Hello from Must Love LLaMas
Sep 1, 2026 · 1 min read
A separate tech blog, MDX when I need it, and a URL I can actually send people.
Tech notes I want to be able to send someone without also sending them my whole personal blog.
don't get high on your own supply
Jul 16, 2026 · 14 min read
An agent-driven Instagram pipeline keeps broadcasting separate from the feed by combining deterministic tools, bounded judgment, and local memory.
good ideas finally clear the bar
Mar 22, 2026 · 5 min read
AI does not erase work. It moves the bottleneck and lowers the bar for neglected projects.
stop letting your llm wing it
Mar 12, 2026 · 2 min read
If the query already works, make the model call it. Do not let it rewrite SQL every morning.
more tools, less trust
Mar 5, 2026 · 3 min read
Fast software creates overlapping tools, conflicting calculations, and a trust problem that another tool cannot fix.
from engineer to leader: stop fixing everything
Feb 28, 2026 · 3 min read
The instinct that made you a good engineer will bruise people if you lead with it.
the wrapper isn't the revolution
Feb 14, 2026 · 5 min read
ChatGPT is a wrapper. The model is the shift. Build on the engine.
attention doesn't parallelize
Jan 21, 2026 · 2 min read
Running many AI agents at once moves the bottleneck from execution to evaluation and fractures the attention needed for deep work.
write it once, click click boom
Jan 12, 2026 · 2 min read
You do not scale. The system you build once does.
you don't have to understand it before you do it
Nov 28, 2025 · 6 min read
Low-risk experiments with Cursor, context windows, and monorepos often teach more than theory alone.