jrflo

6 ms·

91.
by jrflo 2mo ago
That's how language works, it's always evolving...
92.
by jrflo 2mo ago
Other languages use different characters for quotes, if anything that's an indication that's not written by a LLM because it's not favoring the standard English character. https://en.wikipedia.org/wiki/Qu
93.
by jrflo 2mo ago
Hmm, ok. I guess I rarely use the web interface and everything that I have agents record as "memory" in markdown is always accessible locally. If I'm accessing something remotely, I use the ChatGPT app with remote which conne
94.
by jrflo 2mo ago
The desktop app is actually really good (esp on macos). I used to be CLI only, but after trying it recently it's super convenient. Much easier to manage multiple projects simultaneously and it has a built in diff viewer, file browser&#
95.
by jrflo 2mo ago
The conflict of interest is real there, but some benchmarks really ought to be closed source. Otherwise, the second your benchmark is public labs will overfit their new models on it and it will cease to be useful.
96.
by jrflo 2mo ago
Cool idea. Why is this beneficial over just using markdown files and allowing agents to grep for whatever they need? I've tried various MCP things in the past and I've found they tend to slow down the agent and waste tokens more t
97.
by jrflo 2mo ago
There is something to be argued about industry vs academic experience but this post has nothing to do with large industry codebases
98.
by jrflo 2mo ago
I think it's generally fair to assume that you don't become a Stanford CS professor by being bad at writing code and creating abstractions, and that the average professor (let alone one at a prestigious university) is more knowled
99.
by jrflo 2mo ago
I mean, the post was written by a Stanford CS lab, so I'm inclined to believe that they know what they're talking about and are not just bad at creating abstractions.
100.
by jrflo 2mo ago
I'm just saying that any larger model with all of our optimizations of today will always beat a smaller model with the same optimizations. Until the smaller models + optimizations are at AGI levels I don't think anyone will really
101.
by jrflo 2mo ago
Thank you for confirming my suspicions, lol
102.
by jrflo 2mo ago
You're totally right that we're far away from brain-level efficiency, but I'm just saying any efficiency gains we make towards small models will likely be felt on large ones as well, and we'll all move the goalposts to w
103.
by jrflo 2mo ago
Let's not forget the Bitter Lesson. Small models sound really nice but at some point you're just fighting the laws of information theory. Efficiency gains on the small model side are nice, but efficiency gains + giant model tends
104.
by jrflo 2mo ago
To be honest, I wouldn't be surprised if Apple isn't already doing this and just doesn't say so explicitly. There is a crazy amount of image processing going on behind the scenes in each smartphone photo. Would be interesting
105.
by jrflo 2mo ago
I read the original paper they're basing this off of and I think you're right. I do wonder how much of a quality tradeoff there is with perturbing the next token probability distribution. My intuition tells me that a more "pr
106.
by jrflo 2mo ago
The price wasn't that ridiculous IMO for the quality of the discovery, it generated 31M output tokens which is ~$1500 in API cost if it was on Fable. A new lower bound on the biggest unsolved problem in mathematics for less than a coup
107.
by jrflo 2mo ago
I think that false positives are inevitable due to the method of watermarking being embedded in the text itself. The output is intended to mimic human writing, therefore it's entirely conceivable that a human could by chance write text
108.
by jrflo 2mo ago
Yeah you're right, I was thinking of trademarks. I just think the system is very cumbersome and antiquated, these days it mostly serves to benefit patent lawyers rather than inventors and small businesses, aside from highly regulated f
109.
by jrflo 2mo ago
That is why all patents exist. It's ridiculously time consuming and expensive to get a utility patent for anything. I invented something at my old company 5 years ago and the patent process is still ongoing, should hopefully get awarde
110.
by jrflo 2mo ago
Weird, I also have floaters and snowboard but they've never bothered me that much. I can just tune them out and don't notice unless I'm looking for them. Come to think of it, this is probably the first time I've noticed
111.
by jrflo 2mo ago
Really surprising that Inverness and Skye have only 16 pubs considering how much tourism the area has. I felt like it wasn't lacking for pubs but maybe I saw all 16...
112.
by jrflo 2mo ago
Been doing --dangerously-skip-permissions and --yolo for 6 months now, and no nothing bad has happened.
113.
by jrflo 2mo ago
Yeah, they may be over built for sure, I'm just saying the demand is there.
114.
by jrflo 2mo ago
Always insane to me how people will just read the headline of an article and point out "flaws" based on what they assume the article says
115.
by jrflo 2mo ago
I don't know much about this guy other than every time I see him on here he has an axe to grind about AI
116.
by jrflo 2mo ago
But AI use translates directly to time savings (if it works). You only need 1-2 hours of time savings per employee per week to break even on a $100/seat/month subscription.
117.
by jrflo 2mo ago
The thing that's a bit different about AI from a CRM is it translates pretty directly into time savings. You only need each employee to save 1-2 hours of work per month to break even with a $80/seat subscription.
118.
by jrflo 2mo ago
This is the thing I've become increasingly radicalized about. I built an app (link below if you're curious, it's $5 [1]) to cure my particular brand of phone addiction. I've been using my phone for probably ~1hr per day
119.
by jrflo 2mo ago
Might not actually be that much cheaper, we don't know what margin OpenAI is charging on Luna API. Open models likely have much less margin.
120.
by jrflo 2mo ago
I'm using the term "one" loosely, it's a chain of exploits rather than a single weakness, but the argument is the same: it's much harder to find every chain than a single chain.

More ›