Viewing profile — wolttam

As written it sounds like you're saying that it was incompetent of the labs to disable the sandbox internet access? They tried to disable open internet access but the models zero-d…

Automated defence is going to use so many tokens.

Yours is the only mention of GPT-OSS in the whole thread. I count about 2-3 comments saying the model is meh and many more saying it’s a big step up. I am among those with real lif…

I think we will eventually reach a point where this is the case, but at the moment it seems like you can throw virtually any non-trivial use-case at a model today and end up being …

It feels like the rate of stupid we have achieved in the last (checks watch) ~18 months has surely eclipsed anything that may have occurred during the roman empire. By sheer volume…

It's a terrible moat. You etch the silicon then nobody wants to run it in 6 months because models have advanced that much further.

We’re already seeing incredible knowledge compression out of the models the GP mentioned like Qwen 35B-A3B, which feels well within the realm of “runs on a phone” in the next handf…

Essentially, MCP is the harness’ wheelhouse, not the model’s

MCP is the protocol by which the harness can connect with additional tools. The harness does the work of discovering the tools, the model is then fed a description of what tools it…

MCP is invisible to the LLM, or should be, under normal use

The problem is that there is no signal to the folks involved whether you have bothered to understand what you’re posting or are simply offloading that work to others. Same issue as…

2 sparks currently run this model at 60 t/s single session, up to just over 100 t/s aggregate with concurrency of 4. Going local has as opened up a world of use-cases I never would…

Hooray! This model makes me very optimistic about the future of local inference. The CyberGym score stands out to me.

It runs really well on 2 DGX Sparks - 60t/s

> and used tiktoken to approximate tokens, by dividing character count by four. This is just the type of thing that stands out. You used a tokenizer library to determine the number…

Not always. An email has a valid SPF when its return path email’s domain permits the sending server’s IP. But that email may have a forged From: header (which causes an SPF mis-ali…

> smaller model that doesn't have any domain knowledge or facts built into its weights I’m not sure it works this way. Language modelling itself is a “domain”, and if it didn’t hav…

Well, one thing it will still have is a boat-load of compute.

Each one of those employees maps to several fold times more spending on compute.

No way this can go wrong for the U.S. economy! Do everything you can to protect the existing incumbents, pay no mind to the wider consequences. Winning strategy.