gwerbin

6 ms·

31.
by gwerbin 1mo ago
They would have to be colluding with any organization that runs a serious benchmark. Which is totally possible! But that would be one hell of a conspiracy theory.
32.
by gwerbin 2mo ago
What's silly about it? Content-addressable decentralized storage -- not tied to a cryptocurrency -- is really valuable for censorship-resistant archiving of politically-sensitive data (such as climate change research).
33.
by gwerbin 2mo ago
I can't speak for anything that came later, but in the early videos the majority of the benefit was that the presentation was exceptionally clear, and focused on developing one little piece of a subject at a time, it gently paste forma
34.
by gwerbin 2mo ago
Claude Code more or less does have the tools to do this: plan mode, todo lists, user question prompts, et al. What it does not have is a "guided" mode where the agent (or harness) interviews you and helps you structure a work plan
35.
by gwerbin 2mo ago
I believe they are genuinely getting better for fully autonomous tasks. Anthropic seems to have gone all in on this, at the expense of more typical usage patterns.
36.
by gwerbin 2mo ago
I found that I need the big model and the high reasoning effort on tasks where I have a large amount of details of varying levels of importance to keep in mind, all affecting different aspects of the project that might be interrelated to va
37.
by gwerbin 2mo ago
> I don't even try to steer it away from that with harness, because it doesn't work and just ends up agonizing over whether it should write some comment. Three paragraphs waffling on that on verbose output If I give it the spec
38.
by gwerbin 2mo ago
That's not the point, the point is that the company making the product is optimizing for the benchmark and/or the apparently idiosyncratic preferences of their own team, and not for the user experience of their paying customers.
39.
by gwerbin 2mo ago
If you watch the thinking traces of just about any modern LLM, you might be surprised at how much "uncertainty" is in there. Weak models with no thinking limits vacillate back-and-forth back-and-forth on a topic for potentially th
40.
by gwerbin 2mo ago
The only people who think LLMs would make good lawyers are the people selling LLMs. The more practical among us recognize that LLMs are our amazing tools for searching through and making sense of large amounts of text with a high level of s
41.
by gwerbin 2mo ago
This is a "don't make me tap the sign" moment. LLMs are next token prediction models. If there are factual errors, confused ideas, etc. in the preceding tokens, that will affect the generation of subsequent tokens, and the er
42.
by gwerbin 2mo ago
> Pitch here is a reference: the note a circle of this area would sound. Tension and density are yours, and so is fade, which is material and air. Where each shape's own fundamental lands above that reference is not yours, and neith
43.
by gwerbin 2mo ago
Government entities in the USA already work like this. That's how all this came about.
44.
by gwerbin 2mo ago
Also because there's no real due process associated with these various lists, they can be used to punish political enemies.
45.
by gwerbin 2mo ago
I'm not saying we won't have breakthroughs or significant capability changes. I'm saying that any reasonable expected rate of breakthroughs is not enough to maintain exponential R&D effort and to have it translate into
46.
by gwerbin 2mo ago
N=2 anecdata but just this week we were discussing setting up a couple of seats with OpenAI as a trial for switching. There are other advantages too, such as being able to bring your own harness including Ai-integrated editors / ACP c
47.
by gwerbin 2mo ago
They want you to use Sonnet to explain what Opus is trying to say. They're not optimizing for token efficiency.
48.
by gwerbin 2mo ago
I think it's a deliberate steganography choice. You can spot Claude vocabulary a mile away, which maybe means you can spot distillations a mile away. But I agree, the GPT models are so much simpler to work with, they have so much less
49.
by gwerbin 2mo ago
Wait, is this the group who predicted that we'd have expontential AI progress and AGI in 2027 because we'd have AIs training AIs? Predicated on the preposterous assumption that R&DEffort == RateOfProgress. The more sober and r
50.
by gwerbin 2mo ago
> The establishment Democrats are not complicit in Trump's nonsense Oh yes they are! They vote to approve his cabinet picks, they campaign harder against left and progressive primary challengers than against the other party, they en
51.
by gwerbin 2mo ago
It's an open question. Will Europe rally behind Denmark in WW3 NATO vs USA, or will we see transatlantic appeasement?
52.
by gwerbin 2mo ago
You can elect a Congress that will stand up to him in 2026. Step 1: continue primarying out establishment candidates from the Democratic Party, because they're largely complicit. Step 2: Vote for said non-establishment candidates in
53.
by gwerbin 2mo ago
Based on Google Translate it looks very much like what the article reports: Greenland Energy is moving exploratory drilling equipment without permits and without permission. Then it provides context on what Greenland Energy is and what they
54.
by gwerbin 2mo ago
And now in 2026 with NSPM-7 anyone who speaks up against literally anything the US does, can get put on a domestic terrorist list, which sets them up to be punished extrajudicially such as being de-banked. And then if the administration get
55.
by gwerbin 2mo ago
One correction: the US is not being run by an autocratic rapist. The US is being run by a cabal of authoritarian/monarchist billionaires, and the aforesaid autocratic rapist is just their representative in the executive branch, support
56.
by gwerbin 2mo ago
No news sites change headlines and edit articles all the time. Often there's a "last updated" date on the article and it can be days or weeks later than the original posting date. The last-updated date is better than nothing,
57.
by gwerbin 2mo ago
While you're waffling on whether it's technically "illegal drilling" or "illegal not-yet-drilling": 1. the US president has previously threatened a war of conquest in Greenland, a territory of a NATO nation 2.
58.
by gwerbin 2mo ago
I assume it's "as in travel". Sounds like the usual Trump ego thing. "It's dangerous, but I know you have to follow me because that's your job to report on what I'm doing, so now you're in danger.&quo
59.
by gwerbin 2mo ago
I like that they didn't try to make it some subliminal superhero tie-in and instead proposed the more plausible mechanism of increasing attention to surroundings in the presence of something unusual.
60.
by gwerbin 2mo ago
A few people are downplaying this as an honest mistake that occurred in the context of necessary testing for guardrails development. That might well be what actually happened! But OpenAI certainly has decided to make a business opportunity

More ›