mft_

10 ms·

181.
by mft_ 4mo ago
Thanks for sharing your experience. It's really interesting that you describe a loss of some 'middle' parts but not others. The 'classic' medical/psychological model of memory has three parts (sensory/sh
182.
by mft_ 4mo ago
> It was denied because there are keys on the device that would leak if meta opened it up. (I'm not an android dev so I don't know the ins and outs of that) Any idea what changed? > neglect and stupidity from zuck meant that
183.
by mft_ 4mo ago
Yeah, I had my first refusal with 4.8 today. I wanted it to show me how to create an overlay on an existing web game, and it extrapolated that because this could be used to provide tools to help win the game (if that was the direction it wa
184.
by mft_ 4mo ago
I'd argue that the context window is analogous to short-term memory. It's functional but limited in duration, and if you overload it, it starts to fail. It's the long-term memory (i.e. learned experiences feeding back and di
185.
by mft_ 4mo ago
> This seems like a thinly-veiled ad for AI and for this person's company, at the end. There was a very clear (what's the opposite of thinly-veiled?) advertisment for the person's company at the end! But I didn't rea
186.
by mft_ 4mo ago
> Like all generalizations, this is only partly true. There are bad actors. Of course, but in the organisations I have experience of, I'd estimate that the 'bad actor' rate is very low - certainly a low-single-digit percen
187.
by mft_ 4mo ago
Indeed; keep Claude away from your butcher's block cutting board!
188.
by mft_ 4mo ago
> Creating the python extractor at all was sort of bizarre. The argument given was that it is too difficult for a developer to spin up a whole dev environment just to compile a tool, but the C extractor is a single C file that only depen
189.
by mft_ 4mo ago
It wasn't really, though. Creating a Python extractor was a side-quest that supported the end goal of making the original source code available and easily accessible. (You're going to cut yourself off from interesting content mor
190.
by mft_ 4mo ago
In a similar vein, I've argued in many a corporate meeting that there's no such thing as "empowerment". People start out wanting to achieve things, change things to be better, do a good job. The active issue is disempow
191.
by mft_ 4mo ago
Unless... you have a personal or professional need to use apps that don't work on Linux. I tried very hard to switch to Linux full time some months ago, but I couldn't find a way of getting Microsoft Office to work satisfactorily.
192.
by mft_ 4mo ago
Maybe I'm being overly hopeful, but given the DGX Spark only runs Linux, and this is apparently a sibling in a laptop form factor, maybe it won't be too difficult to get Linux up and running effectively? Probably (a lot) more eas
193.
by mft_ 4mo ago
Fair enough :)
194.
by mft_ 4mo ago
> I think you severely underestimate the power of Copilot. It's the absolute worst thing for windows. I agree with you, but I think you've overestimating the level of technical sophistication of the typical Neo buyer. > (I&#
195.
by mft_ 4mo ago
llama-bench is part of the llama-cpp package, but from recent experimentation, the settings it is able to (or is documented to?) accept lag behind somewhat. Not sure whether it would accept all of the esoteric settings in the article?
196.
by mft_ 4mo ago
No offense intended. I guess someone in the RevMed Scientific Comms team had a more stressful day than they'd envisaged. > Perhaps the abstract of yesterday corresponded with a preliminary version of the study report, but meanwhile
197.
by mft_ 4mo ago
Probably the bottom end an individual would want to consider using due to slow generation time . Sure, you could theoretically take a model compressed in this manner and deploy it on an old netbook and run the calculations on the CPU, but
198.
by mft_ 4mo ago
Your numbers are a little off. The NEJM article was published a few hours ago: https://www.nejm.org/doi/full/10.1056/NEJMoa2605555 RAS G12 population mOS: daraxonrasib 13.2 months / chemotherapy 6.6 mon
199.
by mft_ 4mo ago
Yes, with nuances. The recently-read-out trial is in the 'second line' - meaning patients will have received one 'line' of treatment before this - typically chemotherapy +/- surgery. The chemotherapy regimes used f
200.
by mft_ 4mo ago
Yeah, it's pretty incredible. And I guess that's mostly what's behind the question: whether this is more of an impressive research/technique demonstrator, or a real product advancement solving a need.
201.
by mft_ 4mo ago
Genuine question: is this solving a real problem? IME, the bottleneck when using diffusion models isn't storage space or memory, it's generation time. Lots of models will run on 8-12 GB 1080-generation GPUs onwards, or on Macs wi
202.
by mft_ 4mo ago
I disagree. I suspect the vast majority of Neo sales are simply driven by the ability to get Apple-quality laptop hardware for such a low price. As such, the people driving the Neo sales are competing manufacturers who offer cheap nasty pla
203.
by mft_ 4mo ago
From the perspective of this intermediate/hobbyist level Python amateur who never quite got to grips with Swift, it's a paradigm shift. For all manner of simple-ish personal apps, Claude can one-shot an MVP in a few minutes, pull
204.
by mft_ 4mo ago
Brings back memories. Comanche was incredible when it came out, running on (IIRC) our family’s 386SX-16. I tried to replicate the effect in Visual Basic, albeit with very limited success at the time.
205.
by mft_ 4mo ago
I've spent cumulatively much more time comparing SVGs of pelicans, and we know who's responsible for that...
206.
by mft_ 4mo ago
Eh, I was expecting something far worse from the title. Once a month, an email reminds you to click on a provided link, log in (via saved credentials, one assumes?) and click a single button? I get that it's small frustration, but I s
207.
by mft_ 5mo ago
Agree. It's incredible that for pancreatic cancer (which is one of the single-most lethal and difficult-to-treat cancers there is) we're moving from a choice between a couple of decade-old and brutal chemotherapy cocktails, to di
208.
by mft_ 5mo ago
A very nicely written article (and I don’t say that often!) And the overall premise is spot on: while it’s a shame that the drugs failed, it’s okay, because we want companies to be taking bets on targets that might result in the next big dr
209.
by mft_ 5mo ago
To be fair, while I love the Hyundai you linked to, it’s mostly going to appeal to people raised on a diet of 80/90s cars, with a particular focus on Japanese exports and Initial D. Not Ferrari’s typical market ;)
210.
by mft_ 5mo ago
What is gobsmacking is the price. I know it’s Ferrari , but one of the interesting things about EVs is that there’s minimal technological differentiation marketed to customers after a certain point. As in: a buyer wouldn’t know or care abo

More ›