NeutralForest
Joined 5 years ago
Karma
1,686
Submissions
516
Member since
Jan 4
Recent submissions
Comment · 20 days ago
Damn I got rekt, crazy to read all the regulations and neighbors just dog walking me :'(
Comment · 2 months ago
very cool, thanks for making this!
Comment · 3 months ago
I have a todo list longer than the Tour de France I want to tackle first!
Comment · 3 months ago
Great to see the project is still going strong, I kinda want to try CL but I always feel like I don't have a great use-case.
Comment · 3 months ago
Really interesting follow-up to the "parse dont validate" article and how to understand the trade-off between very specific typing or not.
Comment · 3 months ago
Uh sorry I mistook this for the 3.15 profiler announcement, thanks for the correction!
Comment · 3 months ago
The man himself, thank you for the articles =)
Comment · 3 months ago
Yes indeed! Mostly don't put changing data in the beginning or prepend.
Comment · 3 months ago
Anything Sam Rose is worth reading: https://ngrok.com/blog/prompt-caching
but the implementation will be up to your provider and harness, for deepseek, they expose some numbers: https://api-docs.deepseek.com/guides/kv_cache/ and Anthropic has a list of actions invalidating your cache: https://platform.claude.com/docs/en/build-with-claude/prompt...
Basically, you avoid anything dynamic: model change, tool change, etc it's also important that your system prompt or main prompt doesn't have non-static data like the date/time/place or someone's name (the person you interact with in a chatbot for example). That should be left to tool call or search.
Comment · 3 months ago
Here are explanation on how to use it from one of the people working on it: https://www.youtube.com/watch?v=f1x4X83CDSA and the chunky docs that go with it in the beta of 3.15: https://docs.python.org/3.15/library/profiling.sampling.html
Comment · 3 months ago
Wasting Orval is the biggest sin of all.
Comment · 3 months ago
Interesting article, wonder where we're going with this though, I find it's very difficult to keep LLMs on track and critical enough to be useful.
Just want to say that:
>In our deployment, student-reported reading completion baselines for MATH 010 were approximately 15%, with instructors estimating 10%. Individual student reports of reading compliance ranged from "literally no one does that" to "is this being recorded?"
is hilarious
Comment · 4 months ago
I can only recommend looking at all the Decker author's projects. He's done work on APL languages, GUIs, interactive programming, etc. https://beyondloom.com/things/index.html
A cool person making cool things!