adchurch

4 ms·

31.
by adchurch 3mo ago
But we're not routing via `claude -p`, if you have sub usage available + it's the right choice to route to a Claude model, then the router is approximately a transparent passthrough. So it gets billed like normal `claude` usage ra
32.
by adchurch 3mo ago
We integrate with OpenCode too! OpenCode provides the harness, then the router selects the right model for the task. We haven't yet set up local model routing though, that's really interesting - have you had any success using loca
33.
by adchurch 3mo ago
100%, from what we've seen, for a lot of big companies that 1. don't have subsidized usage and 2. are pushing AI adoption hard, figuring out token costs is P0 or P1 for their eng leadership
34.
by adchurch 3mo ago
Yep it uses the Claude sub if possible and falls back to API billing only if you don't have a Claude sub or it's out of usage! Same deal for Codex
35.
by adchurch 3mo ago
I'd say that a typical main agent loop has 1-3 models (obviously very situationally dependent), but when you have subagents those can get routed independently since they have a fresh context window, so there are a lot more degrees of f
36.
by adchurch 3mo ago
Yeah that's a really interesting point, tbh I think the more relevant variable here is the harness you're using rather than the specific model? i.e. GPT 5.5 in the Claude harness behaves a lot more like Claude than Codex if that
37.
by adchurch 3mo ago
If you have a Claude/Codex subscription then we use that (and account for the subsidized price accordingly when making routing decisions) instead of API billing. So you get the best of both worlds: subsidized usage for frontier models
38.
by adchurch 3mo ago
Oh cool, feel free to reach out to me at andrew@workweave.ai if you ever want to share notes! We've learned a lot in the process of building this so far :)
39.
by adchurch 3mo ago
If you statelessly route each new request: yes it does end up being more expensive! So our routing is cache-aware. It will have a much higher threshold to switch from one model to another if there's already some cache for the first mod
40.
by adchurch 3mo ago
Fun fact: Cursor's "auto" mode is just Composer (or at least it was last time I checked). So it's different in the sense that it actually does route to more than 1 model
41.
by adchurch 3mo ago
It saves money because some agent sessions can be entirely handled by a smaller model (also relevant: subagents use fresh context windows so a subagent with a simple task can be routed to a smaller model even if the main agent needs a front
42.
by adchurch 3mo ago
You're right and that's why we built the router to be cache aware! Once it starts using one model, the threshold to switch to another model will be higher because the additional cost of the cache miss needs to be worth the cost sa
43.
Show HN: Smart model routing directly in Claude, Codex and Cursor(github.com)
216 points by adchurch 3mo ago|113 comments
44.
Show HN: Optimal model routing directly in Claude, Codex and Cursor(github.com)
4 points by adchurch 3mo ago|0 comments
45.
Weave (YC W25) is hiring ML, AI, product, & design engineers(jobs.ashbyhq.com)
1 points by adchurch 4mo ago
46.
Weave (YC W25) Is Hiring ML, Design, and Product Engineers(jobs.ashbyhq.com)
1 points by adchurch 7mo ago
47.
Weave (YC W25) is hiring a founding ML engineer(ycombinator.com)
1 points by adchurch 11mo ago
48.
Weave (YC W25) is hiring a founding AI engineer(ycombinator.com)
1 points by adchurch 1y ago
49.
by adchurch 1y ago
Author here, great to see all the conversation & thoughts you all have shared so far! One thing I've seen a lot of people saying: hiring juniors isn't worth it because they'll just leave for more money in a couple years.
50.
by adchurch 1y ago
If you keep asking questions eventually you hit something they don't actually know. How long it takes to get there + what they say when they get there is what makes this interesting imo!
51.
by adchurch 1y ago
Great idea re: giving hard problems. Same motivation behind why we ask people about past projects and keep diving deeper and deeper. The point is to figure out if they're curious & capable of engaging on a deeper level, vs. just fo
52.
by adchurch 1y ago
Check out the bit about the hiring process, particularly about filtering for the right mindset.
53.
Weave (YC W25) is hiring a founding AI engineer(ycombinator.com)
1 points by adchurch 1y ago
54.
Weave (YC W25) is hiring a founding AI engineer(ycombinator.com)
1 points by adchurch 1y ago
55.
Weave (YC W25) is hiring an AI engineer(ycombinator.com)
1 points by adchurch 1y ago
56.
Weave (YC W25) is hiring a founding AI engineer(ycombinator.com)
1 points by adchurch 1y ago
57.
Weave (YC W25) is hiring a founding engineer(ycombinator.com)
1 points by adchurch 1y ago
58.
by adchurch 1y ago
now that's a beautiful api!
59.
Weave (YC W25) is hiring a founding engineer(ycombinator.com)
1 points by adchurch 1y ago
60.
Weave (YC W25) is hiring a founding engineer(ycombinator.com)
1 points by adchurch 1y ago

More ›