Genuine question, why have you chosen to phrase this scraping and distillation as an attack? I'm imagining you're doing it because that's how Anthropic prefers to frame it, but isn't scraping and distillation, with some minor shuffling of semantics, exactly what Anthropic and co did to obtain their own position? And would it be valid to interpret that as an attack as well?
I don't think that learning from textbooks to take an exam and learning from the answers of another student taking the exam are the same.
Joking aside, I also don't believe that maximum access to raw Internet data and its quantity is why some models are doing better than Google. It seems that these SoTA models gain more power from synthetic data and how they discard garbage.
Firehosing Anthropic to exfiltrate their model seems materially different than Anthropic downloading all of the Internet to create the model in the first place to me. But maybe that's just me?
I don't see the material difference in firehosing anthropic vs anthropic firehosing random sites on the internet. As someone who runs a few of those random sites, I've had to take actions that increase my costs (and burn my time) to mitigate a new host of scrapers constantly firing at every available endpoint, even ones specifically marked as off limits.
Yes, what the LLM providers did was worse and impacted people financially a whole lot more in lost compensation for works as well as operational costs that would never reach the heights they did solely because of scrapers on behalf of model providers.
Very cool that these companies can scrape basically all extant human knowledge, utterly disregard IP/copyright/etc, and they cry foul when the tables turn.
We should treat LLM somewhat like patents or drugs. After 5 years or so, the models should become open source. Or at very least the weights. To compensate for the distilling of human knowledge.
All extant human knowledge SO FAR. Remember, by the nature of the beast, the companies will always be operating in hindsight with outdated human knowledge.
An app like Cal.com can be vibe coded in a few evenings with a Chrome MCP server pointed to their website to figure out all the nooks and crannys. The moat of Cal.com is not the code, it's the users who don't want to migrate.
The real answer is they are likely having a hard time converting people to paid plans
Exactly, that's why most Saas companies are in a very tough position.
You have to bring value that goes beyond the source code and hosting, otherwise your clients are going to vibe code a custom solution instead of paying you.
> otherwise your clients are going to vibe code a custom solution instead of paying you.
How many things do you want to be responsible for? How many vibe coded projects do you want to maintain?
I think this line of reasoning is overblown. Just because you can doesn't mean a significant number of people will. I think the 3D printer comparison is apt.
Same story as always, writing the code in the easy part. Requirement gathering, analysis, consensus, direction, those are all the hard parts. Enterprises have a business to run and don’t want to run a software shop on top of everything else.
The story is usually that businesses don't want to commit to indefinitely expending their limited efforts maintaining software which isn't part of the company's core competencies. Most of the cost and effort of software happens after the first release is delivered.
> Enterprises have a business to run and don’t want to run a software shop on top of everything else.
It sounds like you mostly understand here. The biggest part of "running a software shop" they want to avoid is responsibility for support, bugs, fires, ongoing maintenance, and legal issues, of post-release software.
Dave's Pizza around the corner doesn't make a social media app, not because Dave can't figure it out, not because he can't vibe code one, not because he can't contract someone to do it, but because running a social media site isn't a core competency of Dave's Pizza. Instead, Dave uses existing social media sites, and focuses his efforts and passions on making pizza.
So I work in enterprise tech. consulting, my current project is with a large, global, chemicals company (it wouldn't be right to call out my client by name). This client is extremely competent from their multiple enterprise architects down to their analysts, they're a pleasure to work with. One of the business requirements could be met by a very simple in-house developed and hosted API, it's a perfect use case for GenAI assisted coding too. There's no magic, it's a problem solved over and over already. However, they don't want to touch inhouse dev with a 10 foot pole for the reasons we're both talking about. They don't want to support it, extend it, back it up, monitor it, and all the other things that have to happen after the code is done. They're perfectly happy to buy licenses from a saas so if anything goes wrong they can tell the CTO "it's not me, it's them". And when the CTO says "why doesn't it do this too!?!" they can say "i'll call our rep and ask".
saas value to an enterprise is more than just the functionality provided and I think that is lost on a lot of the heads down software devs here.
They are, and always have. Looking over "software engineer" roles in my local area, I see folks at companies in a variety of industries: finance, health, logistics, health care, and the local power utility, all well outside the software industry.
Most enterprise companies don't develop everything in house, but usually do have a varied mix of in-house infrastructure, IaaS and PaaS solutions, and SaaS products. Large organizations across varied industries often have multiple internal dev teams, and the availability of increasingly sophisticated AI tools is going to enable the same teams to be effective at more, and more complex, projects. AI will definitely start shifting make-or-buy decisions, especially for mature, commodity use cases, to 'make'.
I don't think it's much cheaper. Writing some code to do some CRUD has always been easy. Getting to a proof of concept is definitely quicker. But creating something that can be relied upon in production? That's as difficult and time consuming as it has ever been.
Yup. I've explained it as okay, some software is free as in beer and others are free as in speech. DIY software is free as in yacht.
It sounds nice, but now you have something that takes an enormous amount of time and effort to use and maintain, plus you need to have someone with the skills to run it.
They won’t, because specialization is a key aspect of capitalism.
This is why companies outsource anything. Google, Inc. is big enough to own farms and ranches to grow the food eaten in its cafeterias. They could make trucks to transport that food. They could operate factories to make cutlery, etc. Why do they instead choose to pay layers of margins to layers of middlemen?
Absurd example? How about Apple? They outsource production of their chips, instead of capturing the margin they are currently gifting to their partners. Why?
Delta Airlines doesn’t operate oil fields or even refineries even though a major cost of their operations is jet fuel. Why?
Once you can reason through these very simple examples, you will understand why enterprises are unlikely to walk away from SaaS.
s/Delta/United/ or s/Delta/Southwest/ or s/Delta/Lufthansa/. Or if you prefer, s/refinery/oilfield, or s/refinery/pipeline. Or even s/refinery/farm/ because Delta also buys food in vast quantities (I would not be surprised to find they have interests in ag producers that offset a small % of their food purchases, which does not diminish the argument).
Delta also does not make airplanes, jet engines, seats, radios, GPS, glass, or even wires. They don't distill the spirits they serve on their flights. They don't own and operate a satellite Internet capability. They don't even make movies for in-flight entertainment.
The point is that Delta, like most successful firms, outsources key aspects of core service delivery.
The second article you linked says plainly that the refinery is an offset/hedge. QED Delta still outsources the vast majority of its fuel costs. (They could, for example, own large swathes of the Permian and do E&P as well. They choose to leave that to others.)
Vertical integration has been a common practice in industry for 150 years.
Yes, very few firms fully control their upstream supply chains, but very few conversely produce nothing but their core market offering in-house. Most companies are somewhere in between, doing some things in-house, and obtaining other things from vendors.
Most large firms have in-house software dev teams responsible for at least some portion of their development work. I know software engineers locally working, variously, at banks, pet supply distributors, power companies, soft drink bottlers, and many other non-tech industries. And AI can and will extend these teams' capacity to internally manager larger segments of their companies' tech stacks.
> How many vibe coded projects do you want to maintain?
here comes the next SaaS idea - vibe coded services as a service. You tell what service you want, may be point out a couple examples, and you get that service vibe coded and hosted for you for a small monthly fee!
I agree with the other poster that mention this is likely a publicity stunt but all it's really showing is that VC is still incredibly stupid with their money. All the more reason to seize it from them then properly fund useful software and not subsidize vanity projects for stanford grads.
About the friction, not the capabilities...I haven't switched off my biz calendar/appointment provider I'm paying for even though I've kinda outgrown it.
Email is actually a excellent example of something with network dependence. Changing email providers requires that you change your email address too (unless you own and use your own domain). An address change causes friction from having to update the network of contacts and services which used your old email address.
For many use cases, maintenance doesn't matter. At this point, using LLMs to one-shot a tool/service for a single use or time-limited use case is becoming more appealing than signing up with some vendor, even for free.
May be trying creating one and see how much effort and time is required to clone such a functionality to a proper working state! Something for personal use can be created in about 5-10 days, but even then the skill that is required and the amount of tokens to burn, hosting and security etc, will easily kill. This is exactly the thought process of many, but it will surely kill many opensource contributors. I've stopped committing anything to any open source repos as a personal choice. I do not want to train a LLM which will eventually create more slop and headaches since for me, time is the only important factor which holds the maximum value! Nothing else!
At risk of self promotion, I think more people should adopt something like the Ship of Theseus license (https://github.com/tilework-tech/nori-skillsets/pull/465/cha...). It's not obvious if this will patch the clean room hole in licensing, but I'd rather see it play out in court than assume opensource is just fully dead
I am incredibly skeptical that license is legally meaningful. (but obligatory IANAL.)
Generally speaking it is very very difficult to have a license redefine legal terms. Either this theseus copy is legally a derivative work or it isn't, and text of a license is going to do at most very very little to change that.
> It's not obvious if this will patch the clean room hole in licensing, but I'd rather see it play out in court than assume opensource is just fully dead
IANAL, but I don't think there is any "clean room hole in licensing": licensing is downstream of copyright law, and clean-room reverse engineering, if done properly, results in products that do not infringe the copyright of the originating work to begin with, so the license therefore never applies to them.
The "Ship of Theseus" license you've linked to attempts to define for itself what constitutes a derivative work, but what is and is not a derivative work is determined by copyright law itself, and there's no concept of imposing licensing conditions on works that your copyright never extended to in the first place.
Simply put, if something isn't infringing your copyright under the criteria established by the law, then your permission was never needed to do it in the first place, so the conditions under which you would or would not be willing offer that permission are irrelevant.
If someone spends years using your software and they have learned a mental model of how your software works, they can build an exact replica and there is nothing you can do about that since there is no copy you can sue over. Said user is also allowed to use AI tools to aid in the process.
What you want is an EULA, which is a contract users explicitly have to agree with. A license file only grants access or the right to copy, it doesn't affect usage of your software.
"AI slop is rapidly destroying the WWW, most of the content is becoming more and more low-quality and difficult to tell if its true or hallucinated. Pre-AI web content is now more like the golden-standard in terms of correctness, browsing the Internet Archive is much better. This will only cause content to go behind pay-walls, allot of open-source projects will be closed source not only because of the increased work maintainers have to do to not only review but also audit patches for potential AI hallucinations but also because their work is being used to train LLMs and re-licensed to proprietary."
Replace AI with "open source and Linux", and "open source" with "Windows" in the statements. That's what Microsoft's PR team would have said about open source and Linux about 20 years back in the 2000s.
After the unsuccessful FUD era, now Microsoft is running away with Linux by running its Windows alongside via WSL to combat MacOS Unix-like popularity, and due to Linux and open source dominance in the cloud OS demographic.
Even worse, in that Microsoft's FUD was mostly right. The joke about Open Source being communism played out straight - FOSS pretty much destroyed the ability to make money on software products, accelerating transition to SaaS models where you can carefully seek rent from the shelter of your secure company servers (later, cloud), and that is in large part responsible for modern surveillance economy - as it turns out, some SaaS segments decayed to "free with ads", where - much like with OSS and locally-run software - you cannot compete on price with free.
Given how many developers here use LLMs daily, how do you think about defensibility? Tools like this seem relatively easy to reverse-engineer and replicate with enough time and LLM assistance. Did that influence your decision to charge a subscription or the change to a personal license?
That's the reason why I added a subscription in the first place - you would pay a dirt-cheap price for a "boring" product with an added insurance that someone will be there to support it.
People will replicate it, sure, but supporting it regularly is another thing. I guess the majority wanted a perpetual license - so it's a win for the masses.
defensibility nowadays is app support and development. the more work you pour into it the more defensible it will be.
I personally would gladly pay to have app constantly polished and improved. What I would not use is some vibe-coded alternative that was slopped with AI in a day and pushed to github with a tweet "i made a free X alternative" and then abandoned.
Honestly, I have tried to really cut down on my usage of 3rd-party dependencies when possible. In a way, it's kind of freeing. Whatever I still need, I write myself. If I cannot write it, then I try to find something FOSS. If I find nothing, then I consider purchasing something.
For example, I am rolling my own window manager (that needs some much needed TLC). I ditched Alfred for Spotlight. Though Alfred is better, I will survive just fine. And the list goes on.
I am not trying to take a dig at the OP. I am sure he or she put effort into this application. But I am genuinely curious -- does anybody actually need this software? Cmd+Tab, a decent window manager, and Spotlight would solve the same problems for free.
I fresh install to give myself a different perspective when I feel like I have too many 3rd party solutions to problems that no longer exist. Spotlight is better and I only casually use my macbook nowadays, so I don't need the power of Alfred. I don't need dock extensions because Stage Manager is mediocre but works well enough for the browser, chat / music apps, and whatever document I'm working with at the time.
how much is there to improve and polish for a taskbar? at most it will be keeping up with macOS throwing breaking changes at you and maybe one or the other weird bug.
Personally, I dare not replace the Dock with Windows-style task bar for fear that my OLED display might have burn-in on it.
Yet, when I need an alternative, I would rather make an APP for my own.
the notation is just an array of move tuples, each tuple contains 1 move for white and 1 move for black, where each move is written as <1st letter of piece name><destination square>
> In this Friday’s magic demonstration, I’m going to show how what you see in Privacy & Security settings can be misleading, when it tells you that an app doesn’t have access to a protected folder, but it really does.
Yeah, but don't expect this level of fine granularity from government, any government, thats ridiculous expectations. Heck, I wouldn't expect it neither from any private company as a default state.
I wouldn't expect it from any goverment either, but I would expect it from someone who allegedly read the story and is trying to share relevant information.
My first focus is sharing, search and discoverability.
The biggest issue is that I have thousands of photos but I don't have a good way of filtering by person, date, location, and finding exactly what I'm looking for. Basically browsing my photos like I would browse an online store like Zalando. Search for something and refine with filters.
In addition - I implemented S3-like storage. Makes it cheap and easy to host in the cloud.
Promotional pricing that will probably be 9x when promotion ends, and soon to be the only Opus option on github, that's insane
reply