Fight enshittification. For whatever reason, many travel sites no longer send full details in the e-mail confirmation, they want you to click through to the site...which means I can't forward it to plans@tripit.com for automatic import.
Immediately after booking something,I tell Gemini to add it to my TripIt. Works great. I have a little prompt explaining how I like it formatted that I cut and paste, so I can just make this a one-click prompt. I could also have it add flights to my.flightradar24.com.
I also use Gemini in Chrome to add appointment confirmations to my calendar. Or remember things in Google Keep.
There is a cognitive ceiling for what you can do with smaller models. Animals with simpler neural pathways often outperform whatever think they are capable of but there's no substitute for scale. I don't think you'll ever get a 4B or 8B model equivalent to Opus 4.6. Maybe just for coding tasks but certainly not Opus' breadth.
The only thing that we are sure can't be highly compressed is knowledge, because you can only fit so much information in given entropy budget without losing fidelity.
The minimal size limits of reasoning abilities are not clear at all. It could be that you don't need all that many parameters. In which case the door is open for small focused models to converge to parity with larger models in reasoning ability.
If that happens we may end up with people using small local models most of the time, and only calling out to large models when they actually need the extra knowledge.
> and only calling out to large models when they actually need the extra knowledge
When would you want lossy encoding of lots of data bundled together with your reasoning? If it is true that reasoning can be done efficiently with fewer parameters it seems like you would always want it operating normal data searching and retrieval tools to access knowledge rather than risk hallucination.
And re: this discussion of large data centers versus local models, do recall that we already know it's possible to make a pretty darn clever reasoning model that's small and portable and made out of meat.
I guess we can imagine a pure reasoning model (if that's even the right word any more) with almost zero world-knowledge. How does it know what to look for? How does it do any meaningful communication at all?
So I think it's useful to have an imprecise-but-fairly-accurate set of world knowledge as part of an otherwise reasoning-heavy model. It's a cache.
And if the it's an LLM, or something like that, I think it basically has to have world-knowledge built in, because what is natural language if not communication about the world?
I find it difficult to understand the distinction between parametric knowledge and reasoning skills in LLMs. I still think of them as distinct but I understand there is significantly overlap. Arguably, they are the same thing in LLMs. So I would assume that if reasoning is high quality, using RAG could be logical (if much slower). However if the lack of parametric knowledge impacts reasoning, then use of larger models seems warranted. A dumb LLM wouldn't offer sufficient results even with all the RAG in the world.
> we already know it's possible to make a pretty darn clever reasoning model
There's is a problem though: we know that it is possible, but we don't know how to (at least not yet and as far as I am aware). So we know the answer to "what?" question, but we don't know the answer to "how?" question.
I think you underestimate the amount of knowledge needed to deal with the complexities of language in general as opposed to specific applications. We had algorithms to do complex mathematical reasoning before we had LLMs, the drawback being that they require input in restricted formal languages. Removing that restriction is what LLMs brought to the table.
Once the difficult problem of figuring out what the input is supposed to mean was somewhat solved, bolting on reasoning was easy in comparison. It basically fell out with just a bit of prompting, "let's think step by step."
If you want to remove that knowledge to shrink the model, we're back to contorting our input into a restricted language to get the output we want, i.e. programming.
I think you are underestimating the strength a small model can get from tool use. There may be no substitute for scale, but that scale can live outside of the model and be queried using tools.
In the worst case a smaller model could use a tool that involves a bigger model to do something.
I don’t really understand this, but I hear it a lot so I know it’s just confusion on my part.
I’m running little models on a laptop. I have a custom tool service made available to a simple little agent that uses the small models (I’ve used a few). It’s able to search for necessary tool functions and execute them, just fine.
My biggest problem has been the llm choosing not to use tools at all, favoring its ability to guess with training data. And once in a while those guesses are junk.
Is that the problem people refer to when they say that small models have problems with tool use? Or is it something bigger that I wouldn’t have run into yet?
except you don't want knowledge in the model, and most of that "size" comes from "encoded knowledge", i.e. over fitting. The goal should be to only have language handling in the model, and the knowledge in a database you can actually update, analyze etc. It's just really hard to do so.
"world models" (for cars) maybe make sense for self driving, but they are also just a crude workaround to have a physics simulation to push understanding of physics. Through in
difference to most topics, basic, physics tend to not change randomly and it's based on observation of reality, so it probably can work.
Law, health advice, programming stuff etc. on the other hand changes all the time and is all based on what humans wrote about it. Which in some areas (e.g. law or health) is very commonly outdated, wrong or at least incomplete in a dangerous way. And for programming changes all the time.
Having this separation of language processing and knowledge sources is ... hard, language is messy and often interleaves with information.
But this is most likely achievable with smaller models. Actually it might even be easier with a small model. (Through if the necessary knowledge bases are achievable to fit on run on a mac is another topic...)
And this should be the goal of AI companies, as it's the only long term sustainable approach as far as I can tell.
I say should because it may not be, because if they solve it that way and someone manages to clone their success then they lose all their moat for specialized areas as people can create knowledge bases for those areas with know-how OpenAI simple doesn't have access to. (Which would be a preferable outcome as it means actual competition and a potential fair working market.)
TLS cipher X25519MLKEM768 is recommended to be enabled on servers which do support it
last time I checked AI didn't even list it when you asked it for a list of TLS 1.3 ciphers (through it has been widely supported since even before it was fully standardized..)
this isn't surprising as most input sources AI can use for training are outdated and also don't list it
maybe someone of OpenAI will spot this and feet it explicitly into the next training cycle, or people will cover it more and through this it is feed implicitly there
but what about all that many niche but important information with just a handful of outdated stack overflow posts or similar? (which are unlikely to get updated now that everyone uses AI instead..)
The current "lets just train bigger models with more encoded data approach" just doesn't work, it can get you quite far, tho. But then hits a ceiling. And trying to fix it by giving it also additional knowledge "it can ask if it doesn't know" has so far not worked because it reliably doesn't realize it doesn't know if it has enough outdated/incomplete/wrong information encoded in the model. Only by assuring it doesn't have any specialized domain knowledge can you make sure that approach works IMHO.
My GitHub fork of anthropics/claude-code just got taken down with a DMCA notice lol
It did not have a copy of the leaked code...
Anthropic thinking 1) they can unring this bell, and 2) removing forks from people who have contributed (well, what little you can contribute to their repo), is ridiculous.
GitHub's note at the top says: "Note: Because the reported network that contained the allegedly infringing content was larger than one hundred (100) repositories, and the submitter alleged that all or most of the forks were infringing to the same extent as the parent repository, GitHub processed the takedown notice against the entire network of 8.1K repositories, inclusive of the parent repository."
wow, it's also not like their code was actually good (though this apply to most enterprise software). To hide a client behind closed source (it's also typescript, so even more baffling) is laughable behavior.
They constantly love to talk about Claude Code being "100%" being vibe coded...and the US legal system is leaning towards that not being copyrightable.
It could still be a trade secret, but that doesn't fall under a DMCA take down.
IIUC, a person can only claim copyright if they have significantly transformed the output. Unaltered LLM output is not copyrightable per US court decisions.
The whole thing is a legal mess. How do you know the LLM did not reproduce existing code? There is an ongoing legal battle in German between GEMA and OpenAI because ChatGPT reproduced parts of existing song lyrics. A court in Munich has found that this violates German copyright law.
I think you're misunderstanding copyright and ownership.
A copyright over code means that ONLY you can use that code, and nobody else; otherwise, you can sue them. For example, if you are an arist, you want to protect your IP this way.
Yes, AI generated code is not copyrightable but so is most code in general. It is very hard to truly get a copyright for a piece of code. But just because you don't have copyright to something doesn't mean it's not your property.
For example, you can buy several movies on DVD and those DVDs will still be your property even though you don't have copyright and if someone does steal those DVDs, it will be considered theft of your property. Similarly, just because the code is AI-generated/not copyrightable, doesn't mean others can just steal it.
Think about it - so many codebases are not legally protected as copyrighted material but are absolutely protected by IP laws and enforced by the companies that own them.
I think you are fundamentally misunderstanding the concepts of copyright and licensing.
> but so is most code in general.
That's definitely not true. All the code I write has my copyright, unless I waive that right to some other entity. If there was no copyright, there would no licensing. How else could you license your code, if you were not the copyright holder?
Have you never seen "Copyright (c) <Authors> 2025" in source code files?
The very fact that your code has your copyright is also the reason for things like CLAs.
> For example, you can buy several movies on DVD and those DVDs will still be your property even though you don't have copyright
That's because artistic works are distributed under a license. Just like software. Licenses have terms under which circumstances a work can be used, modified and (re)distributed. In the case of DVDs, you are generally not allowed to make your own copies and then sell them. In the case of software, that's why you have the various software licenses (proprietory or open-source).
> Similarly, just because the code is AI-generated/not copyrightable, doesn't mean others can just steal it.
You can't set licensing terms for something that is not copyrightable.
Huh? Normal property law is plainly not applicable to a non-rival good like information (unlike for instance a physical DVD: if someone takes a DVD from me, I don’t have it anymore). “Intellectual property” is, but it is not so much a legal regime as confusing shorthand for a number of distinct ones:
- Trademark law, which applies to markings on copies rather than copies themselves;
- Trade secret law, which stops applying when the information escapes into the wild through the secret-holder’s own actions;
- Patent law, which definitionally only applies to public knowledge as an incentive to not keep it secret instead;
- Publicity rights, which only apply to depictions or discussions of natural persons;
- Moral rights, which are mostly about being recognized as the author and even in their strongest incarnations do not restrict unmodified copies;
- Database right, which isn’t applicable as we’re not talking about a compendium of things, and anyway does not exist in the US and most other places outside the EU;
- Copyright, which you’ve conceded is not applicable here.
There’s no “intellectual property” distinct from these things, and none of them are relevant.
No the human cannot hold the copyright also. They can own the property rights to the code and protect it. It's not like the rule is "AI cannot copyright stuff but humans can" but rather code is rarely copyrighted and in its case, ownership is much more important.
If your code was generated by you and you store it in your system and have property rights over it, you can enforce legal actions even without holding a copyright over the code.
In general, it is kind of weird to want to copyright code. How do you patent a for-loop for example
You can definitely copyright code. I think the English term "copyright" is a bit misleading. In German it is "Urheberrecht" (= author's right), which I think is much clearer.
If you author something, you have the sole copyright. In fact, in Germany you can't even waive your copyright away. However, you can grant licenses for the use of your work.
The difference between copyright and licenses is crucial! By licensing your work, you do not waive your copyright. You still remain the owner. If you publish your code under the GPL and you are the sole author, you can always relicense your code or issue commercial licenses under different terms.
> In general, it is kind of weird to want to copyright code. How do you patent a for-loop for example
There is a fundamental difference between copyright and patents! Patents require a novel technical contribution and they must be granted by a patent office.
This is even worse. My Claude Code instance can theoretically write the same code as your instance for a similar prompt. Why should one of us be able to have the copyright?
Yea this is the thing that makes no sense to me. Any frontier model can unmiminize minified JS pretty decently. Obviously not everything comes through, comments and such, but I always assumed the reason it wasn't open source was to prevent an endless shitstorm of AI slop PR's, not because they were trying to protect secret sauce.
I would look at how podman for Mac manages this; it is more transparent about what's happening and why it needs a VM. It also lets you control more about how the VM is executed.
Everyone close to Anthropic leadership has claimed they’re the real deal and it’s not a stunt. I don’t think it’s bull. They are trying to find a reasonable middle ground and settled on some red lines they won’t cross.
In my case, Mudita Kompakt phone and Boox Go tablet for reading textbooks. Mudita has the stock launcher, Boox is full of bloatware and I installed Niagara
I built an internal CI chat bot with it like 6 months ago when I was learning. It’s deployed and doing what everyone needs it to do.
Claude Code can do most of what it does without needing anything special. I think that’s the future but I hate the vendor lock in Anthropic is pushing with CC.
All my python tools could be skills, and some folks are doing that now but I don’t need to chase after every shiny thing — otherwise I’d never stop rewriting the damn thing.
Especially since there’s no standardizing yet on plugins/skills/commands/hooks yet.
> I hate the vendor lock in Anthropic is pushing with CC.
Accepting any kind of vendor lock in within this space at the moment is an incredibly bad idea. Who knows what will get released next week, let alone the next year. Anthropic might be dead in the water in six months. It's unlikely but not impossible. Expand that to a couple of years and it's not even that unlikely.
Immediately after booking something,I tell Gemini to add it to my TripIt. Works great. I have a little prompt explaining how I like it formatted that I cut and paste, so I can just make this a one-click prompt. I could also have it add flights to my.flightradar24.com.
I also use Gemini in Chrome to add appointment confirmations to my calendar. Or remember things in Google Keep.
There's lot of use cases for this kind of thing.
reply