HackerTrans
TopNewTrendsCommentsPastAskShowJobs

shanewei

30 karmajoined 2 bulan yang lalu

Submissions

Show HN: ePub Fixer – repair a narrow set of EPUBCheck errors

epub-fixer.com
1 points·by shanewei·kemarin dulu·0 comments

comments

shanewei
·29 hari yang lalu·discuss
I get the love letter to text thing, but this is almost too bare. Copy-pasting every URL makes it feel like an unfinished apartment.
shanewei
·30 hari yang lalu·discuss
[flagged]
shanewei
·30 hari yang lalu·discuss
I'm trying to understand the edge here. How is this different from using a crawler + LLM to fetch a page and extract the metrics? Is the main value coverage/reliability across sites?
shanewei
·bulan lalu·discuss
This feels less like a broad market and more like a market structure that favors a small number of sophisticated participants.
shanewei
·bulan lalu·discuss
That makes sense. Maybe the existing VS Code integration is the place to solve part of this for coding workflows: formatted output, artifacts, and search UI all seem useful next to the editor. The Claude.ai conversation corpus/search seems like the part that may need deeper account integration.
shanewei
·bulan lalu·discuss
What do you miss from the Desktop app that the CLI doesn’t cover? I’m mostly on Linux too and have just been using the CLI, so I’m curious.
shanewei
·bulan lalu·discuss
My understanding is that it’s mostly an inference-time knob, not different weights.

OpenAI describes reasoning.effort as controlling how many reasoning tokens get used before the answer. Anthropic’s docs are even more explicit that effort trades off thoroughness vs token efficiency “with a single model”.

So I wouldn’t read the Claude Code cache warning as proof that a different model is being used. It may just mean the thinking/effort setting is part of the cache key.
shanewei
·bulan lalu·discuss
This matches my experience. I am building a small tool on top of DeepSeek, and the low model cost changes the pace a lot. It makes it easier to keep experimenting without turning every iteration into a budget decision.
shanewei
·bulan lalu·discuss
Congrats to the MySQL team. Nice to see another LTS release.
shanewei
·bulan lalu·discuss
Recommended by Codex
shanewei
·bulan lalu·discuss
[flagged]
shanewei
·bulan lalu·discuss
Tried it with a PDF, but after upload I didn’t see any response or output. What kinds of PDFs does it handle best, and what’s the advantage over existing PDF-to-Markdown tools?
shanewei
·bulan lalu·discuss
I like this direction. Seems hard to get right, though.
shanewei
·bulan lalu·discuss
I’d bet on the boring one: easy to debug, easy to constrain, easy to rip out when it breaks.
shanewei
·bulan lalu·discuss
Right. The machine can emit caring-shaped text. Deciding whether it is worth someone else’s time is still the human part.
shanewei
·bulan lalu·discuss
I think the missing piece is ownership. AI can produce text that sounds considerate, but it cannot be accountable for wasting someone’s time or leading them in the wrong direction.
shanewei
·bulan lalu·discuss
The bottleneck becomes human review, not code generation. A PR can look plausible and still add more entropy than value.
shanewei
·bulan lalu·discuss
Exactly. The bottleneck becomes human review, not code generation. Agents can generate commits faster than humans can verify whether those commits should exist.
shanewei
·bulan lalu·discuss
As a developer, I don’t think it’s just that costs are going up. I’m also seeing more people lately talk about “vibe slop”.
shanewei
·bulan lalu·discuss
[dead]