Y
HN Search
Hacker News Search
new
|
comments
|
top
|
jobs
aray07
searching PlanetScale…
1.
▲
2.
▲
3.
▲
4.
▲
5.
▲
6.
▲
11 ms
·
31.
▲
by
aray07
4mo ago
i have never seen this before - where you offer something and then take that away
32.
▲
Why images use 3x more tokens in Claude Opus 4.7
(claudecodecamp.com)
3 points
by
aray07
5mo ago
|
0 comments
33.
▲
Images cost 3x more in Opus 4.7
(claudecodecamp.com)
2 points
by
aray07
5mo ago
|
0 comments
34.
▲
by
aray07
5mo ago
Came to a similar conclusion after running a bunch of tests on the new tokenizer It was on the higher end of Anthropics range - closer to 30-40% more tokens https://www.claudecodecamp.com/p/i-measured-claude-4-7-s-new..
35.
▲
by
aray07
5mo ago
yeah thats the part that is unclear to me as well - if our usage capacity is now going to run out faster.
36.
▲
by
aray07
5mo ago
im running some experiments on this but based on what i have seen on my own personal data - I dont think this is true "given that Opus 4.7 on Low thinking is strictly better than Opus 4.6 on Medium, etc., etc.” Opus 4.7 in general is m
37.
▲
by
aray07
5mo ago
effort level is separate from tokenization. Tokenization impacts you the same regardless. I find 5 thinking levels to be super confusing - I dont really get why they went from 3 -> 5
38.
▲
by
aray07
5mo ago
i think the new qwen models are supposed to be good based on some the articles that i read
39.
▲
by
aray07
5mo ago
anthropic’s pricing is all based on token usage https://platform.claude.com/docs/en/about-claude/pricing So if you are generating more tokens, you are eating up your usage faster
40.
▲
by
aray07
5mo ago
are you okay with paying more for your services without any perceived improvement in the service itself?
41.
▲
by
aray07
5mo ago
yeah thats is my biggest issue - im okay with paying 20-30% more but what is the ROI? i dont see an equivalent improvement in performance. Anthropic hasnt published any data around what these improvements are - just some vague “better instr
42.
▲
by
aray07
5mo ago
isn’t caveman a joke? why would you use it for real work?
43.
▲
by
aray07
5mo ago
yeah opus 4.7 feels a lot more verbose - i think they changed the system prompt and removed instructions to be terse in its responses
44.
▲
by
aray07
5mo ago
yeah similar for me - it uses a bunch more tokens and I haven’t been able to tell the ROI in terms of better instruction following it seems to hallucinate a bit more (anecdotal)
45.
▲
by
aray07
5mo ago
yeah i am still not clear why there are 5 effort modes now on top of more expensive tokenization
46.
▲
Measuring Claude 4.7's tokenizer costs
(claudecodecamp.com)
714 points
by
aray07
5mo ago
|
498 comments
47.
▲
by
aray07
5mo ago
good point - i analyzed text tokenization. will run some experiments to see how visual tokenization has changed
48.
▲
Claude Opus 4.7's new tokenizer: 1.47x on English, 1.01x on Chinese
(claudecodecamp.com)
4 points
by
aray07
5mo ago
|
2 comments
49.
▲
Verification Is the Next Bottleneck in AI-Assisted Development
(opslane.com)
6 points
by
aray07
6mo ago
|
2 comments
50.
▲
Claude Code's sandbox.denyRead doesn't stop the Read tool
(claudecodecamp.com)
4 points
by
aray07
6mo ago
|
0 comments
51.
▲
I Tried to Reverse Engineer Claude Code's Usage Limits
(claudecodecamp.com)
3 points
by
aray07
6mo ago
|
0 comments
52.
▲
Show HN: I tried to reverse engineer Claude Code usage limits
(github.com)
5 points
by
aray07
6mo ago
|
0 comments
53.
▲
by
aray07
6mo ago
Yeah this is a bit crazy and not surprising at all. The limits have always been opaque and you never know when they change. I started building an open-source local proxy that logs every rate-limit header Claude Code sends. I am using it to
54.
▲
You shouldn't use ultrathink in Claude Code
(claudecodecamp.com)
3 points
by
aray07
6mo ago
|
0 comments
55.
▲
High Agency
(highagency.com)
2 points
by
aray07
6mo ago
|
0 comments
56.
▲
by
aray07
6mo ago
you can choose your own model in claude code and it generally defaults to Opus
57.
▲
Claude Code's System Prompt
(claudecodecamp.com)
3 points
by
aray07
6mo ago
|
0 comments
58.
▲
I Can't Stop Running Claude Code Sessions
(claudecodecamp.com)
4 points
by
aray07
6mo ago
|
0 comments
59.
▲
by
aray07
6mo ago
Yeah - its definitely a new way of working and getting used to!
60.
▲
I Take My Laptop to the Gym So Claude Doesn't Have Downtime
(claudecodecamp.com)
3 points
by
aray07
6mo ago
|
2 comments
More ›