dchu17

6 ms·

1.
by dchu17 5mo ago
Sorry, I may be misunderstanding the question. The way this works is that it stores workstreams and session state in a local SQLite DB, and links each ctx session to the exact local Claude Code and/or Codex raw session log it came from
2.
by dchu17 5mo ago
that's interesting, I hadn't at this point but this sounds potentially useful
3.
Show HN: Ctx – a /resume that works across Claude Code and Codex(github.com)
72 points by dchu17 5mo ago|28 comments
4.
by dchu17 7mo ago
Gave this a run and it was pretty intuitive. Good work!
5.
by dchu17 7mo ago
Yes this is a major problem I thought about. The makeshift solution here was to redact the “identifying information” on the press release. Even then, I benchmarked that GPT-5 could still match it back to the right TIKR around 53% of the tim
6.
by dchu17 7mo ago
Yes we know of a few. Honestly, it was pretty hard to even find a good catalyst calendar for this space. I'll give it a read to learn more. Thanks for the note!
7.
by dchu17 7mo ago
That's interesting. I am curious, what kind of analyses did you work with on the molecule and drug itself? Was it like mostly reading papers/patents or did your team do anything experimental?
8.
by dchu17 7mo ago
Our initial goal with this project actually wasn't trying to get an edge in terms of better evaluating information, but rather, we wanted to see if an LLM can perform similarly to a human analyst at a lower latency. The latency for the
9.
Show HN: BioTradingArena – Benchmark for LLMs to predict biotech stock movements(biotradingarena.com)
35 points by dchu17 7mo ago|15 comments
10.
by dchu17 10mo ago
Hi, thanks for the comment! Just wanted to respond to some of comments here: >> First, your business model isn't really clear, as what you've described so far sounds more like a research project than a go-to-market premise.
11.
by dchu17 10mo ago
Yeah I think one of the things that would be interesting is to see how well it generalizes across tasks. It seems like the existence of pathology foundation models means there is certainly a degree of generalizability (at least across tissu
12.
by dchu17 10mo ago
Thought about this too. I think there are two broad LLM capabilities here that are kind of currently tangled up in this eval: 1. Can an LLM navigate a slide effectively (i.e find all relevant regions of interest)? 2. Given a region of inter
13.
by dchu17 10mo ago
I've been thinking a bit more about better ways to build the tooling around it, I don't know much about video compression to be fully transparent but will read up on it. I have been running into some problems with memory managemen
14.
by dchu17 10mo ago
Nope I haven't, I can take a look and see if I can fit it in
15.
by dchu17 10mo ago
I think so. It feels like there is more to be squeezed from just better prompts but was going to play around with fine-tuning Qwen3
16.
by dchu17 10mo ago
Nope, I just did some prompt engineering on ootb models. I thought about doing some fine-tuning on like Qwen but think that there is still more performance to be squeezed out with just prompts here.
17.
by dchu17 10mo ago
> What's the performance of the models trained specifically on this task, and random guessing, compared to the expert pathologist? I should probably first clarify here, the disease classification tasks are about subtyping the type o
18.
Show HN: Cancer diagnosis makes for an interesting RL environment for LLMs
46 points by dchu17 10mo ago|20 comments
19.
Show HN: Postgres Platform for Global Replication and Sharding(fortress.build)
2 points by dchu17 2y ago|0 comments
20.
by dchu17 2y ago
Yep! Just felt it wasn't for us. We originally built it for e-commerce as just a better way to get product data than through affiliate APIs but felt like most of the market pull weren't in things we were interested in.
21.
by dchu17 2y ago
We really like Nile and it is definitely a company we look towards for inspiration! While we hope to share similar DXs, our fundamental difference is that we are focused on a BYOC-first platform instead of a serverless Postgres platform. We
22.
by dchu17 2y ago
Thanks for the criticism! We still definitely need to work on our language to best communicate this; we'll work on keeping it more concise and straightforward to best highlight what we offer.
23.
by dchu17 2y ago
This is a great question :) We initially targeted startups at the moment that they are moving from a 3rd party DBaaS to databases on their own cloud. However, we just realized this was super tough to time. We experimented a bit with enterpr
24.
by dchu17 2y ago
This looks super interesting! Will check it out. Thanks for letting us know!
25.
by dchu17 2y ago
Thanks for the feedback! We've seen most SaaS companies use some sort of tenant_id column and this is definitely the most popular method that developers currently use. We want to provide a few things for SaaS developers. For one, many
26.
by dchu17 2y ago
Thanks for letting us know, fixing it now!
27.
Launch HN: Fortress (YC S24) – Database platform for multi-tenant SaaS
121 points by dchu17 2y ago|82 comments