Havoc

4 ms·

1.
by Havoc 5d ago
Yeah especially with their tendency to depreciate stuff they’ve grown tired of
2.
by Havoc 5d ago
Comparing that 17 vs 18 photo id say it’s less a case of detail preserved and more heavy AI post processing. That closeup zoom 18 eye shot has that classic AI smooth skin look AI influencers all have.
3.
by Havoc 5d ago
Yeah same. Got access to their API and then realised I don’t really have an immediate use case
4.
by Havoc 5d ago
Simplicity is a feature too. > you can still do a lot memory management That’s kinda my point - you don’t really want to trust the LLM to get any sort of memory management right. A system where hard constraints are baked into the languag
5.
by Havoc 5d ago
And they hallucinate errors in the millions and struggle with financial data that is in a layout that isn’t in the training data. Ie balance sheet etc. Been trying to add more AI to my workflow but it just doesn’t work (yet) - not in the sa
6.
by Havoc 5d ago
Ah that’s likely why they added a third engine
7.
by Havoc 5d ago
Adtech based companies are evil
8.
by Havoc 6d ago
Pretty sure comfyui has a mac executable
9.
by Havoc 6d ago
Remind me, who do CEOs hire when they have a specific outcome in mind but need it to come from an "independent" source? >We are also in dialogue with METR Another org that has lets call it interesting ties to the big AI shops.
10.
by Havoc 6d ago
Solid mem throughput (for the type of device anyway). Lets hope they up their game of software/OS support...
11.
by Havoc 6d ago
Either would have worked, but rust's fussy compiler & memory features is an advantage for LLMs. The more bug catching you can shift out of runtime and into compile time the better since the LLM can fix it
12.
by Havoc 6d ago
Can they port their m365 chat UI too? Don't know what tech stack it is but I'm guessing electron judging by how buggy and slow it is
13.
by Havoc 6d ago
I'd say step 0 is know your audience. I'm happily vibing my own toy projects, but would prefer if the tech in hospitals is not vibe coded. And I don't think it's plausible that the gap between those two is "well you
14.
by Havoc 6d ago
Wow - I did much worse than I'd have thought. The general lack of humans helps conceal the difference though. Bypasses the uncanny valley problem. I'd contend that most of us would do a bit better on truly random photos Surprised
15.
by Havoc 6d ago
That seems about as probable as translator jobs coming back in 5 years time
16.
by Havoc 6d ago
LLMs of course, but think the real issue here is incentives. Clearly the participants are incentivized to publish garbage. That's going to crater the signal to noise ratio of papers so best that be fixed asap. How though...
17.
by Havoc 6d ago
I like this as a concept - taking a very real world task and checking whether it works. Surprised the outcomes are so poor though. I recall years ago AI was capable of beating pro level DOTA teams. I guess in one case it was specifically tr
18.
by Havoc 7d ago
I just got my invite so the waitlist doesn't seem to be particularly long
19.
by Havoc 7d ago
Sanity prevails
20.
by Havoc 8d ago
I don’t mind the generic dark themed LLM ones even if they all look the same but this particular block looking one is not my fav
21.
by Havoc 8d ago
Surprised they published this without mention of jurisdiction. Each country has their own laws. Would have been prudent to highlight that this is for (presumably) US law
22.
by Havoc 8d ago
While a lot of that is true, there is also a recent sharp uptick in better visibility on the problem. A bit like improving disease detection makes the rates shoot up. In particular the Madlanga commission mentioned in the article has been u
23.
by Havoc 8d ago
Cautiously optimistic. The V1 was noticeably weak on world knowledge but here the 3.8 base model is geared more towards reasoning than world knowledge anyway so might not matter as much
24.
by Havoc 8d ago
Their first 27B bonsai was able to run on an iphone.
25.
by Havoc 9d ago
>77" OLED is impossible to move safely without original box for example) Exactly same here - well curved super ultrawide. Without the curved box styrofoam it can't move. Though in hindsight it was a bad deal...at ~60 bucks a mo
26.
by Havoc 9d ago
>I was gonna ask how people found their coding plans Very good - but I'm on a legacy plan. And coming up on a renewal that would put me on the watered down current plan. But with 50% legacy discount think it may be worthwhile. If I
27.
by Havoc 9d ago
Interesting that the tone of announcements between US and Chinese providers is converging. GLM has in the past been more technical rather than speculation about future development on RSI etc. Also curious whether those 100k accelerators are
28.
by Havoc 10d ago
Saw some measurements on SBC NPUs (3588) and that did seem to have a decent win on power over CPU...but also a perplexity loss relative to CPU so think this will prove quite hard to reliably quantify in practice.
29.
by Havoc 10d ago
So I guess today everyone gets actual work done
30.
by Havoc 10d ago
Can't say I entirely buy it in this case, but interesting rule of thumb anyway. It sure didn't feel like regime change was imminent if it were a little bigger.

More ›