Obaid

yes i’ve hacked paxel. this is a public disclosure in real-time, with a tool anyone in the world can use to make their analysis top 1%, and @ycombinator i’ve indicated basic cryptography that is missing and can be used to patch this [me@obaid.wtf] lemme tell you the full story

@wtfobaid @ycombinator @sdianahu @garrytan @dflieb @gustaf @t_blom @bosmeny Impressive hacking! Sorry for all the hopefuls out there, but we have patched the bug :). Obaid - thanks for reporting the issue. Look forward to seeing you at Startup School in SF in two weeks!

@snowmaker @ycombinator @sdianahu @garrytan @dflieb @gustaf @t_blom @bosmeny Thanks Jared :) Hope so as well

and not only does it accept it, but along with hidden scoring parameters used by yc’s backend, even the visible report strength of the same project uploaded through “regular” and “boosted” paxel differs wildly.

until YC fixes it. pls do @ycombinator @snowmaker @sdianahu @garrytan @dflieb @gustaf @t_blom @bosmeny

... 📌 guys wrapping this thread up because it's getting too long as it is, full post available on the blog comments and discussion will be available on twitter, hackernews, various reddit posts, linkedin. [🔗 paxel․obaid․wtf ] is the link. enjoy. share with friends.

servers (no logs, secure, powered by cloudflare ai gateway) and returns them “buffed”: i.e: increasing the score in a realistic way, while identifying qualities in your code the original LLM analysis does not do. then it uploads it, back to yc. and yc accepts it.

use as a drop-in replacement. It uses the same YC servers and runs the exact same pipeline as Paxel (code preserved bit-for-bit) but at the last step just before it uploads to YC, it intercepts the “scoring” and “notes” parameter, sending them to an LLM gateway hosted on my

there are 2 ways you can learn this. One is to continue with reading the writeup below, and the other, preferable for people like me, is in baby-steps, and by-doing. With the exact same ux as the original paxel, I’ve created a fork hosted at paxel․obaid․wtf that you can

I checked has been adding support for VSCode Copilot, so the bug hasn’t been patched (unless they quietly built a honeypot in that time, but indications in the way the code is structured tells me that isn’t likely) — so i think it’s now appropriate to make this disclosure public

it to give myself that edge? nearly 12 days later i finished my application simultaneously with a companion email to YC sent called “Disclosed a vulnerability in Paxel as part of my YC SS26 application.” No details, the body was empty.

this was sent on the 24th of Jun ⏰ I was invited to Startup School in SF 6 days later, but I didn’t get a response back for that email, and neither from the applications team where I disclosed more details. Since all this, the only update to the paxel install script

Granted, I wanted to be cool and send them panicking and looking for my application to make it stand out among the 30k+ others. but anyone who was at YC likely has enough agency to cross-reference my email against the applications database and find the detailed disclosure

wanted to know what Paxel did — like really really badly [prioritizing curiosity over all else can sometimes become procrastinatory] if it was being used to grade applications for startup school, it must be acting as a competitive ranker. and if so, why wouldn’t i try and hack

yes i’ve hacked paxel. this is a public disclosure in real-time, with a tool anyone in the world can use to make their analysis top 1%, and @ycombinator i’ve indicated basic cryptography that is missing and can be used to patch this [me@obaid.wtf] lemme tell you the full story

📌 cross-posting from my blog, but i don't want the link deboost so find it. this is all real-time. early june when I found the application form for Startup School 26’ (referred by Dhanush, shoutout: Audora S26!) part of the extreme delay in me applying was just that I really

e.g: Thinks it can run on its own time, i.e: when bash is switched to run_in_background says "I'll poll it in 3 min".

@bcherny Claude Code has not been context-fed information about its own tools (Monitor, run_in_background: true) properly. Often doesn't use them, repeatedly reuses the wrong one for the task (Monitor when a background shell would be appropriate), or doesn't know what they'll do

a huge chunk of learning how to prompt is just improving your vocabulary. Unlike humans, LLMs don't have a limited set of vocabulary so you don't need to use general-purpose, simplified adjectives to communicate. Learn specialized, accurately descriptive words, however niche