macNchz

I’ve taken early stage apps through a bunch of security review processes and never encountered questions about the specifics of the auth backend, beyond whether it can support the client’s specific SSO requirements.

These days I tend to favor having auth built-in, via an "old school" web framework that provides an extensible auth system out of the box. Then we’ll extend that system with a managed 3rd party service to handle SAML when that starts to come up in sales conversations, because the setup is annoying and we can lean on the vendor to deal with whatever weird old IdP the client shows up with.

I definitely think the bigger context window helps. The code quality quite visibly drops across all models I've used as the context fills up, well before the hard limit. The editor tooling also makes a difference—Claude Code pollutes its own context window with miscellaneous file accesses and tool calls as it tries to figure out what to do. Even if it's more manual effort to manage the files that are in-context with Aider, I find the results to be much more consistent when I'm able to micromanage the context.

Approaching the context window limit in Claude Code, having it start to make more and worse mistakes, then seeing it try to compact the context and keep going, is a major "if you find yourself in a hole, stop digging" situation.

I'm all about clear prompting, but even using the verbatim prompt from the OP "ffmpeg command to convert movie.mov into a reasonably sized mp4", the smallest current models from Google and OpenAI (gemini-2.5-flash-lite and gpt-4.1-nano) both produced me a working output with explanations for what each CLI arg does.

Hell, the Q4 quantized Mistral Small 3.1 model that runs on my 16GB desktop GPU did perfectly as well. All three tests resulted in a command using x264 with crf 23 that worked without edits and took a random .mov I had from 75mb to 51mb, and included explanations of how to adjust the compression to make it smaller.

Lack of complete e2ee is a feature for many large organizations—they still want everything encrypted, they just want a master key to be able to audit communications for compliance/investigations/insider threat identification. They also want strict control over who does what with the app, and where all of the associated data lives. Teams is just a totally different product from WhatsApp in that regard, with all sorts of functionality that will never exist in WhatsApp—tons of control over user identity and access management, integration with all sorts of other security tooling, etc.

Yeah I've mostly just embraced having to remove them as part of a code review, helps focus the review process a bit, really.

I've been intending to try some side by side tests with and without a conventions file instructing it not to leave stupid comments—I'm curious to see if somehow they're providing value to the model, e.g. in multi-turn edits.

Using all of the popular coding models pretty extensively over the past year, I've been having great success with Gemini 2.5 Pro as far as getting working code the first time, instruction following around architectural decisions, and staying on-task. I use Aider and write mostly Python, JS, and shell scripts. I've spent hundreds of dollars on the Claude API over time but have switched almost entirely to Gemini. The API itself is also much more reliable.

My only complaint about 2.5 Pro is around the inane comments it leaves in the code (// Deleted varName here).

Not a doctor, but to my understanding there are two main issues with CT scans in this context: the radiation from the scan itself, and the fact that it can lead to "incidental findings" which then might cause a cascade of other, more invasive tests with risks of their own.

CT scans are super valuable for all sorts of things, and the radiation dose/risk for an individual is small, but at a societal level using them freely as a "might as well" kind of test results in a lot of excess radiation exposure. This recent research suggests up to 5% of cancer diagnoses each year could be from CT scan radiation: https://www.ucsf.edu/news/2025/04/429791/popular-ct-scans-co...

The "incidental findings" side is an interesting dilemma: CT scans often reveal other stuff beyond what they were initially ordered for, but which can't be identified solely by the CT scan itself. So, if something looks a little weird on the CT scan it creates an imperative to figure out what it is, so there are additional tests, each of which carries its own risks and also consumes limited resources that might otherwise have been used for people with more definitively concerning findings. https://en.wikipedia.org/wiki/Incidental_imaging_finding

It's effectively malware—this article has some more detail: https://arstechnica.com/security/2025/06/meta-and-yandex-are...

Basically, they created a channel between the browser and a localhost webserver running in their native apps, by abusing the ability to set arbitrary metadata on WebRTC connections. That way, they were able to exfiltrate tracking cookies out of the browser's sandbox to the native app, where they could be associated with your logged-in user identity.

The NYT does have a direct-sold ads business and first-party data platform for targeting them: https://open.nytimes.com/to-serve-better-ads-we-built-our-ow...

I love when I ask it to remove things and it doesn't want to truly let go, so it leaves a comment instead:

   # Removed iterMod variable here because it is no longer needed.

It's like it spent too much time hanging out with an engineer who doesn't trust version control and prefers to just comment everything out.

Still enjoying Gemini 2.5 Pro more than Claude Sonnet these days, though, purely on vibes.

I find the way that it handles errors to be super annoying—I type fairly quickly and hitting backspace is built into my muscle memory in a way that just seems totally incompatible with that site, even with the "Stop cursor on error" setting unticked.

> Disagree. Some learn, not all and decreasing numbers career to learn

This is an organizational issue then—someone who is operating at a junior level who demonstrates that they don’t care to learn should be let go.

It was founded by two Radio Shack employees!

Where I've found LLM coding assistants really effective in my client consulting work is around iteration: I can deliver so many more versions of an application UI in a given amount of time that it really changes the entire process of how we dig into a project. Where I might previously have wanted to start with low fidelity wireframes and go through approvals to avoid proto-duction and pain down the line when the client didn't like something, now we can rough out the whole thing in a functional proof of concept and then make sweeping changes live on a call as we test different interaction paradigms.

The Walmart in the area from this article is separated from the main town by a four lane road with no sidewalks, across which the nearest crosswalks are more than half a mile away in either direction—so you’re either playing high stakes Frogger, or, depending on your starting location, you might conceivably have to walk nearly two hours out of your way round trip along the shoulder of this road to use a crosswalk. They also get five feet of snow per year, so a good part of the year that walk is extra dangerous and miserable.

I can’t say for sure, but I think this is much more typical of American Walmarts than it is to be able to easily walk to them.

Streetview of your opponent as a pedestrian trying to access the Massena Walmart: https://maps.app.goo.gl/ufTWTHxHCReFP8VA9

There is certainly a wide spectrum in how people behave, and an important factor is also how they would like to be perceived by others. Rundowns on how wealthy people live will tend to overindex on the behavior of people who want others to know they’re wealthy.

As an example, there’s a culture among what people refer to as "old money" families in the US northeast (with generational wealth from long ago), wherein they tend to avoid seeming outwardly wealthy or really talking about money at all…generally aiming to project an unpretentious vibe, eschewing designer clothes and driving 20 year old Volvos, but still spending vacations at long-owned family getaways worth tens of millions, flying first or charter, and send their kids to specific, expensive schools to socialize with others of similar backgrounds.

The great majority of lower-income people commuting into Manhattan are not doing so in cars, and those that do are eligible for a discount.

As an anecdote about any amount of parking winds up being used if it's free: there's a guy a block over from where I live in dense, transit-connected Brooklyn who has, at any given moment, at least 6 cars parked on the street, that he tinkers with and seems to occasionally buy and sell. The houses on that block sell for around $3-4 million these days, but the parking is free.

The neighborhood is also quite clearly full of cars that basically never get used—given there's no cost to store them, the owners save them for an occasional trip and in the meantime they accumulate snow/leaves/pollen etc the rest of the year, in proportion to how long they sit without being used.

The dysfunction of this system is particularly aggravating when you have an actual need to park nearby your home, e.g. if you're loading/unloading heavy things, and there's never any space. I would love if every block had temporary loading zones, but people freak out about the idea because it requires taking away some free parking.

Somehow appropriate for a crowdsourced experience of driving.

Windowless office spaces are so miserable to me. Often under-ventilated and overlit with unpleasant artificial light as well.

On this topic, something window-related that’s common in France but rare in the US is functional shutters—it’s so nice to be able to completely close a shutter outside your bedroom window and have pitch blackness, regardless of the light outside. The best you get in a typical American house is "blackout" curtains or shades that leak tons of light around the edges.

I've wanted to do a similar thing for "sweatiness" for a while, to try quantify the amount of time each year you can do outdoor activities or have the windows open without feeling too hot, based on dew point/humidity + temperature.

I grew up somewhere that didn't require air conditioning, but live in NYC now and find that as an adult the only weather I find truly unpleasant is when it's too hot to do anything outside without immediately being drenched in sweat. I really dislike the feeling of being trapped inside in the A/C, since at least even when it's very cold you can be comfortable in proper clothing.

Even without an actual map, though, I've been enjoying many of the visualizations on https://weatherspark.com/ for comparing cities on these kinds of things.

Having been on both sides of this—working behind a counter and answering phones at various jobs long ago, and being someone who often surprises family and friends with my ability to extract good outcomes from customer service—I think it’s somewhat of a misconception that being as unpleasant as possible is actually effective at getting results.

I fully understand that the godawful CS mazes many companies set up wind up pushing people in that direction, and that it feels like the only option, but I believe quite strongly that being patient and polite but persistent winds up being much more effective than being unpleasant.

As a small case in point: I worked summers in a tiny ice cream shop, most of the time solo. The shop had a small bathroom for employees only—it was through a food prep area where customers were not allowed by health code. I had some leeway to let people back there as it was pretty low-risk, and I would in the evenings when no other businesses were open, or if a little kid was having an emergency. People who were unpleasant from the get-go when placing their order, however, were simply told we had no bathroom at all. People who started shouting when I told them I wasn’t supposed to let people back there (not uncommon!) and suggested a nearby business were never granted exceptions.

Works great, I've been using it in production for a few years. DRF was one of my least favorite bits of the Django world and Ninja has been an excellent alternative.

I still love Django for greenfield projects because it eliminates many decision points that take time and consideration but don't really add value to a pre-launch product.

Well yeah, scrolling through and liking a picture of your friend's vacation and commenting "Adorable!" on a video of your cousin's toddler only gives you, say, 10 minutes to see ads, whereas getting fed an endless stream of progressively more intense and precisely-tuned content to tickle at your inner psyche (be it most susceptible to anger, lust, envy, greed etc) means you might spend hours on there scrolling past ads.

I think people in general are not all that great at doing this.

Anecdotal, but I grew up in a small town in rural New England, a few hours from NYC and popular with weekenders and second-home owners from there. I don’t think that people from NYC are inherently rude, but there’s a turbulence to life in NYC to where jockeying for position is somewhat of a necessity. It was, however, transparently obvious in my hometown that people from the city were unable to turn it off when they arrived. Ostensibly they had some interest in the slow-paced, pastoral village life, but they were readily identifiable as the only people being outwardly pushy and aggressive in daily interactions. I’ve lived in NYC for some time now, and I recognize the other side of this, and feel it stemmed less from inherent traits and more from an inability to context switch behavior.

> I live in NYC now, and fortunately we still have Micro Center which is still fun, though I do worry that it might go the way of Fry's Electronics soon enough.

There are actually five in the metro area: Brooklyn, Queens, NJ, LI, and Yonkers. I live in biking distance of the Brooklyn store and drive past (stop at) the Yonkers one pretty frequently and find they’re both usually quite busy, so hopefully they’ll stick around.

I make a point of trying to buy from them because I value having a place nearby that employs actual purchasing staff so I don’t have to sort through junk and counterfeit products, and because I like having a place where I can deal in person, have things same day, return without shipping etc.

Ordering pizzas cash-on-delivery to someone’s house is already a fairly common form of online harassment (sometimes also a precursor to swatting).

Well now we'll have four different ways to format strings, since removing old ones is something that doesn't actually happen:

    "foo %s" % "bar"
    "foo {}".format("bar")
    bar = "bar"; f"foo {bar}"
    bar = "bar"; t"foo {bar}" # has extra functionality!

Yeah I see a lot of stuff in America where decades of single-minded optimization towards cheap & convenient has had knock-on effects that ultimately don't serve people very well, or wind up creating second order costs—like hollowing out downtowns for cheap stuff at Walmart being a big part of making car ownership a prerequisite for daily living, or the cheap and convenient food options that have precipitated a nationwide public health crisis, or the way we've constructed our built environment in a way that has seemingly left people unhappy and isolated.