HN user

wonderfuly

808 karma
Posts71
Comments90
View on HN
news.ycombinator.com 10d ago

Ask HN: How do you review AI code?

wonderfuly
6pts4
developer.microsoft.com 14d ago

Don't rewrite your CLI for agents

wonderfuly
9pts3
github.com 3mo ago

Show HN: A local GitHub-like code review tool to work with agents

wonderfuly
2pts0
github.com 4mo ago

Show HN: Browser extension that takes you to the HN discussion for current page

wonderfuly
1pts0
claude-commits.vercel.app 5mo ago

Show HN: A live feed of commits authored by Claude Code across GitHub

wonderfuly
5pts0
manus.im 9mo ago

Manus 1.5

wonderfuly
2pts0
chromewebstore.google.com 1y ago

Show HN: NotebookLM Web Importer

wonderfuly
4pts0
thishackernewsdoesnotexist.com 1y ago

Show HN: This Hacker News does not exist

wonderfuly
32pts16
manus.im 1y ago

Show HN: ManusAI – the first general AI agent

wonderfuly
2pts2
news.ycombinator.com 1y ago

Ask HN: Best practice for deploying LLM API with streaming

wonderfuly
1pts0
app.chathub.gg 1y ago

Show HN: Free webpage summarizer powered by Gemini 2.0 Flash

wonderfuly
1pts0
ai.google.dev 1y ago

Gemini 2.0 (Experimental)

wonderfuly
3pts0
app.chathub.gg 1y ago

Show HN: AI Models with One Subscription

wonderfuly
1pts0
gist.github.com 1y ago

Show HN: Modified Claude Computer Use Demo to Control macOS

wonderfuly
1pts0
old.reddit.com 1y ago

Show HN: Use GPT-4o vision to add events to your Calendar from screenshots

wonderfuly
2pts1
chromewebstore.google.com 2y ago

Show HN: Dead Simple Twitter/X Search

wonderfuly
2pts0
old.reddit.com 2y ago

Someone copied my open-source product and ranked #2 on Product Hunt

wonderfuly
6pts2
app.chathub.gg 2y ago

Show HN: Chat and compare with multiple AI models

wonderfuly
1pts1
github.com 2y ago

Show HN: conversational AI search with <500 lines of code

wonderfuly
10pts0
chathub.gg 2y ago

Show HN: ChatHub 2.0

wonderfuly
7pts0
www.tryemoji.com 2y ago

Show HN: Turn Emoji into Artwork via AI

wonderfuly
3pts2
github.com 2y ago

Show HN: I Brought Edge Copilot to Chrome

wonderfuly
2pts1
github.com 2y ago

Show HN: Block network access for any extension or website

wonderfuly
1pts0
github.com 2y ago

Show HN: Browser Firewall – block network access for any extension or website

wonderfuly
1pts0
chathub.gg 3y ago

Show HN: Open-Source AI Playground

wonderfuly
47pts7
threadreaderapp.com 3y ago

Reverse engineering the Bing AI sidebar in Edge browser

wonderfuly
1pts0
github.com 3y ago

Bing AI sidebar in Edge browser reverse engineered

wonderfuly
1pts0
news.ycombinator.com 3y ago

Ask HN: What happens to James Halliday ( Substack)?

wonderfuly
1pts1
github.com 3y ago

Show HN: Bing Chat sidebar ported from Edge to Chrome

wonderfuly
18pts11
developer.chrome.com 3y ago

chrome.sidePanel API

wonderfuly
3pts2

Why is it more reliable? The human usage point is fair, but I doubt how long it is still necesary.

Last year, I came across NotebookLM and immediately noticed a pain point: importing the web pages I was browsing into NotebookLM required several steps. So in less than a day, I developed this Chrome extension: NotebookLM Web Importer[1], which allows for one-click importing. As NotebookLM has gained popularity this year, my extension has also seen great growth. So, in July, I added paid premium features to unlock additional features. It exceeded my expectations and quickly went over $500 a month. It now has over 100,000 users and is still growing.

[1] https://notebooklm-web-importer.com

I'm a victim of this.

In addition to concerns about npm, I'm now hesitant to use the GitHub CLI, which stores a highly privileged OAuth token in plain text in the HOME directory. After the attacker accesses it, they can do almost anything on behalf of me, for example, they turned many of my private repos to public.

This article tries to argue that the AI bubble has burst by pointing to the failed release of GPT-5. Admittedly, the release of GPT-5 was somewhat of a flop, but I think it's more of a failure in its launch rather than the model itself. In fact, if you use the GPT-5 Thinking model, it's actually quite good. They attempted to make the model automatically route to different levels of thinking intensity, but the routing didn't work very well, which led to the various bad cases people experienced.

I'm not sure if this is a good idea, just like pretending a network request is a function call, it hides a lot of elements that shouldn't be ignored. I still prefer to clearly explicit embedding, LLM generation, etc.