DEV Community

Jackson Ly profile picture

Jackson Ly

AI systems engineer in Germany and lead engineer at a YC-backed startup. I write about running many AI coding agents at once, and checking their work.

Location Germany Joined Joined on  Personal website https://jacksonly.xyz

Education

B.Sc. Business Administration, Frankfurt School

Work

Lead engineer at Openlaw (YC F24)

225 billion Claude Code tokens in 30 days, and the subagents that never named a model
jacksonxly profile

225 billion Claude Code tokens in 30 days, and the subagents that never named a model

#ai #claudecode #programming #agents
2
Comments 8
9 min read

Want to connect with Jackson Ly?

Create an account to connect with Jackson Ly. You can also sign in below to proceed if you already have an account.

Create Account
Already have an account? Sign in
Test the second access path before you call a metric unreadable
jacksonxly profile

Test the second access path before you call a metric unreadable

#debugging #api #programming #webdev
1
Comments
3 min read
Refreshing a token I did not own logged our account out
jacksonxly profile

Refreshing a token I did not own logged our account out

#api #oauth #debugging #webdev
Comments
4 min read
A 200 response is not a page, and your policy check is grepping an empty shell
jacksonxly profile

A 200 response is not a page, and your policy check is grepping an empty shell

#webdev #python #automation #testing
1
Comments 3
4 min read
Your agent's inbox check is measuring read state, not answered state
jacksonxly profile

Your agent's inbox check is measuring read state, not answered state

#ai #agents #api #debugging
1
Comments 4
4 min read
Durable approval is not the same as valid approval
jacksonxly profile

Durable approval is not the same as valid approval

#ai #agents #security #architecture
6
Comments
3 min read
Running local and cloud models in the same coding agent: what actually ships in 2026
jacksonxly profile

Running local and cloud models in the same coding agent: what actually ships in 2026

#ai #llm #opensource #productivity
Comments
4 min read
Your interval-triggered agent pays full price for empty runs
jacksonxly profile

Your interval-triggered agent pays full price for empty runs

#automation #architecture #python #devops
Comments
4 min read
Human in the loop is not a switch: classify the commands
jacksonxly profile

Human in the loop is not a switch: classify the commands

#agents #ai #security #python
Comments
4 min read
Your cache_read_input_tokens is zero. Here is what silently did it.
jacksonxly profile

Your cache_read_input_tokens is zero. Here is what silently did it.

#llm #ai #performance #api
1
Comments 2
5 min read
Your baseline scored 0.000. That's a broken harness, not a result.
jacksonxly profile

Your baseline scored 0.000. That's a broken harness, not a result.

#rag #llm #testing #ai
Comments 3
4 min read
Our Rust file watcher ate 23.6 GB of RAM, and our ignore rules never had a chance
jacksonxly profile

Our Rust file watcher ate 23.6 GB of RAM, and our ignore rules never had a chance

#rust #macos #performance #debugging
Comments 3
4 min read
Point any app at a local LLM on your Mac (OpenAI-compatible endpoints)
jacksonxly profile

Point any app at a local LLM on your Mac (OpenAI-compatible endpoints)

#ai #llm #mac #productivity
Comments
4 min read
MLX vs GGUF on Apple Silicon: which local LLM format should you actually use?
jacksonxly profile

MLX vs GGUF on Apple Silicon: which local LLM format should you actually use?

#ai #machinelearning #llm #apple
Comments 1
4 min read
What belongs in an agent's context window (and what to evict)
jacksonxly profile

What belongs in an agent's context window (and what to evict)

#ai #machinelearning #programming #llm
3
Comments 17
4 min read
What poisoning a RAG store taught us about agent memory
jacksonxly profile

What poisoning a RAG store taught us about agent memory

#ai #rag #security #machinelearning
1
Comments 2
7 min read
Why I built my Mac assistant to run 100% on-device (and what local-first actually cost me)
jacksonxly profile

Why I built my Mac assistant to run 100% on-device (and what local-first actually cost me)

#showdev #buildinpublic #privacy #ai
1
Comments 9
4 min read
loading...