Christopher Yu

Full-stack engineer. I build SaaS end to end, schema to deploy.
5 yrs production, 3 in fintech. Bali, GMT+8. Two slots for Q4.

[Christopher Yu @christopheryuu_]

For a while now I've been posting the SaaS I build

Same thing, new direction: I'm building them for other people
Schema to deploy, one person, no handoffs

Five years in production, three of them in fintech

[Christopher Yu @christopheryuu_]

@AnthonyCastrio Software Engineer here 👋
https://t.co/fgzYU4Aqy5
Would appreciate any leads!

[Christopher Yu @christopheryuu_]

@marclou How do you know this tho ? Is there a pattern you’ve noticed from growing audiences, or is it more about AI-generated content feeling less personal ?

[Christopher Yu @christopheryuu_]

@shownotover No, but it does make me feel better about treating it like an unpaid intern

[Christopher Yu @christopheryuu_]

@dev_niladri @X Let's connect !

[Christopher Yu @christopheryuu_]

@Zohairakhn_ Interested! Is this a new build or an existing product that needs frontend work? I work with React/Next.js and would love to learn more.

Portfolio: https://t.co/fgzYU4Aqy5

[Christopher Yu @christopheryuu_]

@danielkleach This is why I started keeping an idea backlog.
Not every problem needs a solution from me right now. The hard part is finding the 1 idea worth committing to

[Christopher Yu @christopheryuu_]

@konig0000 Interested! I’m a frontend/full-stack developer working with React & Next.js. Would love to hear more about the platform, stack, and timeline.

Portfolio: https://t.co/fgzYU4Aqy5

[Christopher Yu @christopheryuu_]

@PythonDvz Typescript, fav language

[Christopher Yu @christopheryuu_]

@shubh19 PostgreSQL.

After working with different stacks, I've found it gives the best default experience: relational modeling, JSON support, extensions, strong tooling, and fewer surprises as the app grows

[Christopher Yu @christopheryuu_]

@ishtwts Play CTF

[Christopher Yu @christopheryuu_]

Five years ago I played CTF competitively. One challenge cost me an entire night, and it came down to a single syscall.

Final round, binary exploitation. Seccomp had the process locked to eight syscalls and execve wasn't one of them, so the usual ending of popping a shell was off the table. I had to open, read and print the flag file myself.

Getting to that point took most of the night. libc 2.32 XORs heap pointers against the address of the chunk holding them, so you can't forge one until you've leaked the heap and recovered the key. Then a libc leak and a stack leak to beat ASLR, then tcache poisoning to turn all of that into an arbitrary write, then a ROP chain onto the stack.

All of it worked. Then I called open() to read the flag, and it died. Every single time. And the whitelist plainly said open was allowed.

libc's open() doesn't issue the open syscall. It issues openat. Seccomp was killing openat, so the function I was permitted to call was never the instruction that ran. The fix was to stop asking libc for anything and assemble the syscall by hand: pop 2 into rax, set the arguments, hit syscall.

I don't break software for a living any more, I build it. But that night is the habit I kept: what a function is called and what the machine does are separate facts, and only one of them is enforced.

[Christopher Yu @christopheryuu_]

Full write-up from the time:

https://t.co/XPERYU25md

[Christopher Yu @christopheryuu_]

Five years ago I played CTF competitively. One challenge cost me an entire night, and it came down to a single syscall.

Final round, binary exploitation. Seccomp had the process locked to eight syscalls and execve wasn't one of them, so the usual ending of popping a shell was off the table. I had to open, read and print the flag file myself.

Getting to that point took most of the night. libc 2.32 XORs heap pointers against the address of the chunk holding them, so you can't forge one until you've leaked the heap and recovered the key. Then a libc leak and a stack leak to beat ASLR, then tcache poisoning to turn all of that into an arbitrary write, then a ROP chain onto the stack.

All of it worked. Then I called open() to read the flag, and it died. Every single time. And the whitelist plainly said open was allowed.

libc's open() doesn't issue the open syscall. It issues openat. Seccomp was killing openat, so the function I was permitted to call was never the instruction that ran. The fix was to stop asking libc for anything and assemble the syscall by hand: pop 2 into rax, set the arguments, hit syscall.

I don't break software for a living any more, I build it. But that night is the habit I kept: what a function is called and what the machine does are separate facts, and only one of them is enforced.

[Christopher Yu @christopheryuu_]

@HeyAliux AI can write code.
Developers make sure it's the right code.

Architecture, security, scalability, debugging, and knowing what not to build still require experience 😁

[Christopher Yu @christopheryuu_]

@TomTurcotteTech I’d love that!

Thanks so much for offering. I’m just starting out, so any insights on building an audience and finding clients through X would be incredibly valuable. Feel free to send me a DM, and we can sort out a time that works

[Christopher Yu @christopheryuu_]

For a while now I've been posting the SaaS I build

Same thing, new direction: I'm building them for other people
Schema to deploy, one person, no handoffs

Five years in production, three of them in fintech

[Christopher Yu @christopheryuu_]

@TomTurcotteTech Not yet,
I’m still in the early stages. I’ve mainly been a fullstack engineer, so this is my first time actively trying to find clients through X.

Have you had much success finding clients here? Any advice would be appreciated! 😁🙏

[Christopher Yu @christopheryuu_]

@omkarships Getting users. Building is the easy part. Getting peoplecto care is a completely different skill

[Christopher Yu @christopheryuu_]

@VibeCdr Fullstaxk + devops 🔥🔥

[Christopher Yu @christopheryuu_]

https://t.co/KUz93bkLcq