This strategy only works as long as China continues publishing weights. Domestic training capabilities are one way of mitigating that risk.

> that allows us to compare their performance. Opus is considerably better.

In coding? Software architecture? Math? General knowledge?

The diversity of model use-cases is so broad that comments about "model x is better" without any context are largely useless

LLM seems to have become synonymous with Generative Transformer architecture.

While this model may share much with GPT-style models on the encoder side, it clearly has a different decoder architecture. So is a high-parameter count language model an LLM even when it doesn't have a GPT-style decoder? The definitions are in flux.

Has LLM become so synonymous with Generative Transformer that other high-parameter count models that interpret language need a different name?

For all we know this might be a non-language-generative transformer e.g. a transformer where the decoder produces confidence scores rather than language. Please provide more likely architectures if you know them, I'm genuinely curious.

This is likely still an LLM (in the purest definition of a language model with relatively many parameters) since the inputs are natural language, just not a generative LLM as the output is something other than more language.

The inputs aren't natural language. https://docs.typesafe.ai/primitives

The inputs are natural language, they're just also structured into a tree. The first example on that very page shows natural language instructions:

    questions = {
        "refund_requested": Noul(
            instructions="Does the customer request a refund?",
        ),
    }

The hybrid nature of this thing is not a detriment, nor does it make it an LLM.

Public communications from government entities really should be legally required to be disseminated on public forums (ideally properly hosted on .gov domains for easy verification, with RSS feeds) Cross-posting on other platforms should be allowed (with backlink to the .gov), but the .gov should be authoritative.

Unfortunately that requires a level of technical competence that most government entities don't have in the US. Not to mention the ridiculous requirements/processes that are placed on official govt software.

It is 2026, it is not hard to publish a "blog" , if you want to share that same content over social networks that's fine but it should be a requirement that the data doesn't solely live in private walled "social" networks

How is posting on wordpress.com or tumblr.com any different from posting on x.com?

1. It's always publicly readable;

2. It is not reliant on any singular corporation, or trusting any singular corporation.

3. The established domain name system proves authenticity.

And of course the information can still be posted to X, etc.

1 is true: if you control the web server, you control whether the content is published as readable by anonymous visitors.

2 is true: if you disagree, provide the name of that singular corporation. To be clear, you're always relying on corporations, but there's no single gatekeeper.

3 is not the same: if I see something published on "www.water.dcceew.nsw.gov.au" I can be reasonably confident that it is published by my local Government. If I see an X account of @NSWDCCEEW_Water then I need to start looking for context clues, most of which can be faked.

1. You don't control the wordpress.com or tumblr.com web server

2. Wordpress or Tumblr respectively

3. but it's not, it's dcceew.tumblr.com or dcceew.wordpress.com

Assuming this is a goodfaith argument for a second !

You can host a wordpress wherever you like with whichever provider you like or even on a potato in a closet. The format and data are open and can be switched to any platform at any time without an account or login.

RSS !

you can't own wordpress.com though

Technical competence can be acquired as needed. Or more likely the other way around, the government creates an entity that provides this competence to other government entities.

At least in the US most places that sell gift cards require ID when purchased in cash

Halved would be a 50% percent reduction.

0.15/0.22 ≈ 0.68, meaning a roughly 32% reduction on inputs. The 50% reduction is only outputs and cached inputs.

The author misses the difference between cashflow and profit entirely.

Selling a car today gives Tesla the full profits of that hardware production today. Running that car as a robotaxi means that Tesla eats the costs of the hardware today in exchange for a larger total profit collected over several years.

So operating a fleet gives the company access to more long-term profits at the cost of decreasing the bank balance today (negative cash flow), where selling the cars lets the company fill the bank account right now (positive cash flow) at the cost of limiting long-term profitability.

The decision to prioritize immediate cash flow vs long term profits depends on the financial position and overall strategy of the company.

Tesla can do both. There will undoubtedly be a fee to join and use the Tesla ridesharing platform - I would expect that every Cybercab sold to require, at the minimum, the monthly $100 FSD subscription. On top of that, Tesla would take a cut off of all rides that cannot be negotiated.

Any franchisee contract would also need to have some sort of rules that prevent Tesla from running their own taxis themselves and undercutting their business, if it turns out to be too successful.

Tesla is not eating the cost of anything, they want to have their cake and eat it too. this will go as well for them as bunch of other things in recent years but just maybe this time people will realize they have been getting scammed for decade+

The amount of memory required to cache all possible diffs (defined as an ordered pair of commits) would likely be in the exobytes. At current ram prices that's easily a trillion dollars of ram to run that cache lol. Git focuses on making diff calculations efficient largely because the space of possible diffs is very expensive to enumerate.

The following from Claude: """ A diff between two randomly chosen commits usually spans years of history, so it's not a few KB — the tree itself is ~1.5 GB of text, and a multi-year span rewrites a large slice of it. Call it 100–200 MB per pair on average: 8.5×10¹¹ pairs × ~2×10⁸ bytes ≈ 10²⁰ bytes, or ~150 exabytes """

I'm not sure if the website allows for diffs against arbitrary tree states, but if it does than the diff space is completely unbounded, and the ram demand is theoretically infinite.

cache all of them? no... but you could cache some of them, likely the newer are accessed more frequently.

Per the article, the vast majority of traffic is scrapers that are generating effectively random, permuted URLs. Even if they're already caching their popular-with-real-humans most recent data, that efficiency gain is a tiny drop in a very large bucket.

Yes, and it still wouldnt help you since they are trying to fetch everything for each of the 921 forks, as stated in the article (same commits need to be cached again on a forked branch)