We use gitlab at work, and I have to say that it is disappointing. The UI is so complicated to do the simplest things (e.g. to approve a MR you need to click a button that is actually a menu; the diffs are difficult to read; the 'To-do list' includes MRs that were already merged (how is that actionable?)) and it seems that they're struggling to turn around improvements quickly. The issue around the 'To-do list' including MRs that were already merged was raised years ago.

I also have to say that I'm surprised about the backlash against bitbucket. I find the UI incredibly simple and clear, as do all of the new joiners. With Script Runner you can do some pretty amazing things. It handles the huge repo's well too.

> to approve a MR you need to click a button that is actually a menu

It's not really any better on Github. Why do I have to click on "Files changed" to approve a PR? Overall I would say it is on par with Github. Worse in some ways; better in others.

We semi-recently had Gitlab reps stop by at work to demo their new AI features (we're already running Gitlab Enterprise). It was a strong signal that we should get off the train before the next major version. None of the features seemed very thought out and to top it all off, there was an outage while they were demo-ing it, leading to broken pages and no LLM responses.

> e.g. to approve a MR you need to click a button that is actually a menu

I think you're referring to the "Your review" button in the upper-right that's not always there, but there's a plain "Approve" button on the Overview tab that doesn't open any menu and is always there.

try to find the issue boards.. its a mess. And expensive.

First step is to get ssh setup correctly, and second step is to enable a firewall to block incoming connections on everything except the key ports (ssh but on a different port/web/ssl). This immediately eliminates a swathe of issues!

Also use fail2ban. If nothing else to decrease the amount of junk in logs.

I think the better question to ask is what services peak at a few hundred transactions per second?

I think it depends on what you're using it for. If it is a simple kubernetes config then the model doesn't matter too much. Contract that with writing the scenario for a backtest for an algo that trades on a venue: it is not the same complexity and the basic models are terrible. I've had it tell me that it has added tests to find that they're just stubs! Opus seems to be getting there, but on more complex tasks the others are a complete waste of time.

> If it is a simple kubernetes config then the model doesn't matter too much

I guess at least this person https://www.tomshardware.com/tech-industry/artificial-intell... might disagree. I think already to know what Kubernetes even is requires quite a bit of knowledge. Using a tool that manipulate its configuration files IN PRODUCTION without risking data loss is another ball game entirely.

I think that hearing other points of view is important.

I see time and time again that posts insinuate that there's no other point of view, and I think that highlighting one perspective is enough to show that this (as in the article) isn't some mathematically perfect piece of advice.

This is my intention, thanks. It's also not to say that my opinion is perfect either, it isn't.

I think that for the average developer this might be true. I think that for excellent developers, they spend a lot of time thinking about the edge cases and ensuring that the system/code is supportable. i.e. it explains what the problems are so that issue can be resolved quickly; the code is written in a style that provides protection from other parts of the system failing; and so on. This isn't done on average code and I don't see AI doing this at all.

I thought that new models were typically released in October. Have I misremembered or is this an unusual timing vs previous years? If so, I wonder why the earlier release?

They didn't update them last October is why.

I think at this point Apple will just release new versions of laptops whenever new CPU revisions and yields allow. M5 Pro wasn't ready for October so delayed until now.

You remember well, they didn’t update these last fall.

And another rumor said these are going to be updated again this fall but I’m not sure about that. With OLED screens and M6 (supposedly).

Increasing component prices perhaps? Get some sales in before you have to jack up the sale price.

Prices aren’t likely to change. Even when the tariffs were on, Apple’s prices didn’t change; they gave up some margin.

They also probably had RAM contracts in place far enough in advance to avoid the worst of the price spikes.

Maybe they want people to have more money available for the new phones later this year, since that market is in decline.

M6 is rumored to be released in Q4.

I did it today. It took me thirty minutes to fix the networking because I couldn't get Little Snitch to uninstall since it didn't seem to be compatible. Basically I had to reboot in recovery mode to disable security features (csr) to uninstall Little Snitch (via systemextensionsctl). This is the worst update I've ever gone through on Mac, and I started using a Macintosh SE.

In the past, MacOS has automatically made a folder of incompatible software that it leaves on the desktop. Little Snitch seems like something that could have been tested.

As far as I know Little Snitch uses a user-space Network Extension, which uses a public API added by Apple to the OS specifically to help devs move out of kernel space.

It is a commonly used extension, which was restricted so that I could no longe remove it. What's more odd is that supposedly I was running the latest version.

Considering that new speakers don't use SoundTouch, I wonder too. I hope that they keep the app running for a while. This kit is expensive and it can't have a short life time!

Your point was thought provoking. In problem solving the "what" and the "how" are orthogonal since the method doesn't dictate the goal. However, if it takes a long time to do something (how: working slowly eg. because you're coding on a very slow, old machine) it tends to predict that there's less accomplished (what). That suggests that this isn't 'fully' orthogonal.

Someone else raised a good point that if we're working on the wrong thing, it doesn't matter how fast we are. However, I think a more subtle interpretation is more helpful here. I think that we need to be clearer about the consequences of the outcomes: what's the value add. The way I often reason about that is whether the outcome is 'Long-term greedy' or whether the outcome is going to make us a million dollars now. I find the latter really helpful, because if we're going to make a million dollars now but it costs us 100K in tech debt, then (provided there's not a better use of the resources) that is likely a good cost-benefit outcome.