I'm sad. I'm a software guy, not much of a hardware expert, so please bear with me if I"m too pessimistic. However, I feel that if trends like this continue, it might be the end of enthusiast-level personal computing as we know it. No more being able to head down to the electronics store and purchase RAM, motherboards, processors, GPUs, storage, and other components. We're going to be limited to locked-down terminals connected to cloud services, both of which provided by a small number of multinational corporations. If we're lucky, we might still have USB peripherals.
The sad thing is that we enthusiasts are a small market compared to the overwhelming majority of computer users who don't mind locked-down devices, or at least until they've been bitten by the restrictions, but if there are no alternatives other than retrocomputing, then it's too late. For decades we enthusiasts have been able to benefit from other markets with overlapping needs such as gaming, workstations, and corporate servers. However, many on-premise servers have been replaced by cloud services, the workstation market has been subsumed by the broader PC market, and PC gaming has faced challenges, from a push toward locked-down consoles to challenges in the GPU market due to competition with cryptocurrency mining and now AI.
One of the things I'm increasingly disappointed with is the dominance of large corporations in computing. It seems harder for small players to survive in this ecosystem. Software has to deal with network effects and large companies owning major platforms, and building your own hardware requires tons of capital.
I wonder if it's possible even for a company to make even 1980s-era electronics without massive capital expenditures? How feasible is it for a small company to manufacture the equivalent of a Motorola 68000 or Intel 386?
I'd like to see a market for hobbyist computing by hobbyist computer shops, but I'm not sure it's economically feasible.
Yeah, "Engineers don't try" is a frustrating statement. We've all tried generative AI, and there's not that much to it — you put text in, you get text back out. Some models are better at some tasks, some tools are better at finding the right text and connecting it to the right actions, some tools provide a better wrapper around the text-generation process. Certain jobs are very easy for AI to do, others are impossible (but the AI lies about them).
A lot of us tried it and just said, "huh, that's interesting" and then went back to work. We hear AI advocates say that their workflow is amazing, but we watch videos of their workflow, and it doesn't look that great. We hear AI advocates say "the next release is about to change everything!", but this knowledge isn't actionable or even accurate.
There's just not much value in chasing the endless AI news cycle, constantly believing that I'll fall behind if I don't read the latest details of Gemini 3.1 and ChatGPT 6.Y (Game Of The Year Edition). The engineers I know who use AI don't seem to have any particular insights about it aside from an encyclopedic knowledge of product details, all of which are changing on a monthly basis anyway.
New products that use gen AI are — by default — uninteresting to me because I know that under the hood, they're just sending text and getting text back, and the thing they're sending to is the same thing that everyone is sending to. Sure, the wrapper is nice, but I'm not paying an overhead fee for that.
Wow. They're not selling off the business, they're totally exiting it.
This is a big loss. Crucial offered a supply chain direct from Micron. Most other consumer DRAM sources pass through middlemen, where fake parts and re-labeled rejects can be inserted.
> Engineers don't try because they think they can't.
This article assumes that AI is the centre of the universe, failing to understand that that assumption is exactly what's causing the attitude they're pointing to.
There's a dichotomy in the software world between real products (which have customers and use cases and make money by giving people things they need) and hype products (which exist to get investors excited, so they'll fork over more money). This isn't a strict dichotomy; often companies with real products will mix in tidbits of hype, such as Microsoft's "pivot to AI" which is discussed in the article. But moving toward one pole moves you away from the other.
I think many engineers want to stay as far from hype-driven tech as they can. LLMs are a more substantive technology than blockchain ever was, but like blockchain, their potential has been greatly overstated. I'd rather spend my time delivering value to customers than performing "big potential" to investors.
So, no. I don't think "engineers don't try because they think they can't." I think engineers KNOW they CAN and resent being asked to look pretty and do nothing of value.
Ex-Google here; there are many people both current and past-Google that feel the same way as the composite coworker in the linked post.
I haven't escaped this mindset myself. I'm convinced there are a small number of places where LLMs make truly effective tools (see: generation of "must be plausible, need not be accurate" data, e.g. concept art or crowd animations in movies), a large number of places where LLMs make apparently-effective tools that have negative long-term consequences (see: anything involving learning a new skill, anything where correctness is critical), and a large number of places where LLMs are simply ineffective from the get-go but will increasingly be rammed down consumers' throats.
Accordingly I tend to be overly skeptical of AI proponents and anything touching AI. It would be nice if I was more rational, but I'm not; I want everyone working on AI and making money from AI to crash and burn hard. (See also: cryptocurrency)
Valve is one of the few companies regularly seen on HN where the headline is something like "[company] is secretly doing something really great" as opposed to "[company] is secretly doing something evil"
I wasn't aware of Hack Club before and wow, their fiscal sponsorship program is enormous: https://hackclub.com/fiscal-sponsorship/directory/ - looks like they cover more than 2,500 organizations!
The Python Software Foundation acts as a fiscal sponsor for a much smaller set of orgs (20 listed on https://www.python.org/psf/fiscal-sponsorees/) and it keeps our accounting team pretty busy just looking after those. Hack Club must have this down to a very fine art.
I wrote a bit more about PSF fiscal sponsorship here: https://simonwillison.net/2024/Sep/18/board-of-the-python-so...
Everything valve doing for linux is making such a huge impact.
The HL3 memes don't even seem fair to use anymore. I don't even want to un-seriously make joke fun of them at this point. They are just genuinely doing so much for the community.
This vulnerability is basically the worst-case version of what people have been warning about since RSC/server actions were introduced.
The server was deserializing untrusted input from the client directly into module+export name lookups, and then invoking whatever the client asked for (without verifying that metadata.name was an own property).
return moduleExports[metadata.name]
My guess is this won't be the last time we see security fallout from that design choice. Not because React is sloppy, but because it’s trying to solve a problem category that traditionally requires explicitness, not magic.
Me: oh cool, this is interesting, I don’t quite understand what exactly that means, let me read the thread to learn more…
The thread: > Replacing ECCA1 by version with step after the direction change could save something like 1% of the ecca1 bits size. Compiling agnosticized program instead of fixed lane program by ecca1 could save something like 1% as well (just guesses). Build of smaller ECCA1 would shorten binary portion, but it would be hardly seen in the ship size.
> Using agnosticized recipe in the fuse portion would definitely reduce its size. Better cordership seed and better salvo for gpse90 would help…
Dear lord I had no idea there’s this much jargon in the game of life community. Gonna be reading the wiki for hours
>This is saying that people in congressional leadership positions do 47% better than other members of Congress.
It's even more damning than that I think.
>we find that lawmakers who later ascend to leadership positions perform similarly to matched peers beforehand but outperform them by 47 percentage points annually after ascension.
The same person makes more after being put in a leadership position than before. That essentially removes any possibility of 'well maybe they're just more knowledgable and that's why they're in leadership.'
Incorrect title, the finding is "lawmakers who later ascend to leadership positions perform similarly to matched peers beforehand but outperform them by 47 percentage points annually after ascension." This is saying that people in congressional leadership positions do 47% better than other members of Congress.
> With their latest data measurements specific to the game, the developers have confirmed the small number of players (11% last week) using mechanical hard drives will witness mission load times increase by only a few seconds in worst cases. Additionally, the post reads, “the majority of the loading time in Helldivers 2 is due to level-generation rather than asset loading. This level generation happens in parallel with loading assets from the disk and so is the main determining factor of the loading time.”
It seems bizarre to me that they'd have accepted such a high cost (150GB+ installation size!) without entirely verifying that it was necessary!
I expect it's a story that'll never get told in enough detail to satisfy curiosity, but it certainly seems strange from the outside for this optimisation to be both possible and acceptable.
I often find we don't appreciate enough people accepting their failures and changing their mind. For some reason I see the opposite: people respecting those who "stick to their guns" or double down when something is clearly wrong. As you say, the context matters and these edits seem to be learning from the feedback rather than saving face since the sentiment stands, just in a less needlessly targeted way.
On the previous HN article, I recall many a comment talking about how they should change this, leave the politics/negative juju out because it was a bad look for the Zig community.
It would appear they listened to that feedback, swallowed their ego/pride and did what was best for the Zig community with these edits. I commend them for their actions in doing what's best for the community at the cost of some personal mea culpa edits.
The edit history of the announcement is quite a ride:
> [2025-11-27T02:10:07Z] it’s abundantly clear that the talented folks who used to work on the product have moved on to bigger and better things, with the remaining losers eager to inflict some kind of bloated, buggy JavaScript framework on us in the name of progress [1]
> [2025-11-27T14:04:47Z] it’s abundantly clear that the talented folks who used to work on the product have moved on to bigger and better things, with the remaining rookies eager to inflict some kind of bloated, buggy JavaScript framework on us in the name of progress [2]
> [2025-11-28T09:21:12Z] it’s abundantly clear that the engineering excellence that created GitHub’s success is no longer driving it [3]
---
1: https://web.archive.org/web/20251127021007/https://ziglang.o...
2: https://web.archive.org/web/20251127140447/https://ziglang.o...
3: https://web.archive.org/web/20251128092112/https://ziglang.o...
>EU vehicle safety regulations have supported a 36% reduction in European road deaths since 2010. By contrast, road deaths in the US over the same period increased 30%, with pedestrian deaths up 80% and cyclist deaths up 50%
I didn't know this, but it is absolutely crazy. Every EU politician who tries to subvert car safety should be dismissed and tried for endangering public safety.
NotJustBikes just put out a video about this issue - https://www.youtube.com/watch?v=--832LV9a3I
A couple years ago he also made a video about these trucks more broadly - https://www.youtube.com/watch?v=jN7mSXMruEo
What's truly maddening is how many of these vehicles which _do not_ meet European safety standards are _already_ in Europe. Walk around Hilversum in the Netherlands and you will see plenty of Dodge Rams (mostly 1500's, but there's even a 2500 Dually usually parked on the sidewalk ("pavement "for Brits) where my kids used to go to school). They're imported under "Individual Vehicle Approval" rules, exempting them from type safety requirements, and on top of that are almost always registered as "business vehicles" (you can tell from the V plate) which means they pay an absolute pittance in tax.
I moved here to get away from American kindercrushers (among other reasons) and I am profoundly concerned that Europe is being invaded by these machines.
(Edit) Worth noting is that a lot of Dutch street design is based on the idea that people _can_ share space with cars in dense, low speed environments, but that assumption flies out the window when the vehicles are so large you can't even see a kid walking or biking to school.
Further edit - source - https://www.motorfinanceonline.com/news/dodge-ram-registrati... 5,000 Dodge Rams imported in to Europe in 2023 alone.
> IBM has not exactly had a stellar record at identifying the future.
IBM invented/developed/introduced magnetic stripe cards, UPC Barcodes, the modern ATM, Hard drives, floppies, DRAM, SQL, the 360 Family of Mainframes, the PC, Apollo guidance computers, Deep Blue. IBM created a far share of the future we're living in.
I'm no fan of much of what IBM is doing at the moment but it could be argued that its consultancy/service orientation gives it a good view of how business is and is planning to use AI.
Ah Monotype. Knew before reading the article it would be them. Their whole MO is buying up smaller type foundries, massively increasing the licensing fees and shaking down the previous customers. They send in auditors, demand to see traffic data and threaten fines. Happened to me twice now.
Creating type is an extremely difficult and skilled discipline and designers deserve to be compensated fairly. However Monotype’s business practices are such that I won’t approve anything but open source fonts for new projects.
Blameless postmortem culture recognizes human error as an inevitability and asks those with influence to design systems that maintain safety in the face of human error. In the software engineering world, this typically means automation, because while automation can and usually does have faults, it doesn't suffer from human error.
Now we've invented automation that commits human-like error at scale.
I wouldn't call myself anti-AI, but it does seem fairly obvious to me that directly automating things with AI will probably always have substantial risk and you have much more assurance, if you involve AI in the process, using it to develop a traditional automation. As a low-stakes personal example, instead of using AI to generate boilerplate code, I'll often try to use AI to generate a traditional code generator to convert whatever DSL specification into the chosen development language source code, rather than asking AI to generate the development language source code directly from the DSL.
I built Cloudflare Workers?
I've seen a rumor going around that OpenAI hasn't had a successful pre-training run since mid 2024. This seemed insane to me but if you give ChatGPT 5.1 a query about current events and instruct it not to use the internet it will tell you its knowledge cutoff is June 2024. Not sure if maybe that's just the smaller model or what. But I don't think it's a good sign to get that from any frontier model today, that's 18 months ago.
A lot of people seem confused about this acquisition because they think of Bun as a node.js compatible bundler / runtime and just compare it to Deno / npm. But I think its a really smart move if you think of where Bun has been pushing into lately which is a kind of cloud-native self contained runtime (S3 API, SQL, streaming, etc). For an agent like Claude Code this trajectory is really interesting as you are creating a runtime where your agent can work inside of cloud services as fluently as it currently does with a local filesystem. Claude will be able to leverage these capabilities to extend its reach across the cloud and add more value in enterprise use cases
I can generate $20 billion in ARR this year too! I just need you to give me $100 billion and allow me to sell each of your dollars for 0.2 dollars.
I work on Bun.
Happy to answer any questions
OpenAI is basically just Netscape at this point. An innovative product with no means of significant revenue generation.
One one side it's up against large competitors with an already established user base and product line that can simply bundle their AI offerings into those products. Google will do just what Microsoft did with Internet Explorer and bundle Gemini in for 'Free' with their already other profitable products and established ad-funded revenue streams.
At the same time, Deepseek/Qwen, etc. are open sourcing stuff to undercut them on the other side. It's a classic squeeze on their already fairly dubious business model.
Adding to this: it's not just that the apprenticeship ladder is gone—it's that nobody wants to deal with juniors who spit out AI code they don't really understand.
In the past, a junior would write bad code and you'd work with them to make it better. Now I just assume they're taking my feedback and feeding it right back to the LLM. Ends up taking more of my time than if I'd done it myself. The whole mentorship thing breaks down when you're basically collaborating with a model through a proxy.
I think highly motivated juniors who actually want to learn are still valuable. But it's hard to get past "why bother mentoring when I could just use AI directly?"
I don't have answers here. Just thinking maybe we're not seeing the end of software engineering for those of us already in it—but the door might be closing for anyone trying to come up behind us.
I’ll be honest, while I have my doubts about the match of interests and cohesion between an AI company and a JS runtime company I have to say this is the single best acquisition announcement blog post I’ve seen in 20 years or so.
Very direct, very plain and detailed. They cover all the bases about the why, the how, and what to expect. I really appreciate it.
Best of luck to the team and hopefully the new home will support them well.
Overall this article is accurate and well-researched. Thanks to Daroc Alden for due diligence. Here are a couple of minor corrections:
> When using an Io.Threaded instance, the async() function doesn't actually do anything asynchronously — it just runs the provided function right away.
While this is a legal implementation strategy, this is not what std.Io.Threaded does. By default, it will use a configurably sized thread pool to dispatch async tasks. It can, however, be statically initialized with init_single_threaded in which case it does have the behavior described in the article.
The only other issue I spotted is:
> For that use case, the Io interface provides a separate function, asyncConcurrent() that explicitly asks for the provided function to be run in parallel.
There was a brief moment where we had asyncConcurrent() but it has since been renamed more simply to concurrent().
Loved playing with it! https://strudel.cc/?qVv8Cr0OD6cc
"NFTs show relatively small groups of tens of thousands of people can work together collectively for a common good" - Garry Tan, 23 March 2022
https://x.com/garrytan/status/1506769562468958210
"If you own an NFT..." you are an innovator among innovators. - Garry Tan, 30 September 2021 [quote tweet]
https://x.com/garrytan/status/1443460589049704450
Web3 vs Earlier incarnations technology adoption curve [image tweet]
https://x.com/garrytan/status/1521530531568963584
"airdrops going mainstream could truly upend some % of the centralized ad-based economy of Web2" - Garry Tan, 25 Dec 2021
https://x.com/garrytan/status/1474826162408808448
> It turns out you can create a UDP socket with a protocol flag, which allows you to send the ping rootless
This is wrong, despite the Rust library in question's naming convention. You're not creating a UDP socket. You're creating an IP (AF_INET), datagram socket (SOCK_DGRAM), using protocol ICMP (IPPROTO_ICMP). The issue is that the rust library apparently conflates datagram and UDP, when they're not the same thing.
You can do the same in C, by calling socket(2) with the above arguments. It hinges on Linux allowing rootless pings from the GIDs in
$ sysctl net.ipv4.ping_group_range
net.ipv4.ping_group_range = 999 59999
EDIT2: more spelling mistakes
> As a blind person, AI has changed my life.
Something one doesn't see in news headlines. Happy to see this comment.
Apple has a video understanding model too. I can't wait to find out what accessibility stuff they'll do with the models. As a blind person, AI has changed my life.
The length of copyright is absurd. Corporations have hijacked a concept that should exist on human timescales.
Ideally, a child could legally provide their own spin on IP they consumed by the time they reach adulthood. But also, people need to make a living.
I actually think the original 14+14 year copyright is the right balance. It gives people time to make their profits, but also guarantees the right of people to tweak and modify content they consume within their lifetime. It's a balanced time scale rather than one that exists solely to serve mega corporations giving them the capability to hold cultural icons hostage.
It's not about productivity at all. These same companies were commissioning studies during Covid that told their analysts "look how productive our employees are now that they are working from home!"
It's about crushing labor.
WFH forces employers to compete. It gives a lot of power to employees, because they can apply for far more roles, work fewer hours, moonlight for multiple companies, etc, apply for other jobs during work hours, etc. These companies know that white collar workers are not fungible. Their intellectual workers are genuinely very difficult to replace and produce a lot of value.
For talent that isn't fungible, it's RTO. For talent that is fungible, offshoring.
Reading all of these takes stating WFH leads to poor productivity simply doesn’t make sense to me.
If your employees cannot be trusted to fulfil their responsibilities (whether in an office, their home or a tent in a woodland) that is not a geographical issue. It is a mentality issue and you are always going to face productivity issue from that employee regardless of from where they work.
I’ve been told time and time again by an array of managers in a bunch of departments and companies that my productivity never changes. That is regardless of whether I am travelling or at home. This is including being in Sri Lanka during their worst economical crisis and facing power cuts of 8 - 12 hours everyday. As a responsible adult I prepared in advance. I bought power banks which could charge my laptop and ensured they were charged when the power worked. I bought SIM cards for all mobile networks and ensured I had data. It really is simply a matter of taking responsibility of one’s situation and having a sense of respect for, and from, your employer/employee.
Forcing people into working conditions in which they are uncomfortable is only going to harbour resentment towards the company and if you are in a country where workers actually have real rights you will have a hard time firing them.
I fear that this is all simply a smokescreen for the authoritarian shift which has occurred throughout the globe. It started pre pandemic and was exasperated during it. Scary times lay ahead.
To avoid the advent calendar, this may be more useful:
https://en.wikipedia.org/wiki/2026_in_public_domain
I typically jump on the latest macOS with enthusiasm. I once made the mistake to install the beta version of the next os, and well, that didn't go well for me. But typically, within X.1, I'm there.
However something shifted since this "visionOS" melted version of macOS (Tahoe); where I have absolutely no intension to upgrade from Sequoia. I hope they will fix it by the time I'll be forced to upgrade (post support deadline).
It started with the macOS that brought the iOS settings panel. We went from a logical structure of easily findable stuff to a complete mess. Just open the "Keyboard" settings on macOS today and it's bewildering how they could ship this and think this is fine. Steve would roll in his grave.
The process to allow running applications that are unsigned is just a horrible hack. It feels like a last minute "shove it and move on!".
By 2035 I wonder if we'll be all running KDE or WindowMaker and the hell with modern OS GUI.
From a Gestalt standpoint, human relations with desktop computers are not the same as with thumb driven mobile OS or air-pinch driven vision OS, period. The hell with "glass" or "flat" design. Desktop OS should be as forgettable as possible, as it's about having long stints of flow, not giving a feeling of "air" or "play".
> With JSON, you often send ambiguous or non-guaranteed data. You may encounter a missing field, an incorrect type, a typo in a key, or simply an undocumented structure. With Protobuf, that’s impossible. Everything starts with a .proto file that defines the structure of messages precisely.
This deeply misunderstands the philosophy of Protobuf. proto3 doesn't even support required fields. https://protobuf.dev/best-practices/dos-donts/
> Never add a required field, instead add `// required` to document the API contract. Required fields are considered harmful by so many they were removed from proto3 completely.
Protobuf clients need to be written defensively, just like JSON API clients.
Siri is probably among the products which had the most exposure to users (probably a billion+ users throughout iPhone's history) without capturing that opportunity to actually do anything meaningful with the huge user base it got for free.
A decade and a half is insane timeline in tech industry, and huge majority of users use Siri the same way today as 15 years ago, setting a timer or an alarm clock.
If they had 0 improvements over these 15 years the situation wouldn't be much different than today.
Firefox was seriously a better browser, not just "implements standards better". It ran faster, it had tabs (wow!) and at one point it got Firebug which let you have a console INSIDE the browser that showed information you could print with `console.log`, I kid you not.
It was a better browser through and through, maybe because MS slept on IE or maybe not, but in the end it isn't revisionist to say they beat MS's proverbial posterior because the browser was better.
Sounds so authoritarian. Luckily, in the UK you only have to scan your face and ID to access cat photos.
"We're also offering the option to transfer from the MPK to SF office for those people whose commute would be the same or better with that change."
So wait, you'll be able to switch offices even though your team might be in the second one? What's the benefit of working remote from your team but next to random, noisy people?
Instagram chief orders quiet layoffs to please investors in 2026
fixed that title for you
I agree with each and every single slide in this presentation; I do. I also know that in each and every company I have ever worked for, none of this is going to fly. Especially, "Attending meetings is a choice." Just like paying taxes is a choice; got it.
---
For decades, I have been asking for agendas; I have asked for clarification on what to do to prepare; I have even suggested that we have solid outcomes. None of which are followed nor what anyone else wants.
Even as a leader at organizations where I can enforce this on my team, it makes absolutely no difference. Hell, Google Calendar (we use Workspace at my current org) doesn't even have solid support for good meeting invite commentary. And, even if it did, 99.99999% of folks wouldn't read any of it anyway.
> I'm shocked
India is currently run by a nationalist regime headed by the so called "butcher of Gujarat"[1], there isn't much that would shock me wrt to that lot's totalitarian tendencies.
[1] https://en.wikipedia.org/wiki/Public_image_of_Narendra_Modi
I agree that is what he is doing, but I can also justify adding fentanyl to every drug sold in the world as "making it better" from a business perspective, because it is addictive. Anyone who ignores the moral or ethical angle on decisions, I cannot take seriously. It's like saying that Maximizing Shareholder Value is always the right thing to do. No, it isn't. So don't say stupid shit like that, be a human being and use your brain and capacity to look at things and analyze "is this good for human society?".
I'm shocked by people and state using the crutch of cyber crime or scams to push a totalitarian solution to a problem that is better solved by improved education and targeted campaigns against common security pitfalls.
I abhor any decision that robs even a grain of my individual freedom.
Well props to them for continuing to improve, winning on cost-effectiveness, and continuing to publicly share their improvements. Hard not to root for them as a force to prevent an AI corporate monopoly/duopoly.
In my opinion, we're possessed by a cultural epidemic of think pieces doing rich and nuanced science history, but wrongly framed in the form of correcting "myths" that, in their substance amount to quibblings over narrative emphasis. It's easy to get taken in by the framing because it truly is enlightening, and the argument goes down so smooth because its embedded in a rich, curious, and fascinating scientific history that otherwise embodies best practices I would happily celebrate.
But the key details about the story of penicillin are that a moldy plate showed bacteria-free clearing, Fleming saw it, isolated the mold, proved its germ-killing filtrate and published the finding, which is the heart of the story and which is not a myth.
I'm sure it's true enough that St Mary's windows were usually kept shut to keep pathogens in and contaminants out, that London's August 1928 cold snap would have slowed staph growth, that Fleming's first notes Or 8 weeks later than the actual event, and that a modern plate seeded with bacteria first will not produce the celebrated halo unless the mold is given a head start. The article makes much of the fact that today’s researchers cannot reproduce the famous halo if they add staph first, yet that difficulty rebuts a sequence Fleming never claimed to have used.
These points are significant, even fascinating, yet the article inflates them into a strobe-lit "MYTH" banner, turning normal human imprecision about times and temperatures into evidence of wholesale fiction, which abuses the ordinary friction of any retrospective account and punishes the story for the very human messiness that makes it instructive.
The window quibble, the incubator gap, and the replication protocol do not touch the central, uncontested fact that chance contamination plus observational curiosity gave medicine its first antibiotic.
When I was a cartographer in the 1500s I used to hide dragons, sea serpents and the occasional heretical inscription in the blank bits, because at least back then the Holy Roman Emperor had the decency to pretend he didn’t notice as long as the tax broders were correct.
Now look at us: the Swiss federal cartographers, salaried, pensioned, triple-proofread, still cannot resist smuggling a naked woman and a cheeky marmot into the official topography. And the admisntration? They wait until the perpetrator has safely retired on full index-linked benefits, then solemnly announce the marmot will be "removed in the next revision cycle, pending environmental-impact assessment of the pixel."
This is what passes for rebellion inside the European regulatory state: a rodent drawn at 1:25 000 scale that offends precisely no one and will be erased by a civil servant who wasn’t even born when it was sketched. Truly the revolutionary spirit of our continent has been reduced to a change-request ticket with fourteen mandatory approvers and a carbon-copy to Bern.
I fill in another compliance form and weep for the age when men risked the stake for a badly drawn leviathan.
"advertising would make ChatGPT a better product."
And with that, I will never read anything this guy writes again :)
Amongst the discussion of rootkits and anti-cheat, I would like to add that part of the reason it is necessary is caused by the game companies that took away the standard method of playing multiplayer -- players running their own servers.
It used to be pretty easy to just ban people from playing, now we're 100% reliant on their ability to do it. So we have anti-cheat which roots our computer, and still doesn't 100% solve the problem.
> CEO walking up behind my desk and pointing out that I'd done MOV AX, 0 when I could have done XOR AX, AX
Now that's what I call micromanagement.
(sorry couldn't resist)
> This is a search tool that will only return content created before ChatGPT's first public release on November 30, 2022.
The problem is that Google's search engine - but, oddly enough, ALL search engines - got worse before that already. I noticed that search engines got worse several years before 2022. So, AI further decreased the quality, but the quality had a downwards trend already, as it was. There are some attempts to analyse this on youtube (also owned by Google - Google ruins our digital world); some explanations made sense to me, but even then I am not 100% certain why Google decided to ruin google search.
One key observation I made was that the youtube search, was copied onto Google's regular search, which makes no sense for google search. If I casually search for a video on youtube, I may be semi-interested in unrelated videos. But if I search on Google search for specific terms, I am not interested in crap such as "others also searched for xyz" - that is just ruining the UI with irrelevant information. This is not the only example, Google made the search results worse here and tries to confuse the user in clicking on things. Plus placement of ads. The quality really worsened.
In contrast to others, I just want to say that I applaud the decision to take a moral stance against AI, and I wish more people would do that. Saying "well you have to follow the market" is such a cravenly amoral perspective.
I love how a number crunching program can be deeply humanly "horrorized" and "sorry" for wiping out a drive. Those are still feelings reserved only for real human beings, and not computer programs emitting garbage. This is vibe insulting to anyone that don't understand how "AI" works.
I'm sorry for the person who lost their stuff but this is a reminder that in 2025 you STILL need to know what you are doing and if you don't then put your hands away from the keyboard if you think you can lose valuable data.
You simply don't vibe command a computer.
Well as a single datum:
After I got laid off in late 2023 I had a devil of a time finding work (despite having AI experience) to the point my unemployment ran out -
And I was 20 years as a dev and tech lead and full stack, (never had trouble finding work) including stints as a leading EM and CTO, I’ve been an industry award winning innovation lead,a digital studio director, switching tech stacks and cloud certs all my career..mentoring juniors and doing podcasts and writing white papers etc, but peanuts - nothing
Getting ghosted by 25 year olds in interviews and doing rounds and rounds and leetcode and all that but no success- for example I had a 7 round interview with NBCUniversal in 2023 and then got ghosted (I probably doged a bullet since they had subsequent layoffs)
a 12 month stint with nothing - we lost our savings as my wife got laid off too
Since then I pivoted to AI and Gen AI startups- joining incubators and finally got some work or at least cofounded some AI startups- now money is tight and I dont have health insurance but at least I have a job… it sucks as a over 45 year old as I have so much experience but no one cares.. still dont have much grey hair so I can pass for 40 to get noticed
No one stable is hiring or your resume just goes to a dead letter queue or is lost in ether or lost amid all the ai generated resumes out there - young ML and PHds and people under 40 seem to be getting work in Gen AI but thats about it
networking is the only game left and most good recruiters I know got laid off too
At least I’ve built up production experience in agents and context engineering RL, pytorch, langchain/LangGraph, RAG, KGa, etc python, BAML, LLM and LLMops to add to my years of full stack work
I’ve seen Picallilli’s stuff around and it looks extremely solid. But you can’t beat the market. You either have what they want to buy, or you don’t.
> Landing projects for Set Studio has been extremely difficult, especially as we won’t work on product marketing for AI stuff, from a moral standpoint, but the vast majority of enquiries have been for exactly that
The market is speaking. Long-term you’ll find out who’s wrong, but the market can usually stay irrational for much longer than you can stay in business.
I think everyone in the programming education business is feeling the struggle right now. In my opinion this business died 2 years ago – https://swizec.com/blog/the-programming-tutorial-seo-industr...
The overhyped tweet from the robinhood guy raising money for his AI startup is nicely brought into better perspective by Thomas Bloom (including that #124 is not from the cited paper, "Complete sequences of sets of integer powers "/BEGL96):
> This is a nice solution, and impressive to be found by AI, although the proof is (in hindsight) very simple, and the surprising thing is that Erdos missed it. But there is definitely precedent for Erdos missing easy solutions!
> Also this is not the problem as posed in that paper
> That paper asks a harder version of this problem. The problem which has been solved was asked by Erdos in a couple of later papers.
> One also needs to be careful about saying things like 'open for 30 years'. This does not mean it has resisted 30 years of efforts to solve it! Many Erdos problems (including this one) have just been forgotten about it, and nobody has seriously tried to solve it.[1]
And, indeed, Boris Alexeev (who ran the problem) agrees:
> My summary is that Aristotle solved "a" version of this problem (indeed, with an olympiad-style proof), but not "the" version.
> I agree that the [BEGL96] problem is still open (for now!), and your plan to keep this problem open by changing the statement is reasonable. Alternatively, one could add another problem and link them. I have no preference.[2]
Not to rain on the parade out of spite, it's just that this is neat, but not like, unusually neat compared to the last few months.
[1] https://twitter.com/thomasfbloom/status/1995083348201586965
[2] https://www.erdosproblems.com/forum/thread/124#post-1899
Here's 12 Sysadmin/DevOps (they're synonyms now!) challenges, straight from the day job:
1. Get a user to stop logging in as root.
2. Get all users to stop sharing the same login and password for all servers.
3. Get a user to upgrade their app's dependencies to versions newer than 2010.
4. Get a user to use configuration management rather than scp'ing config files from their laptop to the server.
5. Get a user to bake immutable images w/configuration rather than using configuration management.
6. Get a user to switch from Jenkins to GitHub Actions.
7. Get a user to stop keeping one file with all production secrets in S3, and use a secrets vault instead.
8. Convince a user (and management) you need to buy new servers, because although "we haven't had one go down in years", every one has faulty power supply, hard drive, network card, RAM, etc, and the hardware's so old you can't find spare parts.
9. Get management to give you the authority to force users to rotate their AWS access keys which are 8 years old.
10. Get a user to stop using the aws root account's access keys for their application.
11. Get a user to build their application in a container.
12. Get a user to deploy their application without you.
> Claude often ignores CLAUDE.md
> The more information you have in the file that's not universally applicable to the tasks you have it working on, the more likely it is that Claude will ignore your instructions in the file
Claude.md files can get pretty long, and many times Claude Code just stops following a lot of the directions specified in the file
A friend of mine tells Claude to always address him as “Mr Tinkleberry”, he says he can tell Claude is not paying attention to the instructions on Claude.md, when Claude stops calling him “Mr Tinkleberry” consistently
> You don't need a computer science background to participate - just a little programming knowledge and some problem solving skills will get you pretty far.
Every time I see this I wonder how many amateur/hobbyist programmers it sets up for disappointment. Unless your definition of “pretty far” is “a small number of the part ones”, it’s simply not true.
I have a pixel watch, and my main use for it is setting reminders, like "reminder 3pm put the laundry in the dryer". It's worked fine since the day I bought it.
Last week, they pushed an update that broke all of the features on the watch unless I agreed to allow Google to train their AI on my content.
The real creepy thing is the way they force you to give up your data with these products. If it were just useful add ons, it wouldn’t bother me, but the fact that Gemini requires you to turn activity history off for paid plans for the promise they won’t train on your data or allow a person to view your prompts is insanity. If you’re paying $20 for Pro or 249.99 for Ultra, you should be able to get activity history without training or review or storing your data for several years.
The NT paths are how the object manager refers to things. For example the registry hive HKEY_LOCAL_MACHINE is an alias for \Registry\Machine
https://learn.microsoft.com/en-us/windows-hardware/drivers/k...
In this way, NT is similar to Unix in that many things are just files part of one global VFS layout (the object manager name space).
Paths that start with drive letters are called a "DOSPath" because they only exist for DOS compatibility. But unfortunately, even in kernel mode, different sub systems might still refer to a DOSPath.
Powershell also exposes various things as "drives", pretty sure you could create your own custom drive as well for your custom app. For example, by default there is the 'hklm:\' drive path:
https://learn.microsoft.com/en-us/powershell/scripting/sampl...
Get-PSDrive/New-PSDrive
You can't access certificates in linux/bash as a file path for example, but you can in powershell/windows.
I highly recommend getting the NtObjectManager powershell module and exploring about:
https://github.com/googleprojectzero/sandbox-attacksurface-a...
ls NtObject:\
I took CS193P when it was first offered in 2007; one of my favorite classes at Stanford because it was so hands-on. At the time few people had iPhones, so everyone in the class got a free iPod Touch for development. My final project was a photo sharing app with a Polaroid shake to reveal mechanic… lightly influenced Instagram which Kevin and I built a few years later!
I've been messing around with GitLab as a self hosted alternative for a few years. I do like it, but it is resource intensive!
For the past few days I've been playing with Forgejo (from the Codeberg people). It is fantastic.
The biggest difference is memory usage. GitLab is Ruby on Rails and over a dozen services (gitlab itself, then nginx, postgrest, prometheus, etc). Forgejo is written in go and is a single binary.
I have been running GitLab for several years (for my own personal use only!) and it regularly slowly starts to use up the entirety of the RAM on a 16GB VM. I have only been playing with Forgejo for a few days, but I am using only 300MB of the 8 GB of RAM I allocated, and that machine is running both the server and a runner (it is idle but...).
I'm really excited about Forgejo and dumping GitLab. The biggest difference I can see if that Forgejo does not have GraphQL support, but the REST API seems, at first glance, to be fine.
EDIT: I don't really understand the difference between gitea and forgejo. Can anyone explain? I see lots of directories inside the forgejo volume when I run using podman that clearly indicate they are the same under the hood in many ways.
EDIT 2: Looks like forgejo is a soft fork in 2022 when there were some weird things that happened to governance of the gitea project: https://forgejo.org/compare-to-gitea/#why-was-forgejo-create...
I'm actually pleasantly surprised to see a 2025 edition, last year being the 10th anniversary and the LLM situation with the leaderboard were solid indications that it would have been a great time to wrap it up and let somebody else carry the torch.
It's only going to be 12 problems rather than 24 this year and there isn't going to be a gloabl leaderboard, but I'm still glad we get to take part in this fun Christmas season tradition, and I'm thankful for all those who put in their free time so that we can get to enjoy the problems. It's probably an unpopular stance, but I've never done Advent of Code for the competitive aspect, I've always just enjoyed the puzzles, so as far as I'm concerned nothing was really lost.
Advent of Code is one of the highlights of December for me.
It's sad, but inevitable, that the global leaderboard had to be pulled. It's also understandable that this year is just 12 days, so takes some pressure off.
If you've never done it before, I recommend it. Don't try and "win", just enjoy the problem solving and the whimsy.
I have a special spot that gets extremely annoyed when I feel I have to spend a necessary time just figuring out what something is. The page suffers from "developer-brain" marketing, where they describe the technology used to build it (cloud-native images) rather than what the product actually is. Why not just lead with that it is a Linux distro/version focused on gaming. In the beginning, I thought it was new hardware, then I actually thought it was a streaming service. the website does a poor job of simply stating what it actually is. Then again, I might be stupid, but the problem is that a lot of people are stupid. And I guess many non-developers are going to have a hard time just figuring out what the hell this thing is. I’m probably being too cranky now, but the page reads like a combination of the worst from developers-speak combined with the rest of marketing-speak.
Man there needs to be some sort of amusement park where you try out common emergency safety mechanisms.
I've never broke / pulled a fire alarm, I'm sure I can, but let me.
ALSO EVERY AIRPORT SHOULD HAVE MOCK EMERGENCY AIRPLANE DOORS FOR PEOPLE TO TRY OUT.
To say the quiet part out loud, I don't think any serious companies have any intention to build a data center in space. There is no benefit in actually trying this. There is however, benefit in saying you'll do it to advance a narrative and distract from the problems terrestrial data centers are facing to an audience that mostly doesn't understand how heat transfer in a vacuum works.
That's historical revisionism. The percentage of American adults over age 25 who have a college degree was only 20% as recently as 1990. When America was truly at the top of the world in the 1950s and 1960s, it was under 10%. A high fraction of college attendance is better correlated with the 21st century decline in America's situation.
This view seems to be common, but I think it misses what incredible alchemy comes from making people who come in for “job training” (like I did) spend 4 years in close proximity with research, academic freedom, liberal arts, and at least an attempt at some kind of intellectual idealism separate from economic incentive. It’s peanut butter and chocolate that has served democracy and its people well by having a middle class that is not just productive, but truly educated. It’s weird and it has problems, but it’s also wonderful, and we should not try to sever the two so we can more “efficiently” crank out credentials.
Higher ed is like employer based health insurance in that they are both weird path dependent historical accidents.
People want cheap healthcare, and it got shoehorned into an odd employer fringe benefit system that really is not at all related healthcare in any intrinsic way.
People want job training, and it got shoehorned into extra departments at liberal arts universities intended as aristocrat finishing schools. Job training really has little to no relationship to liberal arts.
And now both those two systems are failing to deliver those benefits because those benefits which were initially afterthought add-ons have outgrown the institutions that were their hosts. It's akin to a parasitic vine that is now much larger than the tree it grew on and is crushing it under its weight. Both will die as a result.
Hi, I'm the founder.
While what you're saying isn't impossible, it's unlikely. In the event it did happen, Bazzite is a fork, a signing key, and a couple forked Fedora Copr repos away from being made completely in someone else's control.
Some of the alternatives that the author suggests (Slack, Discord, Matrix rooms) are so much worse to search for answers in. Stack overflow has many disadvantages, but it is extremely good at being a publicly searchable repository of answers to common questions
As much as I love Bazzite at end of the day it's still a custom distro and every single day there is a chance they just close the project down and move on. Happened to so many distros in the past, this is not out of question. I’m not saying “big corporate” distros are better but personally I'd rather stick to something more mainline.
Hopefully Valve will release a general version of SteamOS with Steam Machine coming (and even they are questionable with their track record)
Better than asking "is college worth the cost," and getting into ROI calculations per major is asking "could we provide a similar (or better) educational and social experience at a fraction of the cost"? To that the answer is yes.
Who are you thinking of? Bezos, Zuckerberg, Gates, Musk all came from wealthy families. They all had the safety net of their family wealth if their business didn’t work out. Even someone who had parents to pay their student loans is relatively privileged, not everyone starts their 20’s debt-free and able to take financial risks.
Clippy was definitely hostile. It would constantly monitor user actions just so that it could interrupt us. Wasted CPU cycles and our time when CPUs weren't very fast. Clippy was hated by everyone. It was not just useless. It was intrusive, wasteful, and hostile. I can't believe my eyes that anyone could think that Clippy is an appropriate mascot for anything good. If anything, Clippy would be a perfect mascot for the trillion dollar companies that exploit our data.
The “one that works out” can also give you a misrepresentation of how the world works and a false sense of how lucky one should expect to be over a long period of time.
At an earlier point in my life, I had been applying to many well-known big tech companies right out of school (not a top school either). I never got a reply from any of them so I ended up accepting a local job with a non-tech company after months of searching.
But I didn’t give up my hopes and kept applying to big tech, and while I did manage to get the occasional interview with some mediocre companies or the random startup, I also miserably failed all of them too.
At some point during my long period of despair at never getting a better job, my very top pick (and arguably one of the best tech companies in the world at the time) reached out to me. Even more miraculously, I somehow passed their interview (the only tech interview I passed in the prior year) and accepted a job there.
I really enjoyed working there. Some of the best years of my life. And my performance reviews were great too, so the imposter syndrome from having failed so many tech job interviews sort of faded into the background. But after a while, perhaps due to the “hedonic treadmill” mentality, I thought I could do better. So I left to join a startup.
Well, the startup failed, as startups tend to do, but what I didn’t expect and what caught me off guard was that I was now back in the same situation I was in right after graduating from college. Don’t get me wrong—having “the name” on my resume now meant I could get at least one chance at an interview about anywhere. But much like the first round that I tried to forget about, I once again failed all the interviews.
Unfortunately, this second time around never procured a “get out of jail free” card.
So I guess my lesson is: 1) there’s a lot of luck involved in these things, 2) if life gives you a winning lottery ticket at some point, don’t throw it away for the chance to win an even bigger lottery, and 3) that famous saying about “the only actions regretted are those not taken” is absolutely, totally wrong—almost all of my regrets in life relate to taking some action I shouldn’t have rather than inaction.
This is why having a safety net and resources to try again is so powerful. Given enough chances you will make it, big. That means the #1 factor in success is the number of chances you get to fail and try again, not necessarily how inherently good you are. I try to remind myself of this often. I have been given so many chances, and I took them.
Quite the opposite: clippy was useless but not hostile which is a sobering contrast to software that is hostile and therefore worse than useless.
It's a cute nostalgic way to say "the bar was on the floor and you blew it anyway."
Clippy was never open source or "good" in any way, it not selling your data was a result of its time, not a conscious choice by its creators. The entire forced clippy "movement" is incredibly poorly thought out
This clippy white-washing is annoying as hell. Just like clipppy was.
While I think there's significant AI "offloading" in writing, the article's methodology relies on "AI-detectors," which reads like PR for Pangram. I don't need to explain why AI detectors are mostly bullshit and harmful for people who have never used LLMs. [1]
1: https://hn.algolia.com/?dateRange=all&page=0&prefix=true&que...
this is one of those HN style comments where business acumen and pertinent sarcasm are wholly indistinguishable .
I can't wait for the instructions to start having ads embedded.
2. Place the turkey in your GE Two in One Oven set to 350, cooking for 10 minutes a lbs.
3. While waiting for your Turkey to finish cooking, why not have an ice cold Coke Zero? Click here for nearby locations.
4. Remove Turkey from the oven, let rest for ten minutes while listening to Lady Gaga and Bruno Mars sing "Die with a Smile" on Spotify.
Oh wow. Did I really write that 11 years ago! How time flies.
> Let’s rip the Band-Aid off immediately: If your underlying business process is a mess, sprinkling "AI dust" on it won’t turn it into gold. It will just speed up the rate at which you generate garbage.
In the world of Business IT, we get seduced by the shiny new toy. Right now, that toy is Artificial Intelligence. Boardrooms are buzzing with buzzwords like LLMs, agentic workflows, and generative reasoning. Executives are frantically asking, "What is our AI strategy?"
But here is the hard truth:
There is no such thing as an AI strategy. There is only Business Process Optimization (BPO).
This is well-expressed, and almost certainly true for an overwhelming majority of companies.
> most normal people don't know what Claude or Gemini are
In think the point is that they don’t need to know what Gemini is, they just need to know Google, which they most definitely do.
IMO ads rollout won’t be as simple as you’re describing it. A lot of people have switched from Google search to AI specifically because it isn’t filled with SEO, ad filled nonsense. So they’ll need to tread very, very carefully to introduce it without alienating customers. Not to mention mollifying advertisers who are nervous what their product will be shown alongside and OpenAI will probably struggle to offer iron clad guarantees about it. And people generally speaking don’t like ads. If competitors like Google are able to hold out longer with no ads (they certainly aren’t wanting for ad display surfaces) they might be able to pull users away from OpenAI.
IMO pivoting to ads is a sign of core weakness for OpenAI. Anyone trying to set up their own ad network in 2025 has to reckon with Google and Meta, the two absolute behemoths of online ads. And both also happen to be major competitors of OpenAI. If they need ads that’s a problem.
- ~1 billion users in just 3 years
- Extremely personal data on users
- Novel way of introducing and learning more about sponsored products
- Strong branding for non-techie people (most normal people don't know what Claude or Gemini are)
- An app that is getting more and more addictive/indispensable
I think OpenAI is going to kill it in ads eventually. This is why Meta and Google went all in on AI. Their lucrative digital ad business is in an existential threat.
I think people who kept saying there is no moat in AI is about to be shocked at how strong of a moat there actually is for ChatGPT.
All free LLM chat apps will need to support ads or they will eventually die due to worse unit economics or run out of funding.
PS. Sam just said OpenAI's revenue will finish at $20b this year. 6x growth from 2024. Zero revenue from non-sub users. What do you guys think their revenue will end up in 2026?
I guess this could also have a knock-on effect, in that ChatGPT will steer it's users away from topics advertisers might find distasteful
Like it might not want to tell you about negative health effects from McDonalds, if McDonalds becomes a major source of ad revenue
I’d really, really like to know what microcontroller family this was found on. Assuming that this is a safety processor (lockstep, ECC, etc) it suggests that ECC was insufficient for the level of bit flips they’re seeing — and if the concern is data corruption, not unintended restart, it means it’s enough flips in one word to be undetectable. The environment they’re operating in isn’t that different from everyone else, so unless they ate some margin elsewhere (bad voltage corner or something), this can definitely be relevant to others. Also would be interesting to know if it’s NVM or SRAM that’s effected.
> Why does bad code bother engineers so much?
I’ll take a stab.
Because I’m being held accountable for the bad results of the bad code. Because I’m being held to task to fix the problems caused by the bad code on a schedule that I didn’t agree with. Because management is using someone else’s bad code to deny me a positive annual review.
“You’re a senior engineer - why did fixing this take so long?” Because of the garbage I had to wade through to even find the problem. Because I wasn’t yet working here when this was written and I have to take time to get my head around the insanity.
Yes, these are management problems. I’ve spent years managing managers and attempting to educate them on how bad code causes errors and delays. But for many reasons, it’s a Sysiphean task.
I've read a few of this guys posts now and have consistently been rubbed the wrong way by them. I think I know why now. It's not that he's wrong. His analysis is reasonable and straightforward. I think it's that the basis for his analysis is ultimately a form of nihilism, coming from someone who (maybe?) used to be an idealist but was burnt by a bad experience and must now explain why believing in anything is misguided.
My instinct after reading this article is to pull back a bit and ask some larger questions. Why is it necessary for big tech companies to act this way? Why does bad code bother engineers so much? Are they actually misguided for feeling like bad code is a catastrophe, or is it really the fault of the broader economic sphere we all inhabit? Is it actually maturity to reconcile ourselves to drift powerlessly as faceless and titanic forces sculpt our reality? So many possible questions.
Every time that I read this about remote work, all I can think is how much I miss IRC and the culture that came from it.
We were doing remote work effectively decades ago. Don't have hallway conversations to fix bugs? Easy, just post your problems on the team chat and someone (often one of several people) would love to drop by to help.
I'm not sure exactly all of the forces that have led to this changing so much, but I'm certain that merely blaming "remote work" isn't it.
Somehow we were better at using remote tools while literally in the same office than some teams are at using them now while fully remote.
what I see alot is that the syntax and overall code architecture is text book, but its the completely wrong approach that creates extremely complicated tech debt. All the code reviews will be on the syntax, and none on the big picture of the business problem, or whether the implementation is overcomplicated.
in the short run (1-2 years) there is no repercussion for this, but eventually making changes will be extremely risky and complicated. The individuals that built the software will lord over everyone else with their arcane knowledge of this big pile of junk
The live feed buries the only useful information at the very bottom of the article:
> The plane manufacturer says it has found that intense radiation from the Sun could corrupt data crucial to flight controls.
> It’s thought most will be able to undergo a simple software update.
> The issue was discovered after a JetBlue aircraft en-route from Mexico to the United States in October experienced a ‘sudden drop in altitude’.
> The plane made an emergency landing, with reports at the time suggesting 15 to 20 people suffered minor injuries.
> It’s thought the incident was caused by intense solar radiation, which corrupted data in a computer used to help control the aircraft.
I studied at Utrecht University and all the programming classes in the Bachelor were C#, Visual Studio, XNA, DirectX. Windows. Database class i had to learn in Proprietary Microsoft tools too. All Microsoft stuff. Sure nobody would complain if you did stuff on Linux but all the support by TAs and teachers was on Microsoft platforms only.. The Master was much better but the Bachelor basically was grooming people to become Microsoft consultants.
If the rot starts at the core of your education curriculum there is no saving your dependence on Microsoft.
I always found this choice puzzling to teach people proprietary technologies in a public institution. This was before DotNet core and VSCode was a thing and Microsoft hadnt whitewashed themselves to look like an open source friendly brand yet.
I've used Molly for over a year. Overnight it lost the device registration and will not contact the servers to re-register. The backup feature also does not work which left me dead in the water for several days with no fix. I switched back to signal and had to start a new database. It was a disaster. YMMV
I was hoping, from the title ("Geo-Unblocked") that this would be about arranging an IP address block that wasn't associated with the UK, rather than just selectively running some traffic through a VPN.
> What "proprietary blobs" does Signal have?
As they say in the Github readme, FCM and Google Maps.
FCM doesn't technically require a blob — it's just that Google wants you to think it does. I reverse engineered their library and it turned out to be a criminally over-engineered wrapper around two broadcast receivers. So, the Mastodon app is proudly the first app ever to both support FCM push notifications, and be 100% open-source.
Don't use all-MiniLM-L6-v2 for new vector embeddings datasets.
Yes, it's the open-weights embedding model used in all the tutorials and it was the most pragmatic model to use in sentence-transformers when vector stores were in their infancy, but it's old and does not implement the newest advances in architectures and data training pipelines, and it has a low context length of 512 when embedding models can do 2k+ with even more efficient tokenizers.
For open-weights, I would recommend EmbeddingGemma (https://huggingface.co/google/embeddinggemma-300m) instead which has incredible benchmarks and a 2k context window: although it's larger/slower to encode, the payoff is worth it. For a compromise, bge-base-en-v1.5 (https://huggingface.co/BAAI/bge-base-en-v1.5) or nomic-embed-text-v1.5 (https://huggingface.co/nomic-ai/nomic-embed-text-v1.5) are also good.
> Second, even if I installed a VPN on my main machine, what about my phone? My laptop? My desktop? Every device would need the VPN running, and I’d have to remember to connect it before browsing. It’s messy.
This is what routers are for. My router (a cheap fanless box with several network ports running linux) is the only thing on my network that knows there's a VPN. I can selectively route whatever I want through it, including having a separate SSID/VLAN from which everything is routed through the VPN. It's wireguard based so there's no "installing a VPN", just an interface/network configured in systemd-networkd (once, on the router).
Edit: Routing by domain name could be tricky, though. I haven't had a need for that, and a proxy with local DNS override (as in the article) might needed if it came to that. I'd still do it on the router, though.
> She demanded that I apologize to the women
This is antisocial advice. It's beyond inappropriate to use the pretense of apology to announce your intimate fantasies to strangers.
The main thing I get out of this article is how easy it is to get trapped in a bubble thanks to algorithmic social media.
For the most part, sexy never left, and statistics bear this out. OnlyFans brings in enormous amount of revenue, even after an expensive, failed attempt to be not-just-a-porn-site. Hypersexualized gacha games are pulling in tens of millions of dollars per month, and not just for men; the women-targeted Love and Deepspace had over $50 million in revenue in October. Marvel Rivals, criticized in some circles (such as the social circles of those in the article) for being an oversexualized "gooner game" has remained in the top 10 games played on Steam since its release a year ago. And nothing drives it home more than stumbling across the shady side of YouTube and finding videos in the "woman with large breasts not wearing a bra does something mundane" genre with multiple millions of views.
> I choose these examples from my personal life because they express sentiments that were once the kind of stuff I encountered only in the messy battlegrounds of Twitter, amid discussions about whether Sabrina Carpenter is being oversexualized, whether kinks are akin to a sexual orientation, whether a woman can truly consent in an age-gap relationship, and whether exposure to sex scenes in movies violates viewer consent.
Ultimately, these are the kind of things discussed only by a small, vocal, very online (some might say terminally online) minority. To think that they represent more than a tiny fraction of the world is, again, reflective of how easy it is to get trapped into online echo chambers.
> “Who are you defending yourself against?” To which he answered, to my astonishment: “I don’t know. The world.”
Indeed. Moving our every interaction in daily life plus our innermost thoughts to the internet has instilled a low-key fear in all of us that we'll be raked over the coals and villified as the world's worst villains. The digital tar and feathers are lurking always, a menacing psychological force. And it can even happen without our knowledge; some stranger can post a two second context-less clip or a snippet of a conversation and make us look our worst.
It's shocking how we can have so much outrage over unknown people but we're flush out of rage for the system that makes us so angry all the time.
My advice for building something like this: don't get hung up on a need for vector databases and embedding.
Full text search or even grep/rg are a lot faster and cheaper to work with - no need to maintain a vector database index - and turn out to work really well if you put them in some kind of agentic tool loop.
The big benefit of semantic search was that it could handle fuzzy searching - returning results that mention dogs if someone searches for canines, for example.
Give a good LLM a search tool and it can come up with searches like "dog OR canine" on its own - and refine those queries over multiple rounds of searches.
Plus it means you don't have to solve the chunking problem!
The choice of MIT for a kernel feels like setting up the project to be cannibalized rather than contributed to.
We've seen this movie before with the BSDs. Hardware vendors love permissive licenses because they can fork, add their proprietary HAL/drivers, and ship a closed binary blob without ever upstreaming a single fix.
Linux won specifically because the GPL forced the "greedy" actors to collaborate. In the embedded space, an MIT kernel is just free R&D for a vendor who will lock the bootloader anyway.
Keep this in mind if you _ever_ feel tempted to take A16Z seriously. Absolute charlatans and clowns.
Since the commit history is public, there's a much easier way to tell that AI had a hand in writing that list.
https://github.com/a16z-infra/reading-list/commit/93bc3abb04...
> opus descriptions in cursor, raw
It would have been really great to end the blog post mid-sentence.
Well, the impression of speed is mainly in the head of the headline writers.
What has actually happened is that after about three years of faffing about the Council finally decided on it negotiation position begore the Coreper 2 meeting last week, thought it seems they ran put of time at actual the meeting and had to have the formal approval this week.
The Council is only one of three parties that draft new laws, so now there's are still several rounds of negotiations left.
Nothing substantial has happened to the three texts since last week, it's just that "chat control is back" drives traffic and "Council preparatory body formally approves draft position that got consensus previously but didn't formally get passed because people were fighting over Ukraine stuff for too long" doesn't.
why are specifically the Danish so obsessed with pushing this through? it always seems to come back to them
Even living nearby in the UK it blows my mind how quickly the EU proposes, kills and then revives and passes controversial legislation in such a short timeframe.
About a month ago I had a rather annoying task to perform, and I found an NPM package that handled it. I threw “brew install NPM” or whatever onto the terminal and watched a veritable deluge of dependencies download and install. Then I typed in ‘npm ’ and my hand hovered on the keyboard after the space as I suddenly thought long and hard about where I was on the risk/benefit curve and then I backspaced and typed “brew uninstall npm” instead, and eventually strung together an oldschool unix utilities pipeline with some awk thrown in. Probably the best decision of my life, in retrospect.
I don’t get the negativity.
The specs look impressive. It is always good to have competition.
They announced tapeout in October with planned dev boards next year. Vaporware is when things don’t appear, not when they are on their way (it takes some time for hardware).
It’s also strategically important for Europe to have its own supply. The current and last US administration have both threatened to limit supply of AI chips to European countries, and China would do the same (as they have shown with Nexperia).
And of course you need the software stack with it. They will have thought of that.
https://vsora.com/vsora-announces-tape-out-of-game-changing-...