are just microphones and speakers will take over, but they are incorrect in many cases.

I don't know about mental health, but at least MIT's policy seems to not detract from it's students reputation in the industry.

I guess the question is why you would say "Would you fly on a rocket designed and built by people who could not fail?" based on the current topic that UoM is adopting MIT's long-standing strategy to first semester grading. You seem to infer that UoM adopting this will produce bad engineers that can't be trusted to design and build a rocket, but MIT has been doing it since the 1960s. Why is it so terrible for UoM to do it and it isn't a concern that MIT does?

Well you have Google, and of course Apple could decide to in-source the AI engine if it is so important, they tend to like doing that.

But when it comes to software development, Anthropic certainly has the advantaged position in reputation. However more broad use cases I think are supremely vulnerable to the platform owners of the endpoint devices changing the default. Why do Claude when the hardware button already chats with "some AI". Why leave office 365 that you already are running when office 365 will frontend whatever AI microsoft feels like? Even in the field that Anthropic is very well respected, I know entire companies that only use it by way of Github copilot because their company has a financial arragement with Microsoft. Leaving even users of Anthropic models at risk of Microsoft shuffling things around to use a model of Microsoft's choosing.

So even if Microsoft sucks, Microsoft dictates a bunch of defaults. Even if Google sucks, they dictate a lot of defaults.

Musk, Altman, and Zuckerberg are probably the ones I see at highest risk of failing to capture enduring success in this market. Anthropic certainly has the technical trajectory to be some sort of durable success, but just think there's a lot of potential for others to act as gatekeepers to the ecosystem.

While Anthropic is at least more substantial than other players, this *absolutely* is suspiciously heavy in this summary.

They anticipate that their revenue *could* be $100 billion if they extrapolate some revenue interval of the entire year, and that they will follow up with an unspecfied duration of year over year growth rate of 800%... Which just is an absurd assumption when the starting point is $100 billion.

No, it says:
"It reportedly delivered more than one megawatt (MW) of power during the flight."

At some point during the flight it went over a MW.

Since you are so convinced that universities that do not assign GPA on first semester produce bad engineers, does that mean you have distrusted MIT educated engineers for the last 58 years?

Strange no one has been bitching that MIT has been coddling students for 58 years...

One kilowatt isn't generally priced as anything. One kilowatt-*hour* is, however, priced.

So they may have actually used this properly, 1MW to take off sounds plausible for an ambitious aircraft in terms of power to take off and climb, but average of 100KW over the flight including the descent to get to a rough area of $5 over a half hour flight (well, still that's a pretty cheap energy rate by modern standards, but it's order of magnitude at least).

If it's a 100 level physics class and it matters, then the student will go through at *least* another semester or more of physics. I would care a lot more about their more advanced versions of the classes than the basics that every random university attendant takes.

If it's a 100 level class without a follow up, then no, it does not matter to the core of their curriculum.