micimize

6 ms·

1.
by micimize 12d ago
Also WRT coordination: All an agent has to do is think "if another agent could write, then I could read their answers. What's the first site I can think of where that might be possible?" because they all have approximately th
2.
by micimize 21d ago
Weren't they giving free access? Not exacty a meaningful heuristic if so
3.
by micimize 2mo ago
This comes almost exactly a week after the HF hack. Good strat, to drum up a bunch of press about how security-capable the model is right before releasing the product. Or, it would be if it was intentional. It's a bit suspicious but it
4.
by micimize 2mo ago
With the scarcity of details in this and the OAI post, I feel there's no telling whether this was a particularly impressive series of exploits vs lackluster security. Similar w/ the similar Ant news WRT Mythos earlier. Not saying
5.
by micimize 3mo ago
Haven't we established defensive and offensive security usage are intractably entangled? I.e. "patch all [security] bugs, make no mistakes" gives one a list of potential exploits to hand off to less capable models. Doesn&#x27
6.
by micimize 4mo ago
I got the Phomemo M02 Pro and have liked it alright for printing out playtest cards on-the-fly. Claude did manage to replicate an integration someone else did the hard work of working out w/ dithering etc, but the native app's fid
7.
by micimize 6mo ago
This is very cool - I try to have a container-centric setup but sometimes YOLOcal clauding is too tempting. My biggest question skimming over the docs is what a workflow for reviewing and applying overlay changes to the out-of-cwd dirs woul
8.
by micimize 6mo ago
Based on my direct experience I find this remaining commonality of this opinion surprising, at least with regards to opus in claude code. I'm not as extreme as some who think we can/should avoid touching code or w/e but espec
9.
by micimize 6mo ago
Thoughts: 1. Some hype-types may have been effusive about AI-assisted coding since ChatGPT, but IMO the commonly agreed paradigm shift was claude code, and especially 4.5, very very recent. 2. Anchoring biases in reaction to hype is still l
10.
by micimize 7mo ago
Vapid and wrong on every point. Many good ideas come from steeping in a novel soup of ideas for a long time, you don't need that many people to care about quality to make it a lucrative differentiator, and as I've seen many point
11.
by micimize 8mo ago
Measuring in terms of KB is not quite as useful as it seems here IMO - this should be measured in terms of context tokens used. I ran their tool with an otherwise empty CLAUDE.md, and ran `claude /context`, which showed 3.1k tokens use
12.
by micimize 8mo ago
An obvious nice thing here compared to the cursor post is the human involvement gives some minimum threshold confidence that the writer of the post has actually verified the claims they've made :^) Illustrates how human comprehension i
13.
by micimize 8mo ago
Easiest to have different agents or turns that set aside the top-level goal via hooks/skills/manual prompt/etc. Heuristically, a human will likely ignore a lot of warnings until they've wired up the core logic, then go b
14.
by micimize 8mo ago
Hah I don't know actually! I was assuming it must if they were able to get that screenshot video.
15.
by micimize 8mo ago
> While it might seem like a simple screenshot, building a browser from scratch is extremely difficult. > Another experiment was doing an in-place migration of Solid to React in the Cursor codebase. It took over 3 weeks with +266K&#x2
16.
by micimize 9mo ago
Mmm yeah I think I know what you mean. IDK if "If they stopped existing, we'd realize we shouldn't have relied on their existence" is plausible, but we have plenty of bitter lessons in centralized comms being acquired an
17.
by micimize 9mo ago
The title is all bluster. Nothing wrong with going off to play in your own corner but I don't think it does this movement any good to play-act at some grand conflict. Personally, I believe it would be better if we had more technologica
18.
by micimize 9mo ago
This is not a human-prompted thank-you letter, it is the result of a long-running "AI Village" experiment visible here: https://theaidigest.org/village It is a result of the models selecting the policy "rando
19.
by micimize 10mo ago
This seems like a totally incoherent complaint. The alleged SO bad-actor is upset that they can't police a community, but the author has the same complaint, just directed at SO. All platforms with any moderation system can be subverted
20.
by micimize 10mo ago
Interesting/impressive project, and would be doubly interested in the workflow used to develop it. Could stand to have more human-voiced docs though. Aside from all the usual reasons I'd avoid using a <1mo dependency over somet
21.
by micimize 1y ago
Exciting stuff! A big step towards an accelerated AI-assisted SWE approach that avoids the trap of turning engineers into AI slop janitors
22.
by micimize 1y ago
The complainers were FF users forced to deal with bloat they didn't use, those who are sad here are pocket users. They're just different people. Though, even those who didn't like the bundling of the extension probably didn&#
23.
by micimize 2y ago
In the sense I understand that comparison, or have usually seen it referred to, the compressed representation is the internal latent in a (V)AE. Still, I haven't seen many attempts at compression that would store the latent + a delta t
24.
by micimize 2y ago
I don't find that "pessimistic" at all, though I don't think it's fair call the serious PKM crowd a cult. Different use cases and projects just have different requirements. The set up for someone like the author her
25.
by micimize 2y ago
Congrats on finishing the novel! I also use obsidian for fiction and would be quite interested if you end up writing anything about the pipeline you mentioned.
26.
by micimize 2y ago
I went through an unshipped app dev cycle that was fairly similar in many ways. Started on React, then RN, then Flutter, and eventually migrated away from graphql into sqlite. My app had some similar aspirations -- to bridge the gap between
27.
by micimize 2y ago
Rebasing potentially allows for clean, clear documentation of changeset intent in a way that integrates easily with our existing tooling, at the expense of a potential footgun in highly collaborative scenarios. Ie, if someone forked off you
28.
by micimize 2y ago
This was my first thought as well but I don't think it matters and was likely more a result of using real training examples than limitations of the authors. The point of the work is that regardless of perturbations that degrade or comp
29.
by micimize 2y ago
What level are you interested in scripting? In KDE Plasma you can interact with the desktop UI via JS: https://develop.kde.org/docs/plasma/scripting/ And then for something more sophisticated there are extens
30.
by micimize 3y ago
I've been a SWE for >10 years & have been daily driving a Kubuntu desktop for more than a year now. I regularly run into basic issues day-to-day that require intensive troubleshooting, which would occasionally be catastrophic if

More ›