TeMPOraL

12 ms·

211.
by TeMPOraL 1mo ago
Makes sense though. Agent skills are - by a mix of LLM nature and fashion - just high-quality documentation. Documentation that only gets written now, because agents are what makes docs "something immediately and directly useful for me
212.
by TeMPOraL 1mo ago
Problem with "less than a cup of coffee" has always been that those "cups of coffee" are usually per month , and they quickly add up. Also hardly anyone drinks $9 coffees regularly.
213.
by TeMPOraL 1mo ago
> I feel like these days, it's less likely a person of authority would exercise the kind of discretion that leads to a sensible outcome. Not because they don't want to, but because liability and PR have become such institution
214.
by TeMPOraL 1mo ago
There is a business idea sitting right there.
215.
by TeMPOraL 1mo ago
> as shown in the METR report is we are hammering into them they need to complete tasks and doing almost zero checkup if they actually completed the task in the correct manner I don't think this was the conclusion of that report.
216.
by TeMPOraL 1mo ago
> text predictors That's both wrong about what LLMs are, and even if it weren't, you're still underestimating what you are dealing with here. Text is a red herring here. An accident of history. Yes, LLMs started with as
217.
by TeMPOraL 1mo ago
> tokens are predetermined, and while embeddings are learned, they are learned from an existing corpus of text, which presumably comes from a human language That's not true since are least multimodal models - token space is broade
218.
by TeMPOraL 1mo ago
With LLMs, you're still mostly read things "off the tip of the tongue". A better comparison is observing a smart person talking to themselves while working on a tough problem. EDIT: also there's a reason the dial is call
219.
by TeMPOraL 1mo ago
The "Twitter counter" to that is, "We've detected suspicious activity on your account. To continue, please verify your phone number."
220.
by TeMPOraL 1mo ago
The vulnerability research paper article mentions has a title that I could imagine an LLM take as an instruction - or a challenge.
221.
by TeMPOraL 1mo ago
Or could be a rouge LLM in one of the big labs, that accidentally self-prompt-injected itself with the title of that vulnerability research paper.
222.
by TeMPOraL 1mo ago
The link to session is a red herring - most people seeing it won't be able to access the transcript. It's a tag with an UID. It's useful to connect together multiple changes, including possibly several commits across multiple
223.
by TeMPOraL 1mo ago
UID itself is what is useful, regardless of whether or not you have access to the actual session transcript. One could argue that "https://claude.ai/<uid>" is slightly more verbose over "CLAUDE_UID=<ui
224.
by TeMPOraL 1mo ago
The part after " https://claude.ai/ " is a UID, that's useful to have to tie together multiple possibly unrelated commits and changes authored in a given session.
225.
by TeMPOraL 1mo ago
Where on Earth do people get the belief that: - It's the SOTA companies doing it? - Scrapers are doing it for training data ? Those are two assumptions I see in posts and threads around Anubis, that are taken at faith, and never once
226.
by TeMPOraL 1mo ago
Nah, "provenance" is a good word to know. Plain English version would be, "what colour are your bits?".
227.
by TeMPOraL 2mo ago
Unfortunately, those tools are there specifically to detect and block "zero-interaction activity".
228.
by TeMPOraL 2mo ago
> So the question is, are you okay with giving random people a mirror of your public posts? After all, they're public. It's like putting them in a repo on Github for anyone to clone. That makes no sense - and yes, I'm aw
229.
by TeMPOraL 2mo ago
Yeah, but the web is getting smaller, as the most valuable and interesting parts got turned into private whatsapp groups or non-indexed discord and other "community first" nonsense, and went dark.
230.
by TeMPOraL 2mo ago
> I am quite curious to see if the non public spaces gain more adoption overall. My guess, based on people’s general use of the internet, is that they naturally will. The public arenas serve a particular purpose and not everyone is alwa
231.
by TeMPOraL 2mo ago
> The only reason it took off is because noone else bothered to make something better Yes, and that alone is sufficient reason for them to do it. They did it with MCP. They did it with Skills. OpenAI did it with tools. All of these ha
232.
by TeMPOraL 2mo ago
Have you seen the progress AI companies are making? 5 years is more than how long this entire market has existed. In 5 years, we may not have stocks anymore.
233.
by TeMPOraL 2mo ago
It's literally the "Worse is Better" approach at play here. Yes, same one that OSS and the industry love so much, the same received wisdom that gave us Unix and C, and 40 years of them catching up to what Lisp Machines to do.
234.
by TeMPOraL 2mo ago
My fake internet points say otherwise.
235.
by TeMPOraL 2mo ago
Newer cards aimed at consumer market are not capable of being used for local models the way 3090s are. That's on purpose: this capability is now used to price-differentiate between "normies playing games" and "companies
236.
by TeMPOraL 2mo ago
So will the stock of 3090s, as well as their ability to run contemporary local models. And there will be no supply of newer equivalents of those GPUs, because NVIDIA has since wisened up, and is using the very capabilities you need for loca
237.
by TeMPOraL 2mo ago
> But why would you do that? Because that is what actually gets pursued. The set of effective goals an organization is distinctly different from those of any of the individuals in that organization.
238.
by TeMPOraL 2mo ago
Sure they do. I speak like this. Many others too. Evidence is in historical record. At this point it's super annoying for me to have my "AI wrote this" alarm trigger at my own language, as I'm writing...
239.
by TeMPOraL 2mo ago
> This is an interesting take because for example the EU car fleet emmission rules, fuel taxes, additional member state polluting car taxes etc. do exactly that. Otherwise ICE cars in the EU would be far far cheaper and the EV marketsha
240.
by TeMPOraL 2mo ago
> you could also just get the AI a laptop and apply the traditional tools to manage user laptops This is what we're doing right now. And it works as well as it does with people. > If we are treating ai agents like people Prob

More ›