It sure is suspicious that both Anthropic (adaptive thinking) and OpenAI (Avoid generic brevity instructions) both seem to be suggesting that the best way to improve outcomes is to entirely leave it to them to decide how many tokens get used.
I mean, it's true that it would be ideal of this stuff did just get figured out optimally behind the API, but there is definitely an incentive on their side to burn more tokens.
The backfiring is baked in, if you think about it: This program is going to die in just a few days, while Sony is going to continue selling physical games for years.
Something I haven't been able to figure out.... How are you supposed to actually get an API key to use quota from your subscription? The terms of service still forbid using OAuth authentication and the API keys from the console indicate that you need to pre-load your account with funds when you try to use them.
While the gist of what you say is true, it is hard to get very good at treating them as instruments when they keep getting replaced with new, ostensibly-better versions every few months. But those new versions are not strictly better. They are mostly-better while actually having different strengths and weaknesses.
It's hard to decide when to use the best tool for a job you are aware of to ensure throughput and when to spend time experimenting with a new tool to learn what it's good at.
I believe this is an attempt to try out a possible answer to the problem of "If AI makes it non-viable for individual companies to pay junior devs, there will be no junior devs". The posited theory: Maybe the AI companies pay them off what is effectively a tax on the industry as a whole (that they can extract because every company has no choice but to pay their fees). It's pretty dystopian so I hope this isn't the future, but... maybe worth trying as an experiment?
So, the author gave the model single sentence prompts like "Balatro, but for the game of coin flips" plus generic encouragement like "make it better" three times ended up with netlify-hostable web games each time? That is hard for me to believe.
It's likely that at least some amount of additional context was provided to the model to enable it to reliably create the desired form factor. This introduces the caveat that the author probably views some amount of context as being trivial / beneath the level of mentioning. But then the question becomes where they draw the line.
In the age of AI, perhaps it will be more common to see major forks of open-source projects emerge, with the upstream back-porting a few of the larger features themselves because they're impressed at what the fork is able to do, as opposed to the fork making PR's back into the upstream.
In the second half of last year, I found that agentic coding with proprietary models (≈ vibe coding) reached the point where it actually speeds up my ability to deliver useful code at work. Before that, AI-based autocomplete definitely helped, but (despite the claims of the people selling AI coding tools) letting an agent author more than a file or so at a time (often a function or so at a time) required a very intricate plan or it would create a mess. Creating that plan or cleaning up the mess would take longer than just doing everything myself.
For me, it feels like widely available open models have recently crossed that same canyon. Are they as good as e.g. late-model Claude Opus? I don't think so. But they have absolutely gotten past the point where they are beneficial. This means that, for me, they are about six months behind.
| Technically, once a model is initialised, that's it. That is a model. If released, that would be, even for the most pedantic absolutists, undoubtably open source.
That is true. But it is not the same model as the LLM created by combining the released weights with the released architecture. The thing that is the "binary blob" is the weights. It is pretty much exactly akin to a Linux driver that depends on linux-firmware. It is wonderful that it exists! But it is only partly open.
| Now, what licensing does, and the only thing that licensing can do is to give you rights to inspect, modify and release that model. That's it. A license will never give you (it cannot) the right to have the internal IP, knowledge, know-how or the "why's" on how the model was edited. That's on you. You have the right to modify, but you can't get the right to know how others have modified it, from a license file. Never had, never will.
| In practice, we do have fully open (open data, open training code, open source models) models. Apertus, from Switzerland and Olmo from the US. Don't get me wrong, it's absolutely great that we have these models, they are very important for the community, and they do help inform everyone about what works, what doesn't, and so on.
You seem to contradict yourself here. That said: I appreciate the correction of my perception that there aren't truly open large language models.
I know it comes off as pedantic to point this out but: Those are open weight models not open source models.
Closed weight models are the equivalent of SaaS. Open weight models are the equivalent of binary driver blobs or Windows software. We don't really have actual open source LLMs, which would need to publicly release their training data and technique so you could train a similar model yourself, or use their work as a baseline for your own model.
This distinction matters because an actual open source LLM would be extremely important from an ecosystem point of view, if someone ever actually released one.
I believe they're referring to the fact that if almost all of your code is written by junior developers without mentorship, you will end up wasting a lot of your development budget because your codebase is a mess.
It depends. On its own, UBI puts a downward pressure on the value of money. Some other things (e.g. setting low interest rates) also put a downward pressure on the value of money. However, some things (e.g. taxes) put an upward pressure on the value of money. So it comes down to how all of those factors balance out.
Yep. It's the difference between "Don't do these things, regardless of what the law says." and "Do whatever you want, but please follow your own laws while you do it".
As Paul Graham said, "Sam gets what he wants" and "He’s good at convincing people of things. He’s good at getting people to do what he wants." and "So if the only way Sam could succeed in life was by [something] succeeding, then [that thing] would succeed"
It is in the specific case that you don't have biometric or PIN login set up on the device and you use a password manager that doesn't require authentication. In that case, the only factor is "something you have". Otherwise, it is still a multi-factor authentication because the device itself still represents "something you have", and your device unlock represents "something you know" or "something you are".
I mean, it's true that it would be ideal of this stuff did just get figured out optimally behind the API, but there is definitely an incentive on their side to burn more tokens.