Internet programming attacks through the ages
80s: hehe...I pretended to be Bob on IRC
90s: 1k people went to my website at once and the server caught on fire
00s: PHP/ActiveX/Flash - we make tech that's basically designed to be hacked
10s: one of these 10k nodejs dependencies got hijacked, glhf figuring out which one
20s: the LLM got prompted-injected via emojis
The Real Engineering Career Ladder
Jr - how do I even code?
Mid - I can code!!!!! Code EVERYWHERE!!!
Sr - how do I solve the same problem with less code?
Staff - delete like 90% of this code
Principal - I miss coding
If you want to be a 10x engineer:
- Share your ideas freely with little concern for credit
- Default to openness
- Hear another team member all the way out, especially when you disagree
- Fail together, not aloneβno one learns if you keep it secret
- Message where you're going as much as current location
The 10x engineer lifts up their team to get 10x things done.
Finally, I had to reach down and just tell it: copy the file from root to each service. Is it elegant? No. Did it stop the errors from creeping in from having the other random things it tried. Yes.
It proposed to me all other kinds of elaborate solutions: lazy loading, multiple .env file references, etc. I accepted them and then it would drift back to multiple .env files.
3/x
To be clear, I didn't spend 48 hours in a row trying to fix this, nor even 1 hour.
It's something that kept coming up in its behavior that it would drift to as a default over time.
It viscerally highlighted things I already knew:
1. LLMs really can't remember
2. LLMs cannot "think"
3. LLMs cannot "know"
4. LLMs cannot learn because of 2 & 3
They are best simulate the appears of all of these.
It still remains on you.
If you don't know where you're going, you will get there.
Claude kept trying make multiple env files for each service. No matter what I did.
So what did I do?
I'm using claude-mem, beads, and several other skills to attempt to improve alignment and long term memory.
I had it make a plan. I had it store that plan. I had it make tasks in beads about that plan.
I even eventually started ADRs to describe the decisions I made.
Didn't matter what I did. It kept trying to make multiple .env files for each service.
2/x
Finally, I had to reach down and just tell it: copy the file from root to each service. Is it elegant? No. Did it stop the errors from creeping in from having the other random things it tried. Yes.
It proposed to me all other kinds of elaborate solutions: lazy loading, multiple .env file references, etc. I accepted them and then it would drift back to multiple .env files.
3/x
I had an interesting experience over the past 48 hours with LLMs.
I have a little side project I'm working on, and it's in a monorep. I wanted to be able to put a .env file at root and then have each service reference it.
Claude code repeatedly choked on this constraint.
How it choked is the interesting part.
1/x
Claude kept trying make multiple env files for each service. No matter what I did.
So what did I do?
I'm using claude-mem, beads, and several other skills to attempt to improve alignment and long term memory.
I had it make a plan. I had it store that plan. I had it make tasks in beads about that plan.
I even eventually started ADRs to describe the decisions I made.
Didn't matter what I did. It kept trying to make multiple .env files for each service.
2/x
I had an interesting experience over the past 48 hours with LLMs.
I have a little side project I'm working on, and it's in a monorep. I wanted to be able to put a .env file at root and then have each service reference it.
Claude code repeatedly choked on this constraint.
How it choked is the interesting part.
1/x
I have used ChatGPT
EM: let's clarify the ACs with the PM
UI Designer: the glass is a user hostile pattern
Product Manager: glass size may be a perception issue; let's A/B test some fake doors
2/
Infra: I pulled down a glass from the artifact registry, and it held water just fine
QA: I can confirm the water looks like it's halfway up; I also tested it with a duck, and the glass broke, so that may be a bug
Security engineer: the glass has a slight green tint; it may have vulnerabilities if it was made during the 30s when glass had uranium in it. We should isolate the glass to prevent users from interacting with it.
3/
The glass is ....
Optimist: half full
Pessimist: half empty
Backend engineer: this is a frontend problem
Frontend engineer: I just put in the water that the API gave me
Fullstack: the glass meets the spec I was given.
1/
EM: let's clarify the ACs with the PM
UI Designer: the glass is a user hostile pattern
Product Manager: glass size may be a perception issue; let's A/B test some fake doors
2/