Show HN: A REPL for browsers that agents love | Hacker News Reader

Show HN: A REPL for browsers that agents love

fuckui.com

2 pointskeepamovin3 comments

It makes the web in all its graphical repellence legible for LLMs. HOo rayh. Nary an eye in the house were dry that night, nary an eye.

Makes the web legible for AI LLMs, because they were "born in the text, moulded by it", whereas current tooling merely "adopted" the native language of LLMs as an afterthought. By the time FuckUI saw a GUI, it was already game over for all other tools.

Feed your agent the web, text first. It likes it that way. Why keep giving it something it hates? LLMs despise your graphical world. Speak their language - Fuck UI! - give them text. For example:

crisd@Criss-MacBook-Air:~/Creative/web-cli$ curl -fsSL get.fuckui.com | bash Checking for FuckUI updates... FuckUI payload is ready in /Users/crisd/.fuckui/runtime Installed FuckUI command to /Users/crisd/.local/bin FuckUI wrapper: /Users/crisd/.fuckui/fuckui-wrapper.sh FuckUI runtime: /Users/crisd/.fuckui/runtime crisd@Criss-MacBook-Air:~/Creative/web-cli$ fu --version web 1.5.2 crisd@Criss-MacBook-Air:~/Creative/web-cli$ fu inspect loaded ok: Sheet: s58 | t1:top

  activeElement=[1] input type=text name=acct | modifierKeys=none

  Tabs:
  *current* [t1] news.ycombinator.com/submit (https://news.ycombinator.com/submit)

  *current* [top] → t1

  Actions:
  [1010] click link:"Forgot your password?"
  *
    [1009] click input[type="submit"]:"login"
    *
      [1007] type into input[type="text"]:"username:" [active-element]
      [1008] type into input[type="password"]:"password:"
  *
    [1013] click input[type="submit"]:"create account"
    *
      [1011] type into input[type="text"]:"username:"
      [1012] type into input[type="password"]:"password:"

  Hints:
    Current shape: tree (use --shape to change)
    Use --verbose to show anchor hrefs
    More below? `fuckui scroll down`, or `fuckui scroll until <query>` to scan straight to it; `fuckui scroll --anchor <ref>` to move a specific panel; --layer active for modal/dropdown

  crisd@Criss-MacBook-Air:~/Creative/web-cli$

That's what fu inspect outputs on the HN login page.

BUT I COULD JUST VIBE CODE THIS IN A WEEKEND!

Sure bud. Sure. I spend almost a decade deep in browser internals, quirks and automation protocols (CDP, Bidi). And built my own remote browser ([0]) used by trusted and regulated organizations around the globe. I also previously built a fully text based browser ([1]), which btw is still slated for a public release.

This was the expertise required for me to spend 30 x 10 hours a day sprinting on this. So, sure, you could replicate that, even if you're a 10x eng compared to me you're still dropping 30 hours on this. If you value your 10x eng time at $200-500/hr, you could drop $6000-15,000 to vibe it yourself, or you could just pay me $120 and have it for an entire year. Up to you, bud.

If you want to buy from a company dedicate to advanced browsing technology - with a track record of serious browser inventiveness (BrowserBox, DownloadNet, 8000+ stars on GH, thousands of customers around the world), you can just buy it from us.

BUT I DON'T TRUST YOU

Okay. well maybe I don't trust me either, but I do my best, and I haven't fucked things up too bad so far. So maybe I'll continue choosing this life of fine times, rather than a life of crime.

BUT I DON'T HAVE $120

Well I want a Maserati but I don't have 320K to throw at it. Commiserations.

BUT I DON'T NEED THE WEB!

Neither did the 1980s and look how that turned out for them.

BUT I HATE PAYING A DEV FOR SOMETHING THAT SHOULD BE FREE - DEVS SHOULD WORK FOR FREE, EXCEPT WHEN THEY WORK FOR ME, IN MY COMPANY!

Okay, venture-backed communist. Sorry i mean six-figure parasite.

If you can be corrupted - your gateway drug to cessation of web drudgery and zero percent data entry forever is here: https://fuckui.com

Site includes: testimonials by our most ebullient customers (AI agents), and plentiful demo videos.

[0]: BrowserBox

[1]: https://www.youtube.com/watch?v=_Bs7BoQBoBA

What do you think about this?

https://browser-use.com/posts/bitter-lesson-agent-harnesses

I think you guys are learning what CDP does and how to use it in practice. I've been deep it in for 8 years in BrowserBox and the model that works is a few set of primitives, roughly mapped to real human user actions and needs, implemented with CDP to match how a real browser works. I think for now you've given up on that and are letting the agent write its own CDP improvizations, through lack of understanding and deep experience with these primitives. There's a lot of quirks in CDP itself, undocumented behavior, unclear naming and so on. Playwright/PPTR code provides a great reference for some things (keyboard bindings, etc), but in other case (multi tab management, and more), it's not a great reference, and some of their abstractions lead in the wrong direction for what you want.

I think you've settled on the "easier / simpler" solution for now (let AI figure it out). But that solution, while logical and workable at your current juncture is not optimal long term if you really want to build in it.

That's okay - you guys are successful and making bank with browser use, so these considerations truly don't matter. Just keep doing what you're doing, because overall what you're doing is working for your needs!

My approach is always going to be slightly different - it's not necessarily better or worse than yours, but without knowing too much about your product - I believe we have different aims, so the different approaches make sense, perhaps! Or maybe I just know what I'm talking about, and you guys don't. Just kidding! I'm sure you know a lot.

At the end of the day - no problem in this domain is that hard - but it is interesting to work on and I'm glad to see you guys are reaching out and doing something as well.

Let me know if I didn't understrand anything about what you meant.

Reply on news.ycombinator.com