nuwandavek

8 ms·

1.
Satellite images before and after Nepal's devastating floods(artifactbin.dev)
2 points by nuwandavek 6d ago|0 comments
2.
by nuwandavek 10d ago
Yeah, I wonder how they even test interactions between different parts of the prompt (if at all). All the recent effort with skills seems to be the right direction as their complexity can be arbitrarily fractal.
3.
All of Claude's hilarious "Never Do This" in Fable 5.1 System Prompt(artifactbin.dev)
3 points by nuwandavek 10d ago|2 comments
4.
by nuwandavek 11d ago
(Author here) I’ve to complete part 3. Spoiler: HN is actually not getting meaner and is shockingly consistent!
5.
Show HN: Show HN Hall of Fame(artifactbin.dev)
3 points by nuwandavek 12d ago|0 comments
6.
by nuwandavek 23d ago
There is so much scope with maps! All we need is a realtime-ish image model to tidy up the scene. (was looking into it during a recent side project - https://sf.isopolis.city/ ). So exciting!
7.
The OpenAI-HuggingFace Incident Timeline(artifactbin.dev)
2 points by nuwandavek 26d ago|0 comments
8.
by nuwandavek 2mo ago
sorry! i debated this quite a bit, but the sv opening credits was too good to not include :(
9.
by nuwandavek 2mo ago
more zoom enabled!
10.
by nuwandavek 2mo ago
more zoom enabled!
11.
by nuwandavek 2mo ago
yeah, there are a ton of issues wrt water. I've documented a few of them here: https://sf.isopolis.city/dev.html Planning to fix in the next version!
12.
by nuwandavek 2mo ago
Spurious lakes are an interesting issue I looked into a bit. 2 types of cases ended up becoming lakes: 1. "shadows" 2. flat green areas I think a better data distribution should eliminate this. Also, my "water detector"
13.
by nuwandavek 2mo ago
ty!
14.
by nuwandavek 2mo ago
very inspired by you!
15.
by nuwandavek 2mo ago
glad you liked it! I had to leave out a ton of details in the training process. But at every step, codex/cc built me a dev tool to navigate the complexity. That was such a unique side project experience.
16.
by nuwandavek 2mo ago
yeah, there are a ton of issues especially around water. I've documented a bunch of them here: https://sf.isopolis.city/dev.html I tried out a few ideas, but i think better training data around water will help a LOT.
17.
by nuwandavek 2mo ago
oooof, I can actually allow more zoom (was just embarrassed at bad pixels)
18.
by nuwandavek 2mo ago
haha ty!
19.
by nuwandavek 2mo ago
hey thanks! i wrote it all up here: https://sf.isopolis.city/dev.html
20.
by nuwandavek 2mo ago
woahhh, these are so beautiful!
21.
Show HN: Isopolis – Isometric pixel map of SF(sf.isopolis.city)
350 points by nuwandavek 2mo ago|81 comments
22.
Data Story: The strait went dark(demos.minusx.app)
1 points by nuwandavek 3mo ago|0 comments
23.
Convert any PDF into a fillable form(github.com)
3 points by nuwandavek 11mo ago|0 comments
24.
by nuwandavek 1y ago
(blogpost author here) I've noticed this too. My top guess for any such thing would be that this type of sub-agent routing is outside the training distribution. Its possible that this gets better overnight with a model update. The seco
25.
by nuwandavek 1y ago
There may be other reasons to use ai sdk, but I'd highly recommend starting with a simple loop + port most relevant tools from Claude Code before using any framework. Nice, do share a link, would love to check out your agent!
26.
by nuwandavek 1y ago
Oops, fixed it, thanks!
27.
by nuwandavek 1y ago
(blogpost author here) Haha, that's totally fair. I've read a whole bunch of posts comparing CC to other tools, or with a dump of the the architecture. This post was mainly for people who've used CC extensively, know for a fa
28.
by nuwandavek 1y ago
(blogpost author here) I actually found none of them useful. I think MCP is an incomplete idea. Tools and the system prompt cannot be so cleanly separated (at least not yet). Just slapping on tools hurts performance more than it helps. I&#x
29.
by nuwandavek 1y ago
(blogpost author here) You're right! I did make the distinction in an earlier draft, but decided to use "RAG" interchangeably with vector search, as it is popularly known today in code-gen systems. I'd probably go back t
30.
by nuwandavek 1y ago
(author of the blogpost here) Yeah, you can extract a LOT of performance from the basics and don't have to do any complicated setup for ~99% of use cases. Keep the loop simple, have clear tools (it is ok if tools overlap in function).

More ›