HackerTrans
TopNewTrendsCommentsPastAskShowJobs

FrancescoMassa

7 karmajoined 10 maanden geleden

Submissions

Claude Savings with context caching awareness

github.com
2 points·by FrancescoMassa·3 dagen geleden·1 comments

Save Claude Code Tokens with Smart Routing

github.com
11 points·by FrancescoMassa·10 dagen geleden·3 comments

Brick: SOTA LLM Routing

arxiv.org
3 points·by FrancescoMassa·28 dagen geleden·0 comments

comments

FrancescoMassa
·3 dagen geleden·discuss
We have a new update for brick : with feedbacks from users we inserted context caching aware routing. so that savings are improved even if switching models transfers context from one kv cache to another. we introduced a tier1 compaction algorythm that makes the router cheaper while mantaining performances. Try it out with brick claude settings > cache aware routing > smartsqueeze
FrancescoMassa
·7 dagen geleden·discuss
[flagged]
FrancescoMassa
·10 dagen geleden·discuss
On our tests subagents & well used workflows are 20-30% more expensive for context & token efficiency
FrancescoMassa
·28 dagen geleden·discuss
[flagged]