interpol_p

It’s incredible some of the little details in LEGO instructions.

My son and I were assembling a set, and one of the corner pieces was strangely absent. My son asked, “why can’t we put this piece in yet?”

The answer came when we turned the page and saw the model had to be flipped. That missing corner piece was the only thing we could use to guide our placement of pieces on the underside of the model.

Similar things in more complex sets include using different, bright colours for bricks on the left/right sides of the internal parts of models. It makes it easy to keep your place as the model rotates through the steps.

Their docs seem entirely LLM written. It seems especially obvious in the FAQ. While I'm not against using LLMs for writing assistance, they've left a lot of the unnecessary language and typical stylistic choices in there, which erodes my trust in the project a bit. Perhaps it's a very good game-oriented version control system, but the lack of human attention on the docs makes me wonder how much they care

There are a lot of ways that AI speeds up software development processes that aren't the actual software development.

I am finding that lately I do not allow LLMs to write any code I am interested in maintaining. Or if they do, I have to micromanage them and it usually takes longer. They produce mediocre solutions, and often add redundant state ("Why did you add that state?" "Because we might need it in the future")

That said, they are extremely good at:

- Dev tools: creating debug tooling, debug screens, scripts that get the job done - Auxiliary development: landing pages, "what's new" screens, tedious boilerplate, gathering strings for localization - Prototyping: building full implementations quickly so you can see all the problems rather than having to anticipate them - Pure transformation: porting from one language or paradigm to another

So while I agree with the article that the actual spec of the feature you are building needs just as much human thought, regardless of AI, the speed-ups around that are worth exploring

An example I have from a recent feature development is adding CarPlay support to an existing app. We could have talked about it and designed it for weeks, but with an LLM I was able to get it running in my car in an hour, go for a drive, and feel it to understand whether it was a valuable direction.

The code was a mess, most of it had to be thrown away, and the LLM couldn't even get the initial build functional (not much CarPlay training data, I expect). But it was an accelerator to answer the question "is it worth investing more time in this?"

To repurpose a quote from Walt Disney, I don’t make software to make money, I make money to make more software.

I want my hobby project to be my job, because I don’t want to work for someone else. I want creative control, freedom to explore and ship ideas, and financial stability.

The only way to get there, that I can see, is to charge for my work.

I always wondered how this compares to the 1999 algorithm Texture Synthesis by Non-parametric Sampling [1]. The results look very similar to my eyes. Implementation here [2] — has anyone tried both?

[1] https://www2.eecs.berkeley.edu/Research/Projects/CS/vision/p...

[2] https://github.com/goldbema/TextureSynthesis

I was unclear. I meant it's not difficult to feel nurturing towards them and form a bond. That's what I thought the OP meant about being difficult.

Woah. As an adult man with five kids, two of them infants, the most natural thing in the world is for them to be present in almost every second of my life.

It’s not difficult at all. Minutes after birth, naked baby was on my naked chest, and bonding started. This never felt contrary to my instinct.

We had a contention between MCP / Skills for our product and ended up offering both. We built a CLI tool that could interface with the MCP server [1]. It seems redundant but our app is a coding app on iOS (Codea), and the issue with offering a plain MCP server meant that the agentic coding harness found it harder to do its job.

With the CLI the agent could check out the project, work on it locally with its standard file editing / patching / reading tools, then push the work back to device. Run and debug on device, edit locally, push.

With MCP the agent had to query the MCP server for every read and write and was no longer operating in its normal coding loop. It still works, though, and as a user you can choose to bypass the CLI and connect directly via MCP.

The MCP server was valuable as it gave us a consistent and deterministic language to speak. The CLI tool + Skill was valuable for agentic coding because it allowed the coding work to happen with the standard editing tools used by agents.

The CLI also gave us device discovery. So the agent can simply discover nearby devices running Codea and get to work, instead of a user having to add a specific device via its IP address to their agent.

[1] https://codea.io/cli

I tried to live like this for a while but found I could not separate applications into spaces

I would try setting up a space for, eg, all my communication stuff. But suddenly I’d need to drag-and-drop an image from my image editor into Slack. Or I’d want to drag a graphic from Safari into Final Cut Pro. Or any number of cross-workspace operations

How do you handle this with spaces? Do you initiate the drag, tap the space hot key, then drop?

I had Opus 4.6 running on a backend bug for hours. It got nowhere. Turned out the problem was in AWS X-ray swizzling the fetch method and not handling the same argument types as the original, which led to cryptic errors.

I had Opus 4.6 tell me I was "seeing things wrong" when I tried to have it correct some graphical issues. It got stuck in a loop of re-introducing the same bug every hour or so in an attempt to fix the issue.

I'm not disagreeing with your experience, but in my experience it is largely the same as what I had with Opus 4.5 / Codex / etc.

The example given in the article is acceptance criteria for a login/password entry flow. This is fairly easy to spec-out in terms of AC and TDD.

I have been asking these tools to build other types of projects where it (seems?) much more difficult to verify without a human-in-the-loop. One example is I had asked Codex to build a simulation of the solar system using a Metal renderer. It produced a fun working app quickly.

I asked it to add bloom. It looped for hours, failing. I would have to manually verify — because even from images — it couldn't tell what was right and wrong. It only got it right when I pasted a how-to-write-a-bloom-shader-pass-in-Metal blog post into it.

Then I noticed that all of the planet textures were rotating oddly every time I orbited the camera. Codex got stuck in another endless loop of "Oh, the lookAt matrix is in column major, let me fix that <proceeds to break everything>." or focusing (incorrectly) on UV coordinates and shader code. Eventually Codex told me what I was seeing "was expected" and that I just "felt like it was wrong."

When I finally realised the problem was that Codex had drawn the planets with back-facing polygons only, I reported the error, to which Codex replied, "Good hypothesis, but no"

I insisted that it change the culling configuration and then it worked fine.

These tools are fun, and great time savers (at times), but take them out of their comfort zone and it becomes real hard to steer them without domain knowledge and close human review.

That's a pretty extreme take. I've been using the Mac since about 2001. I like Tahoe and a well designed Tahoe app can look really nice on the platform. There are bugs, inconsistencies and other issues, but it doesn't feel that different than many previous macOS / OS X releases

I believe you can do regular hard edged intersections. You can see in his operator list some are listed as “smoothSubtract” and some are just “subtract”

It’s just easy to do the melding thing with SDFs so a lot of people do it

The reason this happens is because big companies get their software pen tested. Part of the pen test report will include something like “accessible from jailbroken devices.”

The pen test results get put into the ticket system as immovable entries. Engineers will question them, only to be shot down by the cyber security department who organized the pen test. The engineers will eventually accept that they cannot convince cyber to drop the issue, and implement the jail break detection.

Why does cyber mandate it? Because no one in a large company wants to accept the risk, even imaginary risk. They want to be able to say, when security is breached, “we did our due diligence. Look at the report, we implemented everything in it”

Why do firms offering penetration testing keep putting junk like this into their reports? Because their automated tools list them out and they’re getting paid to find issues. The more the better.

It’s insane and entirely about passing off risk.

Depends what you see as “abusing” the system. By working from home, I can take a walk in the garden when I find it hard to think, it energises me. At my office I can (and do) take a walk in the car park, but inevitably I leave the office with a headache caused by constant noise and fluorescent lighting

At home, I can put my family first if needed. When I’m at the office and something comes up at the kids’ school that I need to deal with, it’s a mad dash to get away soon enough that I almost have to drop everything and run

The times working in the office has been good as a software engineer: when we are prototyping on physical hardware I do not have at home. That’s it

It’s great if people love to go to the office. That’s fine. It’s managers that enforce it who are the problem — the people who work for you aren’t children and if you feel like you can’t trust them to make the decision to work from home, why on earth would you trust them in your office?

It's fast, but it's not that fast.

My son regularly borrows my iPhone 14 Pro for shooting video, and I inevitably have to do a large AirDrop transfer to him of all his footage. We usually see about 10 GB per minute, which is really fast

Sure. I catch up with many of them on weekends anyway — we hike together, our families know each other, some live nearby etc.

Regarding knowledge sharing, that happens equally well via Slack. (Actually, I'd say a screen share works better than over-the-shouldering someone else's screen in person)

We have some sort of hybrid policy. Every single time I have showed up at the office, I either end up socialising far too much and get nothing done (I find it extremely hard to work next to people without talking to them).

Or nobody is there and I end up having driven (40 minutes each way) to the office to have Teams meetings with a wonderful view of the car park, under fluorescent lights, using a cheap low-resolution office monitor. When I could have been having those Teams meetings with a view of my garden and a much nicer monitor I have invested in

I'm in the market for this

I've been hoping for Apple to return to "thin" and it's nice that they're trying. I don't know whether I would buy this, but my current iPhone 14 Pro feels like a brick — thick stainless steel

When I go for a run, it's uncomfortable to have in a pocket depending on what running clothes I am wearing. The heaviness makes it feel far more likely to break all the times I have dropped it (and I have dropped it many times, without a case)

I really like the brevity of text-davinci-001. Attempting to read the other answers felt laborious

Safe area can account for things that are not just a notch. It's used across Apple platforms to indicate anything that might need to occupy a dedicated region on the screen: notch on iPhones, home indicator, iPadOS traffic light buttons, menu bar, curved edges of displays, and so on

Your container views can extend the safe areas for their children as well. In our apps, which allow users to run their own custom projects, we increase the safe area for our UI so that users can avoid it in their own rendering

Safe area is a fairly neat and functional API. The unfortunate thing is the older `CGDisplayCopyAllDisplayModes` API is just lumping all resolutions together

I've been diving into Claude Code after reading articles constantly praising its abilities. But I think perhaps it's better suited to web development

Using it for iOS development is interesting. It does produce working output (sometimes!) but it's very hit-or-miss. Recently I gave it a couple hours to build a CarPlay prototype of one of my apps. It was completely unable to refactor the codebase to correctly support CarPlay (even though I passed the entire CarPlay documentation into it). I gave it three attempts at it. Then I intervened and added support for CarPlay manually, following that I added a lot of skeleton code for it to flesh out. Claude was then able to build a prototype

However, over the next few days as I tried to maintain the code I ended up rewriting 60% of it because it was not maintainable or correct. (By "not correct" I mean it had logic errors and was updating the display multiple times with incorrect information before replacing it with correct information, causing the data displayed to randomly refresh)

I also tried getting it to add some new screens to a game I develop. I wanted it to add some of the purchase flows into the app (boring code that I hate writing). It managed to do it with compile errors, and was unable to fix its own build output despite having the tools to do so. Instead of fixing the build errors it caused, Claude Code decided it would manually verify that only its own changes were correct by running `swiftc` on only files that it touched. Which was nonsense

All that said, there was a benefit in that Claude Code writing all this code and getting something up on the screen motivated me to finally pick up the work and do some of these tasks. I had been putting them off for months and just having the work "get started" no matter how bad, was a good kick start

I don't have sympathy for them. I just happen to agree that it's a lot of engineering work to comply now.

Perhaps if they had opened parts of their infrastructure much earlier, they wouldn't be legally compelled to do it now, and have to invest significant resources to do so.

Yeah, who knows if the EU would see it that way. They may require Apple to provide first-party APIs that are equivalent in power to what they offer developers who submit via the App Store. Either way, my post was pointing out that it is non-trivial engineering effort to do this, and I think that's still the case.

Hell, just releasing my own personal code as open source — auditing it, decoupling libraries, removing internal stuff, it's a huge multi-week effort for me to do. For any company with as much code as Apple, it's pretty daunting

Sure, I'd be into that. But that would not comply with the DMA I think? As in, Apple still has a ton of work to do, engineering wise, if they are to make their platform available to all in the way specified by the DMA

For example, I don't think it would fly that they could say to the EU: users who want a third-party browser just have to enable root access and lose access to all Apple services and authentication

That is an oversimplification of what I stated.

Apple has a significant engineering challenge to turn their current operating system into something that allows side-loading similar to what Google offers. It's not a matter of "commenting out an if statement"

The current developer SDKs Apple offers are strongly tied to their services, which cost them money to run. So first thing is, they have to decouple that so developers can implement applications using a baseline SDK that does not use Apple services (no iCloud, no Maps, no HealthKit and so on)

I think it would be great for users if they did do this. It would be akin to what Google does by shipping and updating Play Services separately from the base Android install

The reason I linked BrowserEngineKit is because if you want to do this properly, you have to build something like Apple has built with that framework (which was built to comply with these policies). Take for example, implementing your own JIT: because arm64e uses pointer authentication, the system uses PACs to ensure that pointers into executable code have not been tampered with. Apple now develops and supports a whole slew of APIs like `be_memory_inline_jit_restrict_rwx_to_rw_with_witness()` in order for developers to manage this themselves.

You saying "just let their pocket computers run software users download and install" is not like every single other computer ever made and sold. This is a gross oversimplification of the modern state of computing, both on mobile and on desktop. There are reasons you don't want random developers loading code into your OS kernel, and Windows and macOS both have protections for this (though the CrowdStrike crashes recently shows what happens when those protections are lax!)

Google engineered and maintains the system that allows you to install APK files. This is my point. The fact that they have developed a security model around APK updates is exactly what I'm talking about.

If Apple wants to offer something similar, now, they are going to have a lot of work cut out for them.

You're not thinking this through, it's not a magic button Apple presses. They are going to have to develop a ton of frameworks just to get something like installable APKs.

Apple allows developers to use iCloud and Maps for free. Presumably because you distribute through the App Store. So if they allow for side-loading they're going to have to lock down and split their App Store "services" into a separate framework — hey, sounds familiar? Just like Google Play services.

Separating out all of Apple's authentication layers, paid and cloud services, and ensuring apps can be cleanly distributed without dependencies on those things it not a trivial engineering exercise.

I'm not trying to imply that Apple should not comply with the DMA. I believe they should. I also believe that it would be a seriously complicated thing to extract their App Store services from their developer APIs in such a way that people could develop against a baseline SDK sans Apple services.

You missed my point. My point is that if Apple wants to add this now, it's going to cost them engineering resources.

You think side loading on Android cost Google "nothing" to implement and maintain? No, it costs them engineering resources to support that feature. It's a good feature to support and it's beneficial to users. But it's not free, it doesn't magically insert itself into the Android codebase if they "comment out an `if` statement" as the GP suggested.

Also, Android is gradually adopting many iOS-like permissions and security models. We recently updated our Android apps related to reading and writing to the file system. Why is that? Because the free-for-all they shipped with was heavily abused by developers.

It’s extremely complex. I’m not debating whether they should comply - they should. But it’s gonna cost them years of engineering effort, and maintenance far into the future. See, for example, BrowserEngineKit

https://developer.apple.com/documentation/browserenginekit

They needed to engineer, maintain, document and support a whole class of APIs so that third parties can create their own competitive browser engines (that offer JIT, etc) while still maintaining iOS sandbox security. There are going to be hundreds of frameworks, thousands of APIs, that will need to come to ensure compliance with the DMA

I wouldn't say it's "damn buggy" — I use Notes daily, and have a significant number of notes that are synced between devices. In my notes I use rich formatting, embed videos, voice memos and lots of images. It handles it really well. I even use iCloud Collaboration feature on a few notes for planning, and for splitting regular expenses

I have three notes that exhibit the bug you mention though: the three notes I keep for each of my children's artwork. I scan the artwork using the document scanning tool in Notes, and it gets embedded as a multi-page PDF (if the artwork itself has multiple parts) or a single PDF. After many years of adding high-res scans, when I scroll to the bottom of these files it takes some time for the note to render. I think I picked the wrong tool for the job here, more than anything!