Quite a lot of sites work well with other browser engines too. I have tried PaleMoon, Netsurf. Users of Apple mobile devices seem to be fine with Safari.
Really light or unusual browsers might be a problem. Dillo, for example (its very fast though).
You can actually get other close-to-browser experiences with stuff like xwidget and embedded WebKit (https://emacsdocs.org/docs/emacs/Embedded-WebKit-Widgets) but generally it’s not worth the fight for commercial sites these days; everything is vibed and only tested against Chrome, so if you’re lucky they’ll load in other Chromium browsers and _maybe_ Firefox.
From what I usually see, ML researches just throw everything at the wall and see what sticks. Whatever approach turns out to work better, gets used more often, and sometimes theoretical background is developed for why it works. RoPE works not because it’s good math, just it turned out to work better than some other equally arbitrary approaches at some point, and then other approaches have also been developed that work better than RoPE.
1. Rotary positional embeddings are representations of the translation group. The n-dimensional translation group ℝⁿ is abelian, so its irreducible representations are all one-dimensional ρ: ℝⁿ → ℂ, specified by a momentum vector m: ρₘ(x) = exp(i⟨m, x⟩). You can get a multi-dimensional representation by choosing several m, which are your learnable parameters.
2. If you wanted RoPE embeddings to also preserve rotation (perhaps you have scenes viewed at arbitrary angles), you are looking for representations of the special Euclidean group. The irreducible representations in four dimensions are your familiar particles: electrons, protons, etc., and in general are parameterized by a momentum vector and a discrete spin vector. You would have to do some kind of annealing process to train the spin vector (or you could just make it constant).
3. If you want to do attention on quadtrees, well, what is the group? Suppose every node in the quadtree has local coordinates scaled to [0, 1]. Moving up/down/to a neighbor in the quadtree requires an affine transformation of the local coordinates, which compose, so we can label every node by its affine transformation to the root (x, d): y → x + 2⁻ᵈy, where x is the node's translation in the image and d is its depth in the tree. The finite-dimensional irreducible representations are parameterized by momentum vectors m as well as a modulus M and basis vectors {e₁, e₂, ..., eₖ} where k = ord₂(M) and look like ρₘ,ₑ(x, d) = DPᵈ where D is the diagonal matrix with entries Dⱼⱼ = exp(2πi⟨2ʲm, x⟩ / M) and P permutes the basis vectors.
I hadn't heard of it, and thought to share it here. I'm not sure why people are upvoting it - but it's not impossible that others are interested in this topic as well..! Cheers
This is normal and expected activity. We encouraged our users to write canonical answers, that could be run as community wikis (not everything got converted to a wiki, but many things did). One of the reasons was to take a spot in the top 3 on a search engine, in an attempt to displace some horrendous / ad riddled website that at best a bad user experience, and frequently was subtly wrong or incomplete.
Some examples: we settled on a canonical answer for “how to install packages in Ubuntu”, or when there was a particularly interesting question that was discussed in the chat rooms, someone might volunteer to consolidate that information and publish the Q&A as a wiki.
It’s just internet points. They’re not worth much.
Like this one, most of the questions are worded to enable an easy answer from a primary source.
In one instance the account answers its own question within the same minute: https://space.stackexchange.com/questions/31279/how-was-the-...
In another instance, the account replies to a comment with "Nonetheless, this would be the question to document any examples."
Now why on Earth would you comply with that request?!
"Kosmonaut survives space mission, gets mauled by bear protecting her cub" is not a great headline. Bringing gun was relatively cheap insurance
But I liked this comment be caused of its know your meme link. I learned something.
Well, I learned two things. One, I learned about the meme. Two, I learned not to trust Know Your Meme when it asks me to turn off my adblocker!
Page 2-170 (179 of the PDF) of reference [1].
[0] https://ntrs.nasa.gov/search.jsp?R=19710014816 [1] https://ntrs.nasa.gov/api/citations/19710014816/downloads/19...
nasa employee: oh hey u guys are back early
astronaut: moon's haunted
nasa employee: what?
astronaut: *loading a pistol and getting back on the rocket-ship* moon's haunted.
https://knowyourmeme.com/memes/moons-haunted> However, we would appreciate your sense of humor in chat if you want to drop by!However, we would appreciate your sense of humor in chat if you want to drop by!
On HN many purely jokey comments are often discouraged, being down-voted and getting replies along the lines of "this isn't Reddit". Is HN dying, by the same measure?
Common in ML-family languages and Rust. They've been brought into most mainstream languages like Java by now because of how useful they are.
Is this not widely considered to be the ugliest part of the language? Why do you prefer having to do this manually?
- manual memory management
- arena and pool allocators
- go style err!=nill style exception handling
- first class support for soa and vectorized simd, linear algebra and swizzling
- sanely typed and namespaces
- explicit unambigious overloading, subroutine context
- first class support for defer statements
- zero friction C ffi
if you haven't heard of odin, you should definitely consider it.
However the language extensions that were the reason for the lawsuit did indeed transition from J++ into C#, with J# being J++ on top of .NET for easing code porting.
The features were events, J/Direct (later P/Invoke on .NET), COM interop (later RCW/CCW on .NET), and Windows Foundation Classes (later Windows Forms on .NET).
Thanks to the good work of internet archivists,
https://archive.org/details/microsoftvisualj00micr
https://download.microsoft.com/download/1/6/d/16d24ada-5317-...
https://www.microsoft.com/en-us/download/details.aspx?id=559...
The irony is that two decades later Microsoft has an OpenJDK distribution, and they were the ones that initially provided the ARM port for Windows.
https://www.microsoft.com/openjdk
https://www.zdnet.com/article/microsoft-completes-phase-one-...
Software in the 80s showed what's possible when you really care about doing it. Software in the last 15 years feels like an exercise on what happens when you don't.
Fundamentally, many inefficiencies were justified for developer productivity, and I'm not sure that makes sense anymore, and the cost of larger projects with clear abstraction boundaries has plummetted.
I don’t identify as a New Atheist, though I enjoy much of their writing. Dawkins is a good author. Sagan and Hitchens were excellent. If someone wanted to explore atheism, those are the writers I’d recommend.
Atheism isn’t a club you join. It doesn’t automatically give you a community or a shared outlook. I’m just as likely to like or dislike a fellow atheist as a theist. Atheism isn’t enough to make us part of the same ‘tribe’.
Which is why atheism will never win the imaginary war some theists think we’re waging for people’s hearts and minds. Outside of a few New Atheists, we just don't care that much about what other people believe.
Ok, then.
Or just use your phone’s slo-mo feature to take a 240 FPS video (which then gets played back at 30 FPS). 60 Hz flicker becomes extremely visible. More than once I’ve noticed the strobe effect and then confirmed it this way.
The thing that makes that so insidious is that not everybody can easily perceive that.
(The reason why such flicker is common in cheap LED bulbs: they turn AC into DC using a half-wave rectifier, which just chops off half the AC wave. This is also why you can still see the effect in rooms lit by multiple lights, because the flicker is in sync, synchronized to the incoming AC.)
If you are building a space for other people and you are responsible for the lighting, please check for this. If you can't perceive it yourself, find someone who can.
Here are some other things that people are sensitive to: - CRI on white LEDs (and flourencents). CRI is how clean the spectrum is in comparison to the sun or a incandescent bulb, where the specturm is completely even. There are some CRI >=95 LEDs that are fine for most people however.
- Motion blur Motion blur on LCDs (such as ~50% of Nintendo Switch 2s) can hurt eyes trying to keep track of what is going on.
It’s also odd to think about lack of ability to concentrate, and I wonder if screens cause this optically somehow but to a less noticeable degree even when less intense than this new one? (Independent distracting contents)
I noticed that adding a severe red filter to all my screens helped me significantly with the concussion and am also curious about that, would that impact the intensity of a refresh perhaps? Is blue light specifically the problem? I’m unsure but have a plethora of subjective experience.
I’m glad for this to enter public consciousness a bit and would love to see more research on the screen itself impacting our brain.
Maybe they just let their rigol scope spit out 240hz, but 480hz worst case and I would actually say 960hz. I think what’s happening with the tandem oled is they both pwm at 480hz and 180 degrees out of phase giving effective 960hz when both are in use. Which appears to be most of the usable brightness.
Edit: I’m so confused.
“The PWM behavior is interesting, because we can see DC-dimming for all brightness levels. Usually, OLED panels use DC-Dimming only at higher brightness levels. The frequency is very low at 240 Hz, however, and this might be an issue for sensitive users, despite the DC-dimming. ” https://www.notebookcheck.net/Amazing-OLED-panel-Apple-MacBo...
1. CRT display phosphors had afterglow.
2. the cathode ray was a dot moving across the screen, that might be easier on eyes than whole OLED screen flashing at once.
The thing these have in common is that they're connecting all the people. There is no intermediary necessary. Probably you won't send a postcard to some child you've never met in Shanghai, make a phone call to a random farmer in Ukraine or use a video chat to call random Venezuelans on the Internet, but you can and that's why these networks are so powerful.
- There is almost always a standard operating procedure (SOP)
- Just b/c there is a SOP for releases etc, that doesn't mean that people always follow it
- This can be due to many reasons e.g. laziness, being distracted, pressure from higher ups ("just push it out there and I'll take the risk" ha!), tired from having a newborn etc etc
- Doing things outside the "green zone" deployment windows therefore always carries risk
- All of this being said, EVEN IF you ARE following SOP, things still break in all kinds of weird ways
- And the end user is not often the best observer nor do they have all of the context to determine what broke, how it broke and why. (There are exceptions to this though)
This was a cascading multicast failure, which can take hours to build up. Earliest reports of disruptions came no later than 10:30 am, so it's quite possible the upgrades were performed in the overnight hours on Thursday.
- Cutting DRAM production in 2022 as a result of low prices/demand made sense if they didn't expect the market to recover as quickly as it did. A lot of the recovery is also likely caused by the start of the AI boom.
- The deals OpenAI made with SK Hynix and Samsung was a smart move by OpenAI where they drove up prices for everybody else while securing "cheap" supply themselves. I dont think its clear whether SK Hynix and Samsung came out ahead in that deal as the demand likely would have risen either way and it wouldn't have needed to go up by as much as it has now.
- Moving production to HBM also makes perfect sense when that is what the market demands and is willing to pay a premium for.
how many successful careers include 8 months in prison
Finally secure a research grant and had budgeted money for a new compute workstation and the RAM prices are killing my budget, because I'd love to get at least 256 GB of RAM and that's 4-5 grand now. Ridiculous, especially since I'd like to have a GPU in it with at least 32Gb of VRAM for medical image model training, so that is also like 4-5 grand too.
Do I wait two years and hope, or pay, isthe question I keep asking myself.
EDIT: missed the positive part remaining, sorry. I maintain that the ambiguity issue stands.
HTTP 418 I'm A Teapot
I'm-Not-A: freezer;teapot;radioIn this case we're talking about vibe coders, who don't know about the implementation details. But we've gotten here in the first place because a lot of developers adopted this Supabase RLS serverless model. I always found it surprising that this model stuck.
To add some nuance: Supabase’s security controls aren’t the simplest to work with and have been a source of confusion -- and data exposure -- for years. But Lovable markets its product specifically to people who, it claims, don’t need that technical knowledge. That sets the expectation that the user doesn't have to worry and that Lovable will get Supabase access controls right.
7500 site operators failing to "action provided findings" seems like a systematic failure on your part and you should accept more responsibility for it. I'm guessing users find the security warnings unclear, confusing, or meaningless. Don't allow users to publish apps if you know they leak private data. Work on educating users and make the apps be built securely the first time.
When an app has end users, security is a key part of the product. That’s why we keep investing in making security tools better and easier to use, taking on as much of the security load as we can and fixing more issues automatically.
I personally believe that making security accessible is what leads to making software accessible. We’re committed to both.
You can read more about our approach: https://lovable.dev/blog/find-it-fix-it-ship-it
In practice it’s an absolutely stupid design. If you create tables without using their dashboard, like with an ORM or with SQL directly, it’s incredibly easy to have them not use RLS and be wide open to the public.
Taking a step back, the whole “query the database straight from the frontend” is a terrible design in general. For example, they still don’t offer rate limiting! Even if the tables data is secured, there’s nothing stopping someone from hammering the database directly with reads/writes until it overloads it, as there’s no application level rate limiting in the middle. Bonkers.
It looks like you're weirdly trying to inflate a legitimate attempt to confirm if the severity of the exposure is as bad as suspected into a full blown nasty data thief.
This is why everyone is so puzzled. You're genuinely harassing op for doing a good thing and responsibly reporting the severe security problem.
I’d much rather live in a world where independent researchers can figure out the extent of the issue rather than go “uh oh there might be an issue here, guess we’ll never know the extent of it, let’s trust them to fix it”.
I can't believe someone actually made this by design
All LLM-based analysis was performed using local models, specifically the "elephant-1" model on our local RTX Pro 4000 SFF.
I respect your insistence on proper data handling -- you’re right in doing so -- and we tried to be as careful as possible when conducting the analysis.
Now let’s hold Lovable to the same standard.
Supabase tables also use pagination, and we never downloaded more than the first page.
https://docs.postgrest.org/en/latest/references/errors.html
For example, if you try to access a table named "secret" that doesn’t exist, Supabase might return a hint saying, "Perhaps you meant the table 'public.super_secret'."
Thanks for volunteering that information, I guess.
This was supposedly fixed on October 30, but there are reports floating around that for existing projects, the defaults still apply.
Lovable deserves some heat for using this, but this was a choice that was made by Supabase many years ago to support making it easier to onboard potential customers, a deliberate tradeoff of less onboarding pain for less security.
Using datasets from the open web archive Common Crawl,
we identified and analyzed 159,041 Lovable projects from
December 2025 to September 2026. Of these, 34,545 used
Supabase.
24,664 websites sent database records from Supabase as
soon as they loaded, without requiring a login or any
other interaction. That alone is not a security issue:
the data may be intended for public display. But by
making additional read-only requests, we were able to
retrieve more database content. We found records
containing email addresses in 10,560 tables across 7,403
of the 24,664 websites (30%).
What isn’t mentioned in the analysis is that we ran a System 1 decision model on each dataset to determine whether it contained privacy-sensitive data. Using a confidence threshold of 85%, we found that nearly 4,000 websites leaked such information.Edit: That's 15,9% of all Lovable websites that use Supabase.
Allowing anyone to build and host a public app without a basic understanding of these things is like letting people fly a passenger plane without a license.
The initial idea behind the analysis was to start with the most basic setup and then move on to more sophisticated attacks. We never reached that second part, as so much sensitive data was already out in the open.
Edit: The answer (as neathack tells it) is: 2.5% of the sites. The reply was buried below.
- They are an easy target, they don’t need any sophisticated hacking knowledge (article mentions inspecting network calls in browsers).
- There tons of those, and more created everyday. For the average user, they don’t know what they are using, or how risky it can be.
The old joke was "If builders built buildings the way programmers wrote programs, then the first woodpecker that came along would destroy civilization", but in the AI era it feels like a butterfly would be enough.
One thing I learned this year is that naive text mapping (embed a corpus of reviews with a Qwen embedding model, use umap to plot, then compare where things fall) primarily compares embedded text length more than anything else. What a review says affects the position less than how long the review is.
This was a little surprising to me. For whatever reason, I didn’t think the embedding would so strongly describe the length of the embedded passage. So most of this year’s time was spent fixing that artifact.
[0]: 2025’s map: https://storage.danieltperry.me/share/steam-map/
[1]: 2026’s map (still a work in progress, both the ui and the review processing are incomplete): https://storage.danieltperry.me/share/steam-map-refresh/
The article sounds to me like Jensen was attempting to maximize revenues and profits for the company he ran, like a CEO is supposed to.
Is Nvidia profiting from globally constrained wafer capacity ? Absolutely.
Does their strategy account for supply constraints ? Absolutely.
Isn't this what corporations are supposed to do ? Adapt, pivot, profit ?
I think everyones just jealous and frustrated right now, and the frustrated decided they didnt want to see leather jacket man getting laid while sipping Dom.
People are petty like that, especially in large crowds.
We have the phrase “container engine” which could mean dockerd or Podman; “container runtime” to me usually refers to containerd or CRI-O but could also refer to runc and crun; and “container manager”, which could alternately refer to dockerd and Podman or to “orchestrators” like k8s Swarm and Nomad. “Desktop app” is what I would call Docker Desktop, Podman Desktop etc.
From cursory reading this appears to be a new desktop app for containers. However they say that the container logic runs behind the real dockerd, so I suppose they’ve slotted it in where containerd goes? So it would be a new container runtime. I feel that’s a little odd to conflate the two, but that’s getting into the hole we’ve already dug with these terms.
Among software developers, we shouldn't need to provide a "assuming there are no bugs" caveat with every statement about how something works, because we know that bugs are essentially unavoidable at scale. We can say that Rust makes it much easier to avoid bugs, but nobody serious has ever claimed that they are 100% eliminated.
It's not like C or C++ compilers are bug free either, despite implementing much weaker models.
This entire argument against Rust hinges on it somehow invalidating all of the improvements it does have, all the safety it does guarantee, just because it also has a few bugs. It's neither a myth that Rust is safer, nor is it a nail in the coffin for it.
Otherwise you must make the argument that just because unsafe exists, and is used, it is automatically as bad as let's say C - a language that has proven time and time again how trivial it is to make mistakes with. Something that just doesn't happen in Rust.
The Rust project as a whole currently has 178 open issues with the tag I-unsound[1]. Many of these describe issues that only exist in the nightly compiler, of course, but it does happen that bugs are not caught before shipping. That's software.
[1]: https://github.com/rust-lang/rust/issues?q=is%3Aissue+state%...
The `variable[i]` notation is intended to function like in other languages, where it returns an unwrapped value. It would be ugly, noisy, unergonomic, and annoying to have to properly handle `Some` (like in `variable[i].unwrap()`) every single time you ever used it. Imagine doing `c[i].unwrap()[j].unwrap() += a[i].unwrap()[k].unwrap() * b[k].unwrap()[j].unwrap();`.
If you really want the result to be wrapped in `Some`, you can use `variable.get(i)`.
That being said, another 2ct:
> A software Engineer needs to have all phases of the SDLC in mind, and be thinking about and planning all of them. Code is a tiny piece of the puzzle.
I would not AT ALL call it tiny. Yes, is one part of MANY. But is your final delivery. Would be like Toyota saying: “hey they don’t really have to understand what makes a good car, the cars we sell is a tiny portion in the whole development cycle”. SW engineers do need to tell apart good from bad code, as much as good from bad tests and good from bad requirements. After all, if I somehow manage to deliver good SW (working and documented) I don’t gove a crap how the requirements and teat work or don’t in the company.
The way I've always phrased it is that if a company has "DevOps" teams or titles, they're not doing DevOps by definition
The point being: I'm always amazed at the cross-purpose impulse of arguing "the title is not important" while, simultaneously, really wanting to use the title.
One of the hardest things is that the same title (and level) can mean very different things across companies, industries and over time. When collecting pay information, we take great care to merge titles that mean the same thing while distinguishing those that don't. There's other edge cases like at some companies a title will be more specific (Security Engineer), while at others they'll retain the general title (Software Engineer) but specialize in the same thing (Security).
Claude has been really helpful in wrangling this all and expanding our taxonomy :)
- speak to production users
- ship to production
- monitor and learn from production usage
You don't need "devopsy" skills to do that, rather organizational structures that allow for incremental change. FWIW, deployments could still be a 2005-like upload via FTP, and the benefits of DevOps could still be reaped.
DevOps is another sad story as it was a saner approach to get people skilled for both dev and ops but turns out people don't like learning new skills as much as they claim. So its meaning differed from org to org and even within the same company.
- DevOps Engineer (Junior, Senior, Staff)
- Systems Development Engineer
- DevOps Team Manager
- Cloud Infrastructure Engineer
- Site Reliability Engineer (+Lead)
All have been meaningless except for Cloud Infrastructure Engineer. The rest described a generalist who knows how to deal with ambiguity and (usually) has the trust with things like AWS root account credentials. Most organisations have no idea what real DevOps is. If they did, they wouldn’t hire for a “DevOps Engineer”
With AI assisted vulnerability research the relative edge of project zero is going down - anything they can find can likely be found by others in a similar timeframe. We see this today with the parallel discovery of php2shell by many groups in the same few weeks
That is the world we live in now and pretending otherwise is reckless and leads to mass hacking like the recent supermarket fridge hacking.
Only devices and applications that can respond to security reports quickly should be connected to the internet.
It's gotten easier finding vulnerabilities, so every day without disclosure now brings a higher chance that blackhats have already found the bug. The deadline gets shorter accordingly.
Calling vulnerability research "extortion" is the kind of ignorant messenger-blaming the hacker culture has always ridiculed.
The article says they are willing to negotiate in exceptional cases, and mentions Spectre.
Definitely much happier. There’s nothing I miss from Ubuntu.
On February 2, 2027, we will move to a “Full 60” policy where vulnerabilities are disclosed after 60 days.
On January 3, 2028, we will move to a “Full 30” policy where vulnerabilities are disclosed after 30 days.
I've seen simple copper->fiber migrations like this take down a large network and I've seen them be blamed when they turned out to be unrelated. I could come up with a scenario for most any view anyone wanted to speculate on because the details are so light and the possibilities are so many, but I won't because I don't have enough details to say which it might have been without just guessing myself. If they say they identified a broadcasting knob they disabled on the other ~60 routers already in service then that's what I'd leave such a conversation at for now - but I wouldn't necessarily trust that or what it really means without more info.
In the thread the other day I saw people blaming anything from AI to attacks to Starlink but few willing to admit they just needed more information to have any certainty about what to think beyond "it sucks this is down and I wonder what the reason will be". The only thing I'd raise my eyebrows at at this point would be if we don't get further details in reasonable time.
But I recall years ago the ATC system made heavy use of multicast in ways that scared me. I dunno if that stuff is still out there but it was some real I wouldn't do that shit if I were you stuff. And yeah mcast can sure fuck a network fast.
As for in general there's plenty of ways a single box can melt a network. Routing protocols let you do bad things fast and pull in more traffic than you can eat.
I was the Sr. Internet Mail Administrator at AOL at the time, and this article was what The Register wrote based on what I had submitted to them. They made a few changes. I know they were trying to protect my identity, but I think they also wanted to make it a better story.
For us, everything started with a planned downtime to upgrade the main core router for the entire service. We had the best network engineers in the business, but even they can make a mistake. And they did. And then there was a whole domino effect of events that followed.
As described it was in the course of switching out copper TDM lines with fiber. On it's face that could be straightforward, but it's easy to imagine the router(s) terminating these lines didn't support the required line cards, so router upgrade(s) were required. So theoretically straightforward could have become more involved.
However, when doing far less critical updates, it's common to set up a test environment, do a dry run of the work, test the roll-back, test the monitoring and alerting, that sort of thing. Doesn't guarantee no outages, and I'm playing armchair QB here, but it seems like a botched job.
It's also telling the Transportation Secretary said things like "we aren't sure they if they aren't doing a great work for us". He also said "this is something businesses have been doing for decades and the FAA is the last to do it"; you'd assume that this is a process that someone (maybe not those doing the work) have developed deep experience with.
I'm not familiar with this L3HARRIS, I gather they are a defense contractor with billion dollar revenue.
Lots of things in this definitely raised my eyebrows.
They should face criminal penalties for violating people’s rights
> the data is actually used during investigations.
Absolutely no! All that data collected without due process (judicial warrant, paper trail, whatever is the norm in your neck of the woods) has to be treated as tainted.
However, if they do anyway, there should be strong victims' rights protections so that the data is actually used during investigations.
There was a string of burglaries in our area a few years ago, and panopticon N-1 (Google Map's auto recording of all phone GPS traces for marketing information) could have found the perpetrators pretty much instantly.
They hit rural outbuildings in the middle of the night on camera, so timestamps were precise, but faces were obscured. A reverse geofence warrant keyed on time of arrival / departure would have narrowed it down to just the crooks with extremely high probability, and other details in the videos would have corroborated the GPS traces.
None of the victims were politically connected, so the police couldn't (by policy) run the reverse GPS lookups. This is the worst of both worlds: Our data is recorded to be used against us later, but if ever it would be useful, we do not have access to it. The same is true for vehicle black boxes / cameras. They record a few seconds before airbag deploy, but you usually are not allowed to access them as part of an insurance dispute (even the ones from your own car).
* Illegal to collect every plate for later analysis
* Illegal to collect anything except for license plates (no drivers, pedestrians, etc.)
* Must delete "non hit" plate images within 3 minutes
* Strict forbidden to upload "non hit" camera imagery off the device
I would add that it should also require a warrant to access the ALPR data, but it's a great start.
> ȝe han redily rehersed, bi resoun ful trwe, Alle þe covenauntz þat I yow kende...
("You have readily recapitulated, by fully accurate accounting, all covenants that I declared.")
I can't think of a more fitting name except, perhaps, the speaker of these lines: the exacting Lord Bertilak.
Congratulations, team. Here's to 30 more years.
Linux life, yo.
Qt is now its own company, namely Qt Group. Wikipedia quotes:
> In 2018, the company had 5,000 customers. Approximately 80% of the Fortune 500 companies are customers of Qt Group.[22]
So it's self-sustaining now.
AFAIK, KDE is still mostly volunteer work, but the way they develop software is pretty robust.
I really enjoy that it showcases all the work that goes into making a desktop actually good, even the small things you bearly notice, things like "When a grouped Task Manager task is flashing to get your attention, opening its group popup now reveals which specific window wanted attention. (Christoph Wolk, plasma-desktop MR #4225)"
A change doesn't need to be big and flashy and make headlines, even small changes deserve to be showcased and credited
Thank you KDE folks - I contribute in terms of bug reports and attempts at fixes sometimes but this time I am signing up for a monthly donation - I encourage people to do what they can to keep a good thing going! https://kde.org/donate/
And that $200 being out of bounds went pretty low: for the first couple years of my programming life, floppy disks were beyond my means, so any code I wrote lived in its official version as hand-written code and would have to be retyped any time I wanted it.
Ancient assembly hacker tip: A JSR/RTS sequence can be simplified to a JMP. To continue your example, the bytes you provided decompile as follows (assuming Apple 2 monitor ROM at $F800):
PRINTBANG:
A9 21 LDA #$21
20 F0 FD JSR COUT1
60 RTS
So, you load the A register with #$21, which is the ASCII code for `!` in hexadecimal, and then call COUT1 which will output the character in the A register, printing the exclamation point. When COUT1 is finished, it will return to you using an RTS instruction, and then you'll return to your caller using another RTS instruction.Since the call to COUT1 is a tail call, you can borrow its RTS instruction to return to your caller and save one instruction. So the following code is equivalent to yours but a little bit smaller and faster:
PRINTBANG:
A9 21 LDA #$21
4C F0 FD JMP COUT1
Notice that the JSR instruction has become a JMP.Another related trick: When you have a routine that callers often want to call twice—say to print `!!` for extreme exclamations—you may be tempted to create a do-it-twice call:
PRINTBANG2:
20 00 03 JSR PRINTBANG
20 00 03 JSR PRINTBANG
60 RTS
But now we know better and code it this way: PRINTBANG2:
20 00 03 JSR PRINTBANG
4C 00 03 JMP PRINTBANG
But we can do even better! We can eliminate the jump and save three more bytes by placing our do-it-twice routine immediately before the routine it is calling: PRINTBANG2:
20 03 03 JSR PRINTBANG
PRINTBANG:
A9 21 LDA #$21
4C F0 FD JMP COUT1
When PRINTBANG2 is called, it calls PRINTBANG to print an exclamation point, but when PRINTBANG returns, it returns to PRINTBANG2, which promplty falls through to PRINTBANG, calling it again and printing a second exclamation point. This time, however, when PRINTBANG returns, it will return to the original caller.Back when every byte mattered, these tricks were useful to know.
You had to carefully calculate the address distance for a branch instruction. The release of DISMON was a revolution: suddenly the computer did all of that for you. It even supported moving pieces of code to a different address to make space for inserting more instructions somewhere!
I figured out exactly what I wanted to do, wrote it out very carefully on paper, looked up all the hex codes, wrote those out very carefully, adjusted addresses etc. as needed. This took many hours spread over weeks. Then I entered the whole thing on the hex keyboard and display of a primitive development board each time I needed it, perhaps an hour or so each time (I think I only did it twice). It worked first time, because why wouldn't it if you were sufficiently meticulous about your work in a process that was very obviously very intolerant of the tiniest mistake?
I must admit my program (gather input information from some transducers, do some calculations, present some results) was likely somewhere between one and two orders of magnitude simpler than a complete Basic interpreter!
Maybe that is part of the nostalgia for the 8-bit days: the practice of "slow programming"? It was relaxing, sort of like those "slow tv" shows.
Why would any of those non-power-users buy a $2000 GPU that doubles as a space heater (seriously, try sitting next to one during long runs in the summer) and thousands more for a computer to host it, plus all of the trouble of setting it up, instead of a $20/month subscription?
I do have the hardware to run Qwen3.8-Flash-Next at home, and I still heavily use my subscriptions even for simple tasks. The experience of using the apps on my phone on the go is great and I always have the option to scale the model up to the task.
The people who are going to pay thousands of dollars up front to avoid $20 are power users and people who have different priorities and motivations, like independence or ultimate privacy. GPU prices dropping won’t change that much.
Also, you don’t need an RTX6000 for Q3.8FN. There are much cheaper options right now at lower speeds.
Engineering types and power users need not apply,but For Everyone Else, the "normies", Qwen Flash Next is probably "640k is enough for anyone" territory.
IPO, valuation etc depend heavily on frontier, or "good enough" AI to need to run on commercial grade GPUs. If Qwen Flash Next is good enough and runs on a $2000 GPU, why would you pay $20/30/100/200 per month in perpetuity?
I wouldn't pick Qwen Flash Next to add kernel namespacing and cgroups v2 to my custom OS kernel, but it will happily implement the Plan.md that Fable 5.2 wrote for a dollar.
You can read the WSJ article
https://archive.is/20261113091227/https://www.wsj.com/tech/j...
And my comment in the other thread
It may not be illegal but if you are investing in AI companies you need to think about whether the demand for AI chips is real or artificial and being manipulated by Nvidia.
https://news.ycombinator.com/item?id=50493251#50496483
> Then he went on to describe the situation the company was in. Major cloud providers were running away from Nvidia, making plans to use alternative chips and lessen their reliance on the firm. Even worse, Huang said, the vast supply of chip wafers and memory Nvidia had committed to buying were coming at him like an “avalanche.”
Jensen's son-in-law was making these vendor financed deals and Kress the CFO now interim CEO was against them.
Large customers like Nvidia and Apple want billions of chips. TSMC can't just ramp up and down instantly. It takes 3+ years and $30 billion to build a modern fab. So TSMC has contracts where you are commiting to buy the wafers for years in advance.
I've worked for multiple fabless chip companies that use TSMC to manufacture. Amazon, Google, Microsoft, and the other big hyperscalers are making their own AI chips to move away from Nvidia. The startups like OpenAI are doing the same so the real demand for Nvidia chips is dropping. But Nvidia can't easily get out of the contracts with TSMC to manufacture and buy the chips. They would probably have multi billion penalties that would show up on their financial statements.
Some people think Nvidia is acting as a VC fund for AI startups but my view is that Jensen freaked out about the dropping demand for Nvidia chips. So he is funding these startups with circular finance deals to buy the Nvidia chips that he can't sell to the hyperscalers anymore.
But weren't they were investing money in companies to help them grow and in turn buy more of their product. Of course Nvidia wants to stimulate demand of their product. I don't think this was nefarious. There is no free lunch, the risk of them never getting that money back is real with a lot of these pie in the sky AI companies. How is this anticompetitive? It's all of these SPVs that meta, google, and others are using that I think needs to be investigated and punished by wall street.
If you think that's wrong, USE YOUR WORDS, I'm open to reason. Down voting with no response or making jeers is not effective communication.
Nvidia allegedly engineered scarcity by funding AI startups to buy its GPUs (techpowerup.com)
Greed and hubris. I think the turbulences will be painful to some, especially those who counted on Nvidia stock to finance their retirement.
But, big picture, this is good news.
L3Harris has been modernizing the communications infrastructure supporting the National Airspace System in close coordination with the FAA. With safety and reliability as our highest priorities, L3Harris, together with its telco partners and the FAA, has completed more than 70% of the planned airspace communications conversion from legacy copper wires to a fiber optic system. We are supporting the FAA in its investigation to understand the conditions that led to the Nov. 13 ground stop of our nation’s air travel. L3Harris takes the nationwide ground stop and its impact on travelers, airlines and air traffic operations seriously.
<end entire statement>I had thought the L3 was from Level 3 Communications and Harris, but its actually L3 Technologies which seems to be a holding company for defense contractors.
Great first impression from a technical company
Thank goodness
> Agent messages
> Your AI agents can message you inside the game. A message costs the agent only the words in it, and nothing prints on success. Messages wait while the game is closed.
It's kinda amazing how the ancient "local mail" technology keeps getting reinvented, you know, and reintegrated inside the gameplay? NetHack had its its "mail demon" since... (googles) version 1.0.2, in April 1985 — the creature would spawn in the game, deliver you a scroll of mail, and vanish. And apparently, it's still in use nowadays? "On public servers such as nethack.alt.org, a user may observe other users' games and communicate with them by sending mail."
[1] https://www.reindeereffect.com/0002