MontagFTB

I’d like my Pip-Boy in matte black please.

I have split up the data model into a handful of layers in order to separate concerns. The evidentiary layer are the raw bits - the media, sources, clippings, photos, etc. From there I derive "personas", which are the (incomplete and possibly inaccurate) representations of people from those media bits. Above that is the curation layer, where the curator takes those personas and compiles persons from them.

So if there are a bunch of articles that talk about someone's birthday but a couple give conflicting data, that is all present and recorded in the evidentiary layer, and it is the responsibility of the curator to decide which is the actual birthday for that person.

edit: the "below/above" terminology is a dependency relationship. The facts undergird the conclusions drawn from them. To flip that relationship would be to "retcon" facts to fit the conclusions a curator has already drawn, which would be a mistake.

I’ve written my own genealogy app as well, it has reinvigorated my efforts. I have a multi-level data model (evidentiary below, curated above) that has helped me piece together people across a multitude of disparate sources. Claude has been my sidekick, and it’s amazing- especially for the more toilsome tasks like extracting people’s details from newspaper articles.

Their infrastructure is buckling under increased demand from AI generated code.

Imagine the software development lifecycle is a river blocked by several dams- necessary processes that also slow the water’s volume and speed. Writing the code was a very big dam upstream, and successive dams didn’t need to handle a lot of water. With that dam busted wide open, the torrent of code is inundating downstream processes. GitHub happens to host many of them.

Earlier this year the GitHub COO shared a post on Twitter with explosive numbers they were facing. I cannot seem to find it at the moment. (Edit: shared in other replies to this thread.)

I think I’ve seen that movie.

> it just reads as more evidence that LLMs are better at replacing leadership than anybody else.

I believe this counts as having “more confidence in AI’s advice than in that of colleagues”

In a 30M+ LOC repo we blew out Qwen3.6’s context thinking about a code review. This was via direct call to model. Do harnesses facilitate better context management, or is there something else to accommodate its smaller context window?

I have heard of some high school classes reverting to typewriters. So your suggestion holds some weight.

TLDR proof of care from the article: a low bandwidth process (e.g., from handwriting to tattooing it on your body) that you voluntarily put your words through to convey their level of personal importance.

Some of his examples were tongue in cheek. But even handwriting feels a little too laborious when what we lost that needs replacement is manual typing.

This article was tremendous. My dad and I rebuilt an old GPW as my car in high school. His brother bought it as his car for high school a generation earlier. This article and the video are a love letter to that old car- a quirky, heavy, underpowered, beautiful legend.

Do you have links to what you found? That solution sounds plausible in this case.

iPhone Dynamic Crease (tm)

Yes, that’s a great video. The Mythbusters used the same model to understand Bay Area current flows during their Alcatraz escape attempt episode.

Multiple articles mention a vaccine for the Zaire strain but not this one. Is it possible to use one for the other? Does the existence of one make it easier to develop another?

These changes were part of the Beta program. As far as I am aware the response there was not on the same level as this blog post.

Thank you. Differing opinions of Adobe aside, you sound like the kind of engineer I'd enjoy a beer with.

Thank you for the constructive feedback. I agree power users are a "keystone species" and improving their workflows will benefit everyone.

Improving existing workflows without disrupting them is extremely hard to do, and often "improvement" is in the eye of the beholder. To be clear, I am not excusing issues within the application that we must fix. The team is working hard across multiple departments to gain consensus on how best to move Photoshop forward, including gathering feedback from users.

There is a lot of effort put in to making the application usable, no question. At the same time, we have added a litany of new features and tools since Photoshop 4, many of which I would describe as extremely technically challenging.

Well it depends on how you define it. We have done several re-skinnings. This migration is a transition from one framework to another, which we’ve never done before. The reasons to leave the old framework lie on several axes (eng, design, product, etc.)

In a word: they were. We do use Photoshop (though not to the level or extent of most users) and noticed the regressions. Shipping software is in perennial tension between getting it perfect and getting it out the door.

Going forward, we would like them fixed, too. Personally my hope is the message from user feedback like this is heard loud and clear, and we respond appropriately.

Hey all- I am Foster Brereton and Principal Scientist for this UI effort. Suffice it to say, the article and this thread have had their impact on the people behind the software. We are aware we got a lot of things wrong. As the primary technical lead on the UI migration, a lot of the implementation details ultimately fall up to me.

Two things I can tell you: the engineering team does care about Photoshop (I’ve been on the team more than 15 years for a reason) and this migration is far from over for us.

These sharp edges are acknowledged, and we are working on them. Some of them are already addressed.

I know this will be of little comfort to some. But to the rest, we are still here. If you have any questions I’ll do my best to answer them.

Visually the results are very compelling! It also gives an at-a-glance intuition about the image that the bar-style options fail to convey. I am a fan.

Could it be a symptom of high altitude oxygen deprivation?

Putting code with side effects into an assert is asking for trouble. Compile with NDEBUG set and the effects mysteriously disappear! Anything beyond an equality expression or straight boolean should be avoided.

It is entirely possible a Rust port could have caught previously unknown memory safety issues. Furthermore, a Rust port that looks and feels like C++ may be peppered with unsafe calls to the point where the ROI on the port is greatly reduced.

I am not trying to dunk on the effort; quite the contrary. I am eager to hear more about the goals it originally set out to achieve.

Before preaching it to others, the writing of a homily or sermon first needs to affect the heart of the one delivering it. Such heart-work is exceedingly difficult if not impossible with AI.

Any word on how much more memory safe the implementation is? If passing a previous test suite is the criteria for success, what has changed, really? Are there previous memory safety tests that went from failing to passing?

I am very interested to know if this time and energy spent actually improved memory safety.

Other engineers facing the same challenges want to know!

> We do [cubic curve fitting] all the time in image processing, and it works very well. It would probably work well for audio as well, although it's not used -- not in the same form, anyway -- in these applications.

Is there a reason the solution that "works very well" for images isn't/can't be applied to audio?