Anthropic has been very disciplined and focused (overwhelmingly on coding, fwiw), while OpenAI has been bleeding money trying to be the everything AI company with no real specialty as everyone else beat them in random domains. If I had to qualify OpenAI's primary focus, it has been glazing users and making a generation of malignant narcissists.

But yes, Anthropic has been growing by leaps and bounds and has capacity issues. That's a very healthy position to be in, despite the fact that it yields the inevitable foot-stomping "I'm moving to competitor!" posts constantly.

Honestly at this point I am pretty firmly of the belief that OAI is paying astroturfers to post the "Boy does anyone else think Claude is dumb now and Codex is better?" (always some unreproducible "feel" kind of thing that are to be adopted at face value despite overwhelming evidence that we shouldn't). OAI is kind of in the desperation stage -- see the bizarre acquisitions they've been making, including paying $100M for some fringe podcast almost no one had heard of -- and it would not be remotely unexpected.

Trump recently posted a diatribe about ranked choice voting in Alaska (calling it "disastrous, and very fraudulent").

Do you know why the modern GOP hates ranked choice voting? Because they rely upon getting clown votes wasted on the Tulsi Gabbard, Jill Stein's and Kanye West's of the world as a way to get elected. They just need to entice just enough fool-vote drawers, knowing the cult will not sway an iota.

So now he's a Codex user. OpenAI and Google both have a minimum age of 13.

EDIT: I should note that Anthropic gave him a refund for the whole month that was underway, despite him being nearing the end of it. So good on them.

Can you expand on this? Your teenage son makes more money than you do professionally, by vibe coding video games?

I have zero obligation to detail the work my minor son does to random weird foot-stomping, entitled creeps on HN, and these bizarrely insecure demands by professional failures is...telling.

In this case I assume you're keying off of the other clown who, based upon absolutely nothing (but apparently their own professional failure), is certain it must be "cheating software".

How pathetic. If this is your lot in life, Jesus Christ find a different career. Maybe the trades or something.

Anthropic made the best models by hiring non-technical folks like philosophers to build the best training sets and evaluations. Now, it seems like their philosophers are telling people how they can and can't use their model.

I sense an opportunity for free tokens.

Ideas for prompts that reliably trigger the age check?

It is frequently said that programming directly is obsolete, and the skill you must have now is knowing how to operate agentic AIs.

Yet you aren't allowed to do this until you're 18.

So, developing software is now 18+ only?

Local models are chasing the online frontier models pretty hard.

So worst case, that's the fallback (FWIW, YMMV)

edit: Qwen-3.5 MoE (and other local MoE models like it)

You can quantize it of course, but if the idea is "as close to Sonnet as possible," then while quantized models are objectively more efficient they are sacrificing precision for it.

So next step is to up that speed, so we're at 4x $1300, Nvidia 5090s with 32 GB of VRAM each (128 GB), or $5,200 before RAM/CPU/etc. All of this additional cost to increase your tokens/second without lobotomizing the model. This still may not be enough.

I guess my point is: You see this conversation a LOT online. "Qwen3 can be near Sonnet!" but then when asked how, instead of giving you an answer for the true "near Sonnet" model per benchmarks, they suddenly start talking about a substantially inferior Qwen3 model that is cheap to run at home (e.g. 27B/30B quantized down to Q4/Q5).

The local models absolutely DO exist that are "near Sonnet." The hardware to actually run them is the bottleneck, and it is a HUGE financial/practical bottleneck. If you had a $10K all-in budget, it isn't actually insane for this class of model, and the sky really is the limit (again to reduce quantization and or increase tokens/second).

PS - And electricity costs are non-trivial for 4x 3090s or 4x 5090s.

Qwen3.5-35B-A3B is reported to perform slightly better than the model you mentioned.

It runs fine but non-optimal on a single 3090 with even 131072 tokens of context , and due to the hybrid attention architecture, the memory usage and compute scale rather less drastically than ctx^2. I've had friends with smaller cards still getting work out of it. Generation is at around 20 tokens/sec on that 3090 (without doing anything special yet) . You'll need enough DRAM to hold the bits of the model that don't fit. Nothing to write home about, but genuinely usable in a pinch or for tasks that don't need immediate interactivity.

It's the first local model that passes my personal kimbench usability benchmark at least. Just be aware that it is extremely verbose in thinking mode. Seems to be a qwen thing.

(edit: On rechecking my numbers; I now realize I can possibly optimize this a lot better)

- Qwen is near Sonnet 4.5!

- How do I run that?

- [Starts talking about something inferior that isn't near Sonnet 4.5].

It is this strange bait/switch discussion that happens over and over. Least of all because Sonnet has a 200K context window, and most of these ancdotes aren't for anywhere near that context size.

Only way to know if your own criteria are now matched -or not yet- is to test it for yourself with your own benchmark or what have you.

And it does show a promising direction going forward: usable (to some) local models becoming efficient enough to run on consumer hardware.

[1] released mid-2025

[2] take with salt - only tests personal usability

+ Note that some benchmarks do show Qwen3.5-35B-A3B matching Sonnet 4.5 (released later last year); but I treat those with the same skepticism you do , clearly ;)

That's unsurprising, seeing as inference for agentic coding is extremely context- and token-intensive compared to general chat. Especially if you want it to be fast enough for a real-time response, as opposed to just running coding tasks overnight in a batch and checking the results as they arrive. Maybe we should go back to viewing "coding" as a batch task, where you submit a "job" to be queued for the big iron and wait for the results.

If you ever want to answer this type of question yourself, you can look at the size of the model files. Loading a model usually uses an amount of RAM around the size it occupies on disk, plus a few gigabytes for the context window.

Qwen3.5-122B-A10B is 120GB. Quantized to 4 bits it is ~70GB. You can run a 70GB model in 80GB of VRAM or 128GB of unified normal RAM.

Systems with that capability cost a small number of thousand USD to purchase new.

If you are willing to sacrifice some performance, you can take advantage of the model being a mixture-of-experts and use disk space to get by with less RAM/VRAM, but inference speed will suffer.

Gemma 4 31B Q6: 9tok/s, I'd say it is smarter than GPT-4o, but yeah it's slow. Good for coding.

Gemma 4 26B A4B Q4: 50tok/s. Feels faster than ChatGPT 5.4, but not as smart (as it reasons less). Good for general chatting and research.

Who says this?

Ah, a "seed oil" guy. This is the indication to everyone what level of discourse and scientific fact you're bringing to the table.

In fact, I'll just cite one single bit in your claim-

>Contrary to their purpose, some studies show artificial sweeteners may raise type 2 diabetes risk by up to 38%

This suffers from the same "aha!" nonsense as the garbage submission (and it is garbage). The "study" you are citing was an observational, epidemiological study. And wow, crazy and hard to believe, but it turns out that people with weight problems (and all that comes with that) are more likely to have partaken of "at least one" sugar substitute drink. This is similar to studying people who chew nicotine gum and for obvious reasons finding a higher rate of lung cancer, so then declare that nicotine gum causes lung cancer. It is so staggeringly stupid it belies belief, but it makes for a headline to fool some rubes.

2025: https://www.sciencedirect.com/science/article/pii/S126236362...

2023: https://pmc.ncbi.nlm.nih.gov/articles/PMC10465821/

To address your point: Potential for reverse causality cannot be eliminated; however, many sensitivity analyses were computed to limit this and other potential biases

2023: Clinical review https://pmc.ncbi.nlm.nih.gov/articles/PMC10822749/

multiple studies have shown various side effects associated with the use of these sweeteners. These side effects include gastrointestinal symptoms [9], neurologic [10] and taste perception changes [11], allergic reactions [12], insulin and metabolic effects [13], and cardiovascular effects [14]. In addition, ASs have been shown to affect the gut microbiota that may mediate certain side effects [15]. Most importantly, many researchers have assessed the potential effect of ASs on the cancer risk of people who consume these products [16,17].

and

Human studies performed by Suez et al. evaluated the impact of ASs on the human microbiome. A total of 381 individuals without diabetes who self-reported regular consumption of ASs, as determined by a food frequency questionnaire, were included. The study demonstrated a significant association between the consumption of ASs and the development of central obesity, elevated fasting blood glucose levels, increased hemoglobin A1c levels, impaired glucose tolerance, and elevated alanine aminotransferase levels. In addition, a subgroup analysis was conducted to compare those who consumed higher amounts of ASs with those who did not consume any ASs. The results of this analysis revealed a statistically significant elevation in hemoglobin A1c levels, even after controlling for body mass index.

They outright state this. The first directly says that their higher sugar substitute group had a higher BMI, lower activity, less fibre, and so on.

"To address your point: Potential for reverse causality cannot be eliminated; however, many sensitivity analyses were computed to limit this and other potential biases"

I've actually read the study (given that certain sorts cite it constantly), and do you know what "limited" that bias? Nothing. Literally nothing.

It is a correlation study. People with weight problems are more likely to utilize sugar substitutes. Reversing the causation is the root of an enormous amount of idiot science, though.

And just to be clear, researcher who post this bunk know exactly what they're doing, and usually it is to yield a "more research should be done" conclusion. It's when laymen start building their little notepad.exe listing of everything that supports their nonsense that it becomes a problem.

When you hear about alpha decay of radioactive materials, that is the matter spitting off a highly ionized helium nucleus, freshly birthed into this world. That He nucleus rapidly steals electrons from matter, which is how it can be dangerous to human cells if ingested.

All of that helium underground is the result of alpha decay, and a single uranium-238 element will birth 8 helium atoms as it transitions through a series of metals and one gas (radon), then finally finding stability as Pb206. U235 will birth 7, becoming Pb207.

Anyways, found that fascinating. It's just happenstance that helium often gets blocked exiting the crust by the same sort of structures that block natural gas from escaping, and they are an odd-couple sharing little in common.

One other fun fact -- radon only has a half life of 3.8 days. Uranium becomes thorium becomes radium, then radon where it has an average 3.8 days to seep out of the Earth and into our basements, where it then becomes radioactive metals that attach to dust, get breathed in (or eaten) and present dangers. In the scale of things, crazy. Chemistry is fascinating.

And by stealing those electrons from other molecules it sets off other chemical reactions, which in things like DNA is highly suboptimal. This all generally happens at the birth of the He atom, presuming it isn't in deep space or something with no electrons to cleave from neighbours, and is only an instantaneous state.

“Because they are identical to helium nuclei, they are also sometimes written as He2+…” [1].

[1] https://en.wikipedia.org/wiki/Alpha_particle

H+ is tricky; it’s called hydron (or just H-plus) but it is indeed formed of a single proton, so in practice it is also called that. The names are interchangeable and depend on the field. For example, when looking at irradiation, we call it proton, but either is correct.

If we want to be pedantic, an alpha particle is a He2+ formed during a nuclear reaction, so it typically has quite a bit of kinetic energy as well. In that context they are interchangeable. I am not aware of He2+ being formed in other ways, He is very stable and I am not aware of something that is able to steal its electrons.

  He has risen,
  He has risen,
  He has risen,
  Helium is alive.

Right now the US should get kicked out of NATO. Every American base in Europe should be shuttered. Europe can nuke up -- precisely what America tried to avoid -- and we can enjoy the new nuclear powers of Germany, Poland, Canada, etc. Japan and South Korea might want to build some warheads too.

And hey look, the grifter halfwit wants to increase the military budget $500B and pull out of NATO. Almost like the trillion dollar budget isn't the grand act of benevolence very ill-informed, uneducated, foolish Americans delude themselves into believing.

People become absolutely psychopathic about these sorts of discussions, and become heartless, soulless goblins. Even people at the bottom of the socioeconomic ladder will desperately line up to show how good of little corporate bootlicking subjugates they are.

Apple, just like Microsoft, has a driver signing process because drivers have basically system-wide access to a system. There is no evidence that nvidia has tried to get eGPU drivers signed for years, but now someone did and Apple signed it. So?

And you could always, precisely as the article states in the very first paragraph, disable System Integrity Protection if you want to run drivers that aren't signed.

The API is there. It's straightforward and easy to use. But these users want to piss in the well, tragedy of the commons style.