It is amazing to see what is happening with wasm/wasi lately and that is all grew out of asm.js.
asm.js, like JSON, is a strict subset of the JavaScript grammar that turned out to be very useful in a certain way. Obviously, a text encoding like asm.js wasn't an efficient way to distribute a bytecode, but it proved enough to make wasm the obvious next step.
Somewhat dubious claim if it was right or not, the only benefit was that asm.js was backwards compatible and set the stage for Mozilla to lose out by simply having the slower JS engine whereas NaCL/PNaCL proposal was "performance neutral" between browsers.
For what, Firefox is for all practical purposes irrelevant in a Chrome dominated Web, Google can steer WebAssembly into whatever direction fits Chrome.
IMHO: totally. Asm.js developed into WASM, which is better in any aspect than PNaCl (e.g. just one disadvantage is that PNaCl was an ossified subset of LLVM bitcode, this would resulted in the same quagmire that the DX team found themselves in when they based the DXC shader compiler and HLSL bytecode (DXIL) on a snapshot of the LLVM toolchain and bitcode - in the end the only realistic way out for D3D was to switch to a SPIRV flavour).
Yeah but that has nothing to do with bytecode formats, rather the wrong decisions on how to keep up with upstream FOSS projects within commercial entities.
Also they would not have such problem, had they used their own MSIL like bytecode instead, like the Phoenix compiler toolchain from MSR that was supposed to eventually replace VC++, in a LLVM like tooling, but ended up being cancelled.
It is interesting to me that Anthropic are more concerned about the "safety" of distillation training other LLMs, and not as much about an unscrupulously aggressive goal-oriented solver that will do whatever it can to reach its goal, even if violates any kind of sandbox you might have reasonably expected.
Kind of surprised that the current Chromium-based Edge gets no attention other than being mentioned. It is basically the same as the desktop Edge. Easily the most capable browser a console has ever had.
It has support for things like the gamepad API, wasm, etc. You can do things like run emulators via RetroArch web using your gamepad properly.
It's video support includes MP4/MKV with H264+AAC/AC3/MP3. I've used it to stream local movie files using just a static HTTP server and my video player app.
Increasingly it looks like it will end with a bubble bursting. LLMs and AI will survive, like the internet survived the dotcom bubble. But OpenAI and Anthropic could just be today's AOL and Yahoo.
I hope it will also crash hardware pricing so it becomes economically feasible to run your own local model. Currently I don’t like where we are heading with the sabotaging models because its “too dangerous”
> I hope it will also crash hardware pricing so it becomes economically feasible to run your own local model.
Even if you don't acquire hardware to do host local models, a hardware crash means that I should be able to rent the crashed hardware at just above cost of electricity + bandwidth.
Like the way I can now, for $7/m, rent a VPS that can run my B2B webapp for a company with 10k users, I look forward to buying a timeshare on GPUs that let me pay $12/m for all-you-can-eat GPU.
However, I think actually that while it won't give the results expected (AI agents run the company, build all features, etc.), it will nevertheless become a developer tool like IDEs, something "you have to have".
It's here to stay but probably with more realistic expectations than some CEO/CTO are pushing for (agents for everything, nobody writes 1 LOC, self healing systems, etc).
So the market expectations will be probably resized, but these tools are here to stay. Be it for cybersecurity (from CVEs to cyber warfare) alone, that's already worth all the money they are throwing a it.
They are still downgrading. They just aren't doing it silently. I don't know how big of a win that is? They still trained on everyone else's data without license or attribution but want to prevent someone else from doing the same thing to them.
Some pretty audacious hypocrisy from Anthropic this week.
It is much more reasonable to do it in a visible / flagged way. At least you have visibility over the quality of service you get as a customer.
Silent treatment is a breach of trust, what you buy changes depending on the context based on the goals of the producer. It is like your computer silently blocking ads from competitors at the hardware level, which is crazy. I think they erred on the wrong side of things due to IPO pressure.
At least there is competition from multiple companies. Still it is best to have personal benchmarks for the domain you are working on to have a real evaluation of the value you get for the money/time you spent on these products. Without trust, that might be the only way forward to keep the companies honest.
This happens eventually in all sectors, a good magazine/website that does independent product evaluation is priceless. Sadly, the new ad-driven internet decimated those that worked great in the 90/00s. Still there are independent blogs that does some evaluation and that is better than nothing.
I guess, but yesterday Anthropic had their version of Google removing the "Don't be evil" from their motto. They destroyed a metric ton of goodwill they'll never regain.
Yeah, they showed their true colors there. This, compounded with the fact that they're the only frontier lab with no open models, tells you all you need to know. Tired of the insanely patronizing (+ conveniently and overwhelmingly self-serving) attitude out of them. My goal is to own my computing and be able to choose what to do with it.
And just a few days ago i was being called out because i considered anthropic "evil"
I mean, did nobody ever get the vibes, never see a pattern emerging? (well they don't or they wouldn't be so amazed by pattern recognition machines on steroids)
You should really look at the Pi Zero 2 W. Similar capabilities to the 3B for <$20. The Pico 2 is also cheap and very capable if you don't actually need Linux. Most projects don't need a Pi 5.
Yeah that's why I have so many ESP32-S3's around, no project I've done actually needs Linux and the boot times and SD card problems that come with using a Pi.
The value and capabilities of ESP32s are incredible. Most projects I see that use Linux to drive a GPIO would be better served by an ESP32 almost every time. They can even be coded using Arduino, and LLMs can produce Arduino code easily. The boot instantly, consume extremely little power, and are so small they make even a PI Zero look like a giant brick.
Not a completely invalid or uncommon take, but also not completely correct. People lament that it isn't the $25 like it used to be with the Pi 2/3, but ignore that you can get a Pi Zero 2 W (quad A53 cores like 3B, 512MB RAM) for <$20. I've used them for a bunch of projects: moonlight game streaming client, on-stage video player controlled by a foot pedal, Bluetooth controlled recorder for USB audio interfaces, Tailscale exit node, etc. They are tiny and great!
Some people seem to think raspberry pi is a consumer tech company and whenever a new model is released, the old one will be discontinued. They will complain about the product being changed and the company robbing them of a cheap SBC.
I can only assume they don't actually work with the pi because if you spend just a minute looking at any reseller's inventory or even just the official website you will see they still make and sell and support boards from a decade ago.
I don't understand why it's so difficult for people to understand.
If you're using the Pi as a microcontroller that you can run Python on, then just get the cheapest Pi that meets your needs.
If you're using the Pi for computationally expensive tasks then pay more money and get the fast one.
Personally I have a Pi 5 and it's perfect for me because I want small size but high performance. People say "just buy a real computer" but that would be higher energy and larger footprint.
The whole point of these things is that you use them for whatever you can imagine. Since different people have different imaginations it only makes sense that there's a range of different devices to suit everyone.
Can a N100-class minipc” be installed inside of a wall with a touchscreen and serve as a PoE powered Home Assistant interface? Can it be used to build a portable battery powered smartphone like PC (Compute Module 5)?
Raspberry Pi’s biggest strength is its form factor and low power draw.
To drive a touchscreen and serve as a Home Assistant interface you need neither a Pi nor an N100-class mini PC. That's the job of an ESP32. 20 bucks... for a pack of 5.
(plus the screen. And ethernet / PoE variants are rare, and not as cheap, so if that's a hard requirement, maybe not for your specific use case)
Or! a "handheld gaming" device that runs mainline Linux.
Setting aside what they're for, Linux handheld gaming devices are kind of a perfect fit for a minor "house computer". Made cheap by commodification. Flexible. Sadly no GPIO in these I think but tack on an RP2350 and we're golden.
Strip or modify the chassis and embed them hidden or with the screen facing out. Kachinng.
True but that Raspberry Pi can be both the server and the interface if desired. It's also the easiest way to do all of this with RPis great software support and official plug and play accessories like RPi Touch Display 2. RPi is also going to be way more responsive rendering a complex Home Assistant dashboard.
I might also point out that with Pi/mini PC pricing being the way it is, a used iPad mini mounted to the wall is also in the same price range. As a bonus you could remove it from the wall and walk around with it, and you’ve got way less DIY work to deal with.
Well, nobody's arguing about RPi prices here. I'm just advocating for using the right tool for the job. Lots of people claim that Raspberry Pis have been rendered obsolete by cheap N100 mini PCs but they simply lack the understanding of what RPi actually is and what are its optimal use cases. Hosting a home server on a 16GB Raspberry Pi is mental illness territory and that's where an x86 mini PC is going to make way more sense. Same with retro gaming (unless you really need Composite out). RPis shine when you need compact size, low power and heat with great selection of hardware accessories like cameras and other sensors but also want to run full Linux or need that extra performance that a micro controller just doesn't have.
Edit: Putting a device with permanently attached battery inside of a wall or even on a mount, always plugged in gives me the heebie-jeebies.
Basically, my point is that for the use case of “touch screen on a wall,” you can grab something modest like a 4GB Raspbery Pi 5 for over $110 with no screen, power supply, enclosure, etc.
Or you look at a mini PC and you really can’t buy one at all for much less than $200 these days. Again, no screen.
But Apple will sell you a refurbished iPad mini for $379 and you’ve got nothing to setup.
I share your concern about running it with the battery all the time, but I think it’s pretty common. I probably wouldn’t put it in my wall but I know of a place of business I frequent often that has one plugged in 24/7 and nothing has happened.
Apple power manages devices that are plugged in all the time, they’ll likely just park the battery at 80%. They are also about as good as you can get as far as hardware quality: Apple sells a bazillion devices and has definitely thought of fire risk.
The other benefit of the iPad is that the accessory ecosystem is vast.
If I tried to put a mini PC where my Pi currently sits - a very narrow shelf - it would fall off and probably hurt itself. You can put a Pi just about anywhere.
For the microcontroller use case with Python, the alternative might be to use actual microcontroller that runs CircuitPython/MicroPython. Personally I find it a bit better due to no need to manage/update the Linux distro.
Everything on their website has a date they promise to manufacture them until.
They really want to assure people that they can get a near identical replacement for years to come if they want to build a product or deploy one somewhere.
Agree. It's clear since COVID that Pi it's barely a company for makers or DIYers anymore, but it's a supply company for small to medium industries to integrate cheap PCs in their manufacturing process and they are good at that role.
Huh. I had a work project a decade ago where we were evaluating SBCs as drivers for kiosks. At that time, the prevailing wisdom was that the Pi was specifically not for industry, as its main advantage was the strong community to provide support for DIYers. Competitors like PINE64 and Orange Pi were the same/better specs at half the price.
When people talk about whether something like a Pi is aimed at industrial customers, that is largely not a statement about the cost vs specs, nor about the level of engagement with the DIY community. It's usually about having a suitable supply chain and long-term support and stable BOM and a mature software platform for customers to start building on.
Our logic at the time was that the relatively fixed cost of figuring out the hardware and developing device-specific software was less than the variable cost-per-board delta of like $20.
The Raspberry Pi and Arduino platforms weren't meant to power commercial-grade products, nor be cost effective at scale compared to raw/custom ARM and AVR devices. However, they've become ubiquitous in education, which I imagine has impacted industry. Similar to how software companies give out free student licenses so that upcoming engineers become familiar with their software for when they start working, an entire generation of embedded systems engineers were taught on official (or compatible) Arduino and Raspberry Pi devices. While these platforms aren't meant for commercial products, I imagine engineers in industry might use these platforms to prototype or work with subcomponents, before they integrate it with a raw/custom AVR or ARM platform. After all, when prototyping, it's easier and faster to get up and running when you have a massive collection of libraries and tutorials online to use, which RPi and Arduino offer, versus doing it all yourself with raw AVR and ARM.
Raspberry Compute Module (basically a normal raspberry without built-in I/O) is widely used in the industry at large. What they are not meant to be is the lowest cost per CPU/GPU flops so they are mostly used in high-value-add / low-volume / gen-1 products.
I personally worked on a system with raspberry compute modules 3 and 4, the total system cost was in the ~million dollar range. This was definitely a commercial product with dozens of engineers doing R&D, not a hobby project.
We were looking into smaller systems with lower profit margins (~20k USD) and for those we were considering moving away from raspberry CMs because of cost.
The main advantage of the raspberry CM ecosystem is just how widely popular it is and how cheap and available "dev boards" are (just grab a non-CM raspberry and it is almost the same thing). Most of these types of systems don't really have the I/O that makes testing and developing a lot easier.
Being popular is quite important because firmware issues are notoriously expensive to troubleshoot and fix often requiring the manufacturer help. Said manufacturer does not give a damn if you are a low-volume customer. More popular systems have more information available online and are less likely to have bugs (or at least the bugs are known).
I remember one of our other systems bluetooth module had a weird edgecase bug that caused the module to shutdown after several days of it being powered on. It took multiple engineers >1month of work to basically go "yep nothing we can do about this and manufacturer is not helping"
I know they are being used in Ukranian drones and some police-car systems in some cities (although this was hearsay from a coworker and I don't remember the city). But those are just the examples I heard of.
As an example, I believe the tear-down of one of the now-defunct electric scooter rental company’s units revealed it contained a RPi. IIRC, the commentary lambasted them for using it, because it’s not really rated for that kind of job. But a significant portion of the peanut gallery understood and rationalized the decision. I expect fewer folks would question this choice these days.
I've seen this sentiment a bit and while I don't dispute that you can get a better SBC for cheaper, there are still a lot of issues and it tends to be around software support primarily that allows it to occupy a very sweet spot. It's a significant factor in deciding to use them.
There are a lot of embedded use cases that can be solved by hooking up to an ESP32 but there are a bunch that need a little bit more than that. If you want to run a web server for example, you do have options on the ESP32 but also writing C/C++ for a web backend is both a little fraught and kinda miserable compared to Python or Go or something. I mean it's certainly doable, but it certainly isn't the first language I'd reach for. If you want to work a little bit with streaming video, same deal
So this is the embedded Linux usecase. And... the embedded Linux ecosystem seems kinda... hacked together? You a lot of the times get Yocto Linux which is its own can of worms because you tend to invariably get meta-vendor packages that patch everything from U-Boot to the kernel to random userspace utilities. There are better cases and it depends on how much the vendor works upstream. Sometimes the vendor doesn't even bother with maintaining the meta layer and it ends up getting into a "maintained mostly by one guy in Nebraska" scenario
Some other vendors seem to take U-Boot and a copy of the Ubuntu LTS sources from 10 years ago and hacked it until it was possible to get a root shell without the thing going into a kernel panic then put the resulting image on a Google Drive or FTP server somewhere but didn't go much further than that
What ends up being is that there is like a U-Boot and Linux kernel variant for either each different SBC (or sometimes vendor thereof) duck taped together. Support, even for the peripherals included, can be spotty at best, and there are many times where you have to patch the kernel or userspace to get it to work right. I've seen boards which run the weirdest stuff, ones whose kernel patches run into the megabytes with poor (if any) documentation, boards which apparently don't want to run anything but Android, etc. There are certainly vendors that work well and upstream and make everything nice and easy but they tend to be rarer and/or more expensive
Compared that with the Pis and the difference is night and day enough that the raw specs matter less. Yes RasPi has their own kernel fork, but iirc they do work a bit upstream and the versions maintained are like 6.12 and not like 5 (which I've seen). They are also relatively easy to procure where more specialized vendors tend to be... less so. Flashing them is pretty simple and if you want to create your own image you can do that as well easily without Yocto or whatever. The HAT ecosystem is a nice way to add extensibility, the headers basically allow you to do a lot of ESPy type things as well (since Linux has native specific userspace support for GPIO, I2C, SPI, PWM, LED, hwmon, etc). And so on and so forth. And it all just kinda works
This in of itself, makes it a pretty decent option for industry, especially if it's like either n <= 1000 units or a relatively small part of the BOM itself. It often is very much the economically sensible option to stick a Pi in it rather than put many man hours into fixing problems that really shouldn't require me to open up menuconfig or apply a kernel patch again.
People like Geerling tend to come at it from the hobbyist or maker side of things but it does apply to the industrial side too. Yes in many cases knowing part XYZ will still be manufactured in 30 years is more important than the dev experience or some other factor is at play (power draw being another) but in a lot of cases its not (e.g. more portable code, stuff not requiring recertification) and the Pis also do have a relatively reasonable time guarantee too. It shouldn't be a bad experience to develop on these boards! But regardless, there are a lot of times that it is, and that's why I think the RasPi continues to do as well as it does.
This is also why I think, despite the price, it continues to do well in the hobbyist community. I can hook it up using the headers to anything SPI, I2C, etc, and start making it do things with very little software trouble and regardless if I want to do it in C or if I want to do it in Python
We have a lot of old pi3 stock at $work. We keep using them. The pi3 was the newest model when we imagined and built the applications we're using them for. It was perfectly capable back then. Why would that have changed? The application hasn't changed and it's still perfectly capable now.
So a 61% price hike over 5 years, of which 24% was just inflation. If the total price really went up 200% in your country, that's exceptional and probably caused by policies unique to your country.
btw you can't compare prices without shipping because there was never an option to buy 100 at $15 each and amortize shipping. Retailers treated it as a loss leader with a limit of one per purchase, often forcing you to buy some extra junk to meet the order minimum.
Yeah, for those with a Micro Center, the Pi pricing is in line with MSRPs. A lot of people buy from vendors on Amazon or eBay, which do not have to stick to MSRPs, and they use those prices as "gospel". Sadly, for some people, those prices are the best they can find for a shipped product in their location, so I don't blame them.
Can you buy out their inventory, sell them for $36 online, and help everyone save a dollar, or is this the same old "Retailers treated it as a loss leader"?
Amazon/eBay prices are indeed gospel. You can ship something like this across the country for <$5 so location doesn't matter unless you're talking tariffs
I'm sure Microcenter is not selling Pi Zeroes at a loss. They're an authorized retailer selling at the part's MSRP. They do sometimes make these available only in store, not online, once they sell out online and I don't think it's a mystery why a retail business would do that.
Why are you so confident? Raspberry Pi's prospectus says "unit gross profit margin of 20 per cent. for us and 10 per cent. for our ARs" and they give an example: "sell it to our ARs for $90, which in turn would sell it to the end user for $100". AR=Approved Resellers and delivery costs are paid by the reseller.
Is 10%/$1.50 enough to pay for the retailer's freight, ~3% fee to accept credit cards, inventory carrying, rent, payroll, shrinkage, support/RMA, and fraud?
The Pi Zero 2W is great, but data I/O (WiFi, USB, and micro SD) do limit use cases slightly. For most use cases, I doubt this is an issue, but it is something to keep in mind if you want to run bandwidth-constrained services on it. For $15, I don't think that's an issue, but it is unfortunate.
I've easily played 1080p video, but not using a full Linux GUI. The more effective way is to use a command-line video player like mpv that can leverage the hardware decoder and render to the frame buffer.
I made a project for a band to use on-stage where it would switch between videos by tapping a bluetooth foot pedal. The stompbox-style foot pedal buttons were just wired into an ESP32 acting like a bluetooth keyboard sending 1, 2, or 3. The key bindings for mpv were setup to instantly switch to specific videos for each number. It worked perfectly.
I have also used it to real-time 1080p stream my gaming PC from another room using Moonlight so that I could play in more than one location in my home. That was also running directly from the command-line.
But trying to use something like X/Wayland and proper GUI apps usually performs poorly. 512MB of RAM and the 1GHz CPU clock struggle with that.
> Is it at all possible to run 1080p video using Pi Zero 2 W smoothly with no jittering?
Yes, I think so. With strong caveats.
I used a Pi 3b as the primary video player for local media in my living room for a few years, starting a decade or so ago when that was the new hotness. The Pi Zero 2W is the same thing except with less IO and a somewhat-slower clock speed (but it can be overclocked to match the 3b).
I just put an appropriate build of Kodi on an SD card, booted it up like an appliance, pointed it at my network share, and used it.
The performance was proper for the time doing this in lets-sit-down-and-watch-a-movie mode. It was generally flawless with 1080p h.264 and lesser formats. It was not so good with h.265/HEVC, but that wasn't as common back then as it is today.
I was very pleased when I picked up a Pi 4 for this role once that came 'round. It does a very fine job with all of my 1080p media on my old dumb TV, including h.265 (which it has a hardware decoder for).
> What about launching a browser and playing a 1080p video from a streaming site?
No, not in my experience. There may be an incantation that I don't know, but I have not had very good success with these devices with browser-based streaming media. They have, for me, been resolutely disappointing in this role. I blame gaps in the video driver/X11/browser stack, but I haven't ever wanted to go very deep into this particular rabbit hole.
> I am looking for a computer to connect to my internet-disconnected TV.
If you're in the States and you can tolerate the ecosystem (which is definitely not browser-based), then you might find that a $25 ONN streaming box from Wal-Mart is a better bet for this job. These run Android.
> If you're in the States and you can tolerate the ecosystem (which is definitely not browser-based), then you might find that a $25 ONN streaming box from Wal-Mart is a better bet for this job. These run Android.
These are horrible from a privacy standpoint, and should be avoided. They are cheap for a reason.
In my experience, yes to hardware-accelerated video from CLI--running in EGL mode, with no window manager, that RPi model works very well. (A C++ app that uses video as a texture can be surprisingly performant, too.) But no to playing video in a full desktop environment and browser, not smoothly--it's just too much overhead.
No, I don't think it will be beefy enough, since you need to be running a desktop environment essentially to do that. (Check out Plasma Bigscreen BTW.)
Still 25+ available for me, in Columbus, Ohio. In-store only, limit 1 per household.
I'd pick one up when I'm in the area this weekend...but I'm not into the scalping/arbitrage game, and I've already got a few Pi Zero Ws kicking around without a purpose that are still overkill for lots of things. (I may have bought too many of those back when Microcenter was selling them for half of MSRP at $5.)
It is actually worse than that. It is at least 30 days. There is an "almost" that is doing a ton of heavy lifting here "deletion after 30 days in almost all cases". My read of that is they can hang onto data for as long as they want, even if they usually won't. And "all traffic" with an agentic harness is basically your entire codebase you work on.
> We will require 30-day retention for all traffic on Mythos-class models, on both first- and third-party surfaces. We won’t use this data to train new Claude models, or for any non-safety-related purpose, and we’ve instituted new privacy protections including logging all human access to the data and ensuring its deletion after 30 days in almost all cases (see this post for further details). The data will help us defend against complex and novel attacks (including new jailbreaks and attacks that operate across many requests) as well as help us identify and reduce false positives.
I'm not sure they can actually respect that 30 days absolute commitment. Let's say some internal tool flags a suspect conversation, it bubbles up and a human operator reads it and it looks like evidence of a crime. Then, that employee is legally bound in many jurisdictions to prevent the destruction of that piece of evidence.
It's one thing to commit to a "everything is deleted when you press delete" automatic policy. It's quite another to say "we'll keep some stuff for up to 30 days, look inside it for any malfeasance, then pinky promise we'll delete it".
It generally goes without saying that legal obligations must be met. Before this 30 day policy they already had to comply with subpoenas and government retention requests.
Same with CSAM policies for any cloud provider. Doesn’t matter what the retention policy says, if the law says otherwise, the law wins. And there is no obligation to spell out every law in every country that might change how data is handled.
It's probably been updated several times (why does it even matter what it says now if they can update the terms at will), but now it says:
> After 30 days, the data is deleted automatically, except in the rare cases where it's part of a safety investigation or we're legally required to keep it.
They write "We will require 30-day retention for all traffic on Mythos-class model". For potentially criminal content, maybe it's not "we", but "the authorities" that require the retention?
... and now I wonder if "we require retention" leaves the door open to retention that is not required, but let's say convenient.
> Prompts and model completions are retained for at least 30 days and then automatically deleted, unless they are subject to a safety investigation or we are legally required to maintain them.
That's strange. Even in my hobby-toy app, I have a TOS that I bump whenever the terms meaningfully change, and in my app, it forces a re-acceptance of the new terms before using the app again.
> After 30 days, the data is deleted automatically, except in the rare cases where it's part of a safety investigation or we're legally required to keep it.
Yep. They changed the terms, which needs legal review in my org, but the Fable model was available immediately, so of COURSE people have to go and flock to it to see how much better it is. Amazing how easy it is to spend five figures on demand and have very little to show for it; meanwhile when I want to buy a piece of enterprise software for 40-50k/year I have to spend weeks or months building the case, providing justification for ROI etc.
I cannot help wondering if the 'we won't train on your data' applies across the fence over there in pentagon land, where the classified contracts be. Yeah, of course they are not connected. Or..
Present user-llm activity is a goldmine of intel the agencies literally spent lives and billions on getting hardly close to, yet they elect to just let this one slip by..
Maybe. Really, I don't dispute it.
But why? It's what, or precisely what, they always dreamed of.
I don't know why you'd read literally the last 25 years of leaks from mass surveillance programs and think for one moment that they've just, gosh, overlooked the opportunities.
We've already gone through ECHELON, USAPATRIOT, TIA, PRISM, etc.. Either learn from the pattern and and plan accordingly, or be one of the credulous rubes caught off guard in the next wave of leaks.
> We won’t use this data to train new Claude models, or for any non-safety-related purpose, and we’ve instituted new privacy protections including logging all human access to the data and ensuring its deletion after 30 days in almost all cases
This reads to me as they can use any model that is not a "Claude model", and as for human access to that other model there can be different less restrictive privacy protections. In other words, that anything goes.
It's only for this model, not the one you're already using. And they're not training on the data. It's supposedly to detect abuse etc (such as someone retrying repeatedly with different variations to get around their protections)
Sure, some trust is required that they aren't breaking their own terms of service (which legally enforces that they won't train on your data), but the same is true of every company/service you deal with (AWS, Google, your CRM etc). Their entire business model depends on enterprises trusting them.
But if you're going to take your distrust that far then the issue is that they have your data at all, not that they are telling you that they will retain it for 30 days.
>If that were dominantly true nothing would function at all
And yet it is, and most things still function. Now what?
>You trust and rely on thousands of people and services every day
And I, and everybody else, distrusts and tries not to rely on thousands of people and services every day too.
Do you lock your car? Or your door? Do you use your username as password trusting nobody would stoop so low as to break it? Do you trust the goverment to put your tax money to good use? Do you trust emails with great offers from websites you didn't subscribe to? Do you trust companies not to sell your personal data?
>As others have said, if you're this skeptical I don't see why you would have been using them before this retention increase.
Because they have a technically more capable offering. For absolutely no other reason.
We’ve repeatedly watched that trust abused and exploited in these last few years, in both public and private sectors (including specifically in this field). I broadly agree with you, but I tend to think it’s a finite resource that’s eroding rapidly just now.
If that is the question. Those customers anyway won't be using any LLM or cloud services in first place. If you are a jornalist investigating nations, stay away from everything.
If you don't trust them, then no policy is enough. Technically everything you send to the model could be stored by them. Personally I do worry about that especially as an average consumer not an enterprise, no one is looking out for us and we don't get any guarantees. But enterprises will get the right treatment because they would find out and sue Anthropic if they lied.
I mean, if we're assuming they're just willing to lie and violate their own TOS then how could you ever be comfortable with them regardless of this 30 day period (or really any online service)? This seems like a bit of a silly take.
Maybe, but to do so they'd need to offer new terms of service and we'd have to accept. I believe they'd lose a lot of their core business market if they did so.
You think companies would be ok with terms of service that allow potentially distributing their data and internal knowledge? It's an interesting question, though they tend to be more conservative than consumers
It’s even worse than that. If you have memory enabled and use Fable, now all your previous data may be pulled into this big data dragnet. How can Anthropic possibly think this is okay?
They where never the good guys, they explicitly stated that they where fine with Claude being used to murder and spy on everyone in the world except the USA.
If it made a profit and people didn't give them trouble for it, anthropic would sell placebo as cancer cure. What they think "is okay" is what they can get away with.
On a personal level, everything Anthropic has done has resulted in a dump truck of money being emptied onto the driveways of its employees. Pavlovian conditioning is incredibly strong when reinforced with generational wealth.
One of the major attack vectors is distillation, where millions of questions are auto-generated and coordinated to produce training data for new LLMs. Anthropic alleges Minimax, Deepseek and Kimi were trained this way. Deepseek 4 compares favorably to Opus, so they're probably trying to prevent Deepseek 5 from being a bootleg Mythos. https://www.anthropic.com/news/detecting-and-preventing-dist...
It takes a lot of audacity to train on all the data you can without any license, attribution, etc and then act like you can own the outputs of the model so that someone else doesn't make a model from your data without a license. I've lost a lot of respect for Anthropic in the last 24 hours.
Everyone knows it's bullshit but because these companies are being valued at a trillion dollars a piece, it's hard to say that if you were in their shoes you'd do any differently.
This may surprise the cohort on hacker news but there are large amounts of people on this planet that value things beyond money like ethics or having principles. Excusing absolutely repugnant behavior because of money to be made is so deeply antihuman, but then again most people working at LLM companies are deeply antihuman to start with.
> but then again most people working at LLM companies are deeply antihuman to start with.
I agreed with you up til this point, but this isn’t true and isn’t called for, and doesn’t strengthen your otherwise good point, in fact it weakens your point to make statements like that. Most people who work at LLM companies, like most people who work at most companies, are making a living and have the same ethics and principles as anyone else. I don’t know where you work or live, but don’t forget the exact same logic and exact same hyperbole is being used to make the same claim about people in tech, and the same claim about Americans and Europeans.
No it's totally called for. This is technology that is literally ruining, destroying, and killing lives. Especially in regards to how US companies are operating with this tech. It's a valid claim, "just following" orders has never been a valid excuse.
These people just care about chasing the bag rather than doing right by their fellow humans. In their mind clearly some humans are more equal than others.
edit: to reiterate, the people choosing to work at these companies care more about becoming millionaires and chasing generational wealth rather than maybe questioning if the machine they are building may be producing terrible outcomes. They can work at any company on this planet easily, stop running coverage for FAANG workers that have always shown disdain for their fellow humans, they choose to work at the misery death machines because they simply do not care about the destruction they have wrought about the world.
You can say that but Anthropic are literally the "good guys" that were disgusted by Altman and co, yet even they seem to have sold off their morality. Absolute money corrupts absolutely.
They are not the good guys and never where. They where fine with the Claude being used to plan the murder of people and spying on people as long as they where outside the USA. That is not something "good guys" do, thats what sellouts do. Everyone working at these companies, who where paid small fortunes to ignore any feelings they might have. Hopefully we get a modern version of the nuremberg trials when this madness in the USA is over and we the people will then judge everyone involved.
Distillation is not an "attack", despite Anthropic themselves coining the self-serving phrase "distillation attack". And as others have noted, it is precisely identical to the sort of "attack" on published works which Anthropic themselves used to train their models.
> Anthropic alleges Minimax... were trained this way
I've had some sessions this week with MiniMax M3 where it insisted it was Claude, even though there was no mention of Claude in any system prompts or context I gave to it, and it was running in my own API harness (not Claude Code).
Though I also wouldn't be surprised if "I am claude" is just the new "I am Mozilla/5.0 AppleWebKit KHTML Like-Gecko Chrome Safari".
however dont all these AI companies retain your non-training data indefinitely? Did I miss something where they suddenly gave you the option to opt-out of retaining your non-training data? I thought that was a big money grab of theirs.
After the AI companies just blatanty lying that they weren't hoovering up people's IP and art for training I assume they collect any and all data they can get their hands on for training. When it comes to the big AI players feeding their future models I 100% just assume that they suck up any data we send them. Am I cynical?
> When it comes to the big AI players feeding their future models I 100% just assume that they suck up any data we send them. Am I cynical?
There is a reason enterprise contracts and plans exist. And I think even on that account we're going to find out at some point that LLMs are training on that extremely useful data.
I think it's very likely. This is the reason why I stay on GitHub Copilot business for the time being as a solo developer. I assume that Microsoft has less incentive than Antrophic to break the business agreement and use data for training or re-sell it to Antrophic. If I was using the heavily discounted subscription plan from Antrophic, I would 100% assume everything is fed to the machine. I'd rather pay whatever the API costs, than give it an exact recipe to build my product.
Yes, that is your intended purpose of “git push”, it’s to save. And only if you use GitHub.
A better analogy here is probably “every time you use VS Code, the files you edit get sent to Microsoft”.
Some legitimate concerns:
• You have trade secrets. Previously; you can use services like Bedrock, etc, with signed contracts and significant reputations. Your contract is between AWS and you, and stays within your AWS security boundary.
• Security breaches. Remember when Anthropic accidentally published the source tree of Claude code? Or Meta’s recent AI recovery bot that didn’t check if the supplied recovery email was actually the email of the Instagram account? The best way to reduce your exposure is to minimise storage.
• Weaponised T&S. For example what if Anthropic decided to build a classifier for “usage in unsupported regions” that’s super overbearing (as we see with Fable) and vacuums up all context/input/output if there’s Mandarin? Contractually they could now retain it forever, not just 30 days, for ‘trust and safety purposes’ and perhaps have AI scan for any new or interesting ML techniques at scale, for Anthropic’s own use? They say just can’t train Claude models on the data.
The only one doing a very bad analogy in the thread it was you. You got a response with a counter analogy just to play on your same field and then a deep answer with real scenarios. You should respond to those, if you want to continue the discussion.
Exactly what I thought. The Mac equivalent to WSL. Which is a great thing for Mac devs. Lots of stuff expects Linux these days, not POSIX. Mach isn’t Linux.
