frontpage.
newsnewestaskshowjobs

Open Source @Github

fp.

Open in hackernews

Ask HN: What happens when we do compress the context in Claude Code?

3•tbharath•1h ago
What happens when we do compress the context in claude code? How does the reduce the context with out loosing the details?

Comments

verdverm•1h ago
The idea is that you create a summary with sufficient details, or you trim things away, depending on the method, but you do in fact lose exact details.

The impacts of this are not well known and likely depend on model, harness, and session. (hunch)

tbharath•37m ago
So it does forward pass once again to get the summary from the whole session as input?
bubblebeard•1h ago
Not an expert, but, when you give Claude, or any AI agent a query it will usually need to do research, use various tools, execute commands etc. and all of this accumelates text in the conversation thread. Each subsequent query in a thread submits the entire conversation to the API, potentially consuming an unnecessary amount of tokens. Compression attempts to solve this by removing irrelevant information. For example, the entire output from a CURL query and how the AI thought about it might not be needed, having the resulting outcome might be enough. Since compression is also performed by the API though it may be worth it, or not, it always depends on the situation at hand. This is my understanding of it anyways.

Ask HN: What livestream do you keep open in a tab?

4•missmoss•20m ago•0 comments

What happens behind the scenes when we change effort for same LLM models?

7•tbharath•2h ago•4 comments

Timed YouTube video transcripts in .txt and .pdf

2•cristyg0101•1h ago•0 comments

Tell HN: A.I Labs are headed by people without world manufacturing experience

2•dzonga•1h ago•1 comments

Ask HN: What happens when we do compress the context in Claude Code?

3•tbharath•1h ago•3 comments

Who's making money from using/exploiting AI?

5•jonthepirate•3h ago•0 comments

The Cost of Seamlessness

2•taotuner•1h ago•1 comments

Will it be possible to do LLM Pool Training?

2•slicktux•1h ago•1 comments

Ask HN: Why are Meta products so bad?

4•asxndu•2h ago•3 comments

The things that matter take a minute

3•ishener•1h ago•0 comments

Has anyone figured out how to guard LLMs on kernel level?

2•isaacnsisong•6h ago•1 comments

Tell HN: It's a Lot

5•mettamage•7h ago•2 comments

Ask HN: Is neuromorphic computing going to replace traditional AI?

3•lennart-rth•7h ago•1 comments

Codex Is Down

12•freakynit•7h ago•5 comments

Tell HN: Namecheap gave my account to an unverified third party

480•Thrashed•1d ago•174 comments

Ask HN: How would you harden AI changes to a 1M-line legacy SaaS before review?

3•thegreatkahuna•9h ago•8 comments

Postgress in Rust

2•poldex•10h ago•1 comments

Companies are optimizing models for specific benchmarks

2•mzubairtahir•11h ago•0 comments

ChatGPT Is Down

9•anujbans•7h ago•6 comments

Ask HN: What's Up with Archive.today?

5•nobody9999•18h ago•8 comments

GPL Infection of LLMs?

4•sl_convertible•20h ago•2 comments

Ask HN: Is there a website that tracks excessive writes to SSDs in OS/app betas?

30•amichail•2d ago•27 comments

Ask HN: How do I find startups that are looking for a dev? How to approach them?

5•need_a_work23•23h ago•8 comments

Claude used my pipeline to find a counterexample to the Jacobian conjecture

6•JGPTechCo•17h ago•4 comments

Two Gov Websites Compromised

7•relunsec•1d ago•1 comments

Ask HN: If OpenAI hacked HuggingFace, why aren't OpenAI prosecuted?

29•sam_lowry_•2d ago•36 comments

Stripe closed my account with a 0.18% dispute rate

11•rellyattd•17h ago•8 comments

Tell HN: ChatGPT exports do not contain all conversation messages

5•afpx•1d ago•1 comments

Ask HN: How to get in touch with Paddle (MoR)

6•probst•1d ago•5 comments

Ask HN: What are your current favorite monitoring and observability solutions?

3•Exoristos•20h ago•1 comments